first commit

Close #772, #1128

.gitignore

@@ -51,3 +51,9 @@ scm-source.json
 docs/build/
 docs/source/_static/
 docs/source/_templates/
+
+# Pycharm IDE
+.idea/
+
+#VSCode IDE
+.vscode/

.travis.yml

@@ -1,17 +1,32 @@
-sudo: false
+sudo: true
 dist: trusty
 language: python
-python:
+addons:
-- "3.4"  # 2.7 and 3.5 are preinstalled by default
+  apt:
+    packages:
+      - expect-dev # for unbuffer
 env:
   global:
-  - ETCDVERSION=3.0.17 ZKVERSION=3.4.9 CONSULVERSION=0.7.4
+  - ETCDVERSION=3.0.17 ZKVERSION=3.4.11 CONSULVERSION=0.7.4
-  - PYVERSIONS="2.7 3.4 3.5"
+  - PYVERSIONS="2.7 3.5 3.6"
-  matrix:
+  - EXCLUDE_BEHAVE="3.5"
-  - TEST_SUITE="python setup.py"
+  - BOTO_CONFIG=/doesnotexist
-  - DCS="etcd" TEST_SUITE="behave"
+matrix:
-  - DCS="exhibitor" TEST_SUITE="behave"
+  include:
-  - DCS="consul" TEST_SUITE="behave"
+  - python: "3.5"
+    env: TEST_SUITE="python setup.py"
+  - python: "3.6"
+    env: DCS="etcd" TEST_SUITE="behave"
+  - python: "3.6"
+    env: DCS="exhibitor" TEST_SUITE="behave"
+  - python: "3.6"
+    env: DCS="consul" TEST_SUITE="behave"
+  - python: "3.6"
+    env: DCS="kubernetes" TEST_SUITE="behave"
+branches:
+  only:
+  - master
+  - /^v\d+\.\d+(\.\d+)?$/
 cache:
   directories:
   - $HOME/mycache
@@ -19,7 +34,7 @@ before_cache:
 - |
   rm -fr $HOME/mycache/python*
   for pv in $PYVERSIONS; do
-      if [[ $TEST_SUITE != "behave" || $pv != "3.4" ]]; then
+      if [[ $TEST_SUITE != "behave" || $pv != $EXCLUDE_BEHAVE ]]; then
           fpv=$(basename $(readlink $HOME/virtualenv/python${pv}))
           mv $HOME/virtualenv/${fpv} $HOME/mycache/${fpv}
       fi
@@ -51,6 +66,26 @@ install:
             ln -s $EC etcd
         }
 
+        function get_kubernetes() {
+            wget -O localkube "https://storage.googleapis.com/minikube/k8sReleases/v1.7.0/localkube-linux-amd64"
+            chmod +x localkube
+            sudo nohup ./localkube --logtostderr=true --enable-dns=false > localkube.log 2>&1 &
+
+            echo "Waiting for localkube to start..."
+            if ! timeout 120 sh -c "while ! curl -ks http://127.0.0.1:8080/ >/dev/null; do sleep 1; done"; then
+                sudo cat localkube.log
+                echo "localkube did not start"
+                exit 1
+            fi
+            echo "Check certificate permissions"
+            sudo chmod 644 /var/lib/localkube/certs/*
+            sudo ls -altr /var/lib/localkube/certs/
+
+            echo "Set up .kube/config"
+            mkdir ~/.kube
+            echo -e "apiVersion: v1\nclusters:\n- cluster:\n    certificate-authority: /var/lib/localkube/certs/ca.crt\n    server: https://127.0.0.1:8443\n  name: local\ncontexts:\n- context:\n    cluster: local\n    user: myself\n  name: local\ncurrent-context: local\nkind: Config\npreferences: {}\nusers:\n- name: myself\n  user:\n    client-certificate: /var/lib/localkube/certs/apiserver.crt\n    client-key: /var/lib/localkube/certs/apiserver.key\n" > ~/.kube/config
+        }
+
         function get_exhibitor() {
             ZC=~/mycache/zookeeper-${ZKVERSION}
             if [[ ! -d $ZC ]]; then
@@ -76,7 +111,7 @@ install:
     fi
 
     for pv in $PYVERSIONS; do
-        if [[ $TEST_SUITE != "behave" || $pv != "3.4" ]]; then
+        if [[ $TEST_SUITE != "behave" || $pv != $EXCLUDE_BEHAVE ]]; then
             fpv=$(basename $(readlink $HOME/virtualenv/python$pv))
             if [[ -d ~/mycache/${fpv} ]]; then
                 mv ~/virtualenv/${fpv} ~/virtualenv/${fpv}.bckp
@@ -84,7 +119,7 @@ install:
             fi
             source ~/virtualenv/python${pv}/bin/activate
             # explicitly install all needed python modules to cache them
-            for p in '-r requirements.txt' 'behave codacy-coverage coverage coveralls flake8 mock pytest-cov pytest setuptools'; do
+            for p in '-r requirements.txt' 'psycopg2-binary behave codacy-coverage coverage coveralls flake8 mock pytest-cov pytest setuptools'; do
                 pip install $p --upgrade
             done
         fi
@@ -92,17 +127,21 @@ install:
 script:
   - |
     for pv in $PYVERSIONS; do
+        if [[ $TEST_SUITE == "behave" && $pv == $EXCLUDE_BEHAVE ]]; then
+            continue
+        fi
+
         source ~/virtualenv/python${pv}/bin/activate
 
         if [[ $TEST_SUITE != "behave" ]]; then
             echo Running unit tests using python${pv}
-            $TEST_SUITE test
+            unbuffer $TEST_SUITE test
             $TEST_SUITE flake8
-        elif [[ $pv != "3.4" ]]; then
+        elif [[ $pv != $EXCLUDE_BEHAVE ]]; then
             echo Running acceptance tests using python${pv}
-            if ! PATH=.:/usr/lib/postgresql/9.6/bin:$PATH $TEST_SUITE; then
+            if ! PATH=.:/usr/lib/postgresql/9.6/bin:$PATH unbuffer $TEST_SUITE; then
                 # output all log files when tests are failing
-                grep . features/output/*/*postgres?.*
+                grep . features/output/*_failed/*postgres?.*
                 exit 1
             fi
         fi
@@ -110,5 +149,9 @@ script:
 
     set +e
 after_success:
+  # before_cache is executed earlier than after_success, so we need to restore one of virtualenv directories
+  - fpv=$(basename $(readlink $HOME/virtualenv/python3.6)) && mv $HOME/mycache/${fpv} $HOME/virtualenv/${fpv}
   - coveralls
   - if [[ $TEST_SUITE != "behave" ]]; then python-codacy-coverage -r coverage.xml; fi
+  - if [[ $DCS == "exhibitor" ]]; then ~/mycache/zookeeper-${ZKVERSION}/bin/zkServer.sh stop; fi
+  - sudo kill $(jobs -p)

.zappr.yaml

@@ -1,13 +0,0 @@
-# for github.com
-approvals:
-  groups:
-    zalando:
-      minimum: 2
-      from:
-        orgs:
-          - "zalando"
-# team should be valid team id in team service https://teams.auth.zalando.com/api/teams/:id
-X-Zalando-Team: "acid"
-# type should be one of [code, doc, config, tools, secrets]
-# code will be the default value, if X-Zalando-Type is not found in .zappr.yml
-X-Zalando-Type: code

Dockerfile

@@ -1,59 +1,156 @@
 ## This Dockerfile is meant to aid in the building and debugging patroni whilst developing on your local machine
 ## It has all the necessary components to play/debug with a single node appliance, running etcd
-FROM ubuntu:16.04
+ARG PG_MAJOR=10
-MAINTAINER Alexander Kukushkin <[email protected]>
+ARG COMPRESS=false
+ARG PGHOME=/home/postgres
+ARG PGDATA=$PGHOME/data
+ARG LC_ALL=C.UTF-8
+ARG LANG=C.UTF-8
 
-RUN echo 'APT::Install-Recommends "0";' > /etc/apt/apt.conf.d/01norecommend \
+FROM postgres:$PG_MAJOR as builder
-    && echo 'APT::Install-Suggests "0";' >> /etc/apt/apt.conf.d/01norecommend
 
-ENV PGVERSION 9.6
+ARG PGHOME
-ENV PATH /usr/lib/postgresql/${PGVERSION}/bin:$PATH
+ARG PGDATA
-RUN apt-get update -y \
+ARG LC_ALL
-    && apt-get upgrade -y \
+ARG LANG
-    && apt-get install -y curl jq haproxy zookeeper python-psycopg2 python-yaml python-requests \
-        python-six python-click python-dateutil python-tzlocal python-urllib3 python-dnspython \
-        python-pip python-setuptools python-kazoo python-prettytable python-wheel python \
 
-    && export DISTRIB_CODENAME=$(sed -n 's/DISTRIB_CODENAME=//p' /etc/lsb-release) \
+ENV ETCDVERSION=2.3.8 CONFDVERSION=0.16.0
-    && echo "deb http://apt.postgresql.org/pub/repos/apt/ ${DISTRIB_CODENAME}-pgdg main" > /etc/apt/sources.list.d/pgdg.list \
-    && echo "deb-src http://apt.postgresql.org/pub/repos/apt/ ${DISTRIB_CODENAME}-pgdg main" >> /etc/apt/sources.list.d/pgdg.list \
-    && curl -s -o - https://www.postgresql.org/media/keys/ACCC4CF8.asc | apt-key add - \
-
-    ## Make sure we have a en_US.UTF-8 locale available
-    && localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8 \
 
+RUN set -ex \
+    && export DEBIAN_FRONTEND=noninteractive \
+    && echo 'APT::Install-Recommends "0";\nAPT::Install-Suggests "0";' > /etc/apt/apt.conf.d/01norecommend \
     && apt-get update -y \
-    && apt-get install -y postgresql-contrib-${PGVERSION} \
+    # postgres:10 is based on debian, which has the patroni package. We will install all required dependencies
-
+    && apt-cache depends patroni | sed -n -e 's/.*Depends: \(python3-.\+\)$/\1/p' \
-    # Remove the default cluster, which Debian stupidly starts right after installation of the packages
+            | grep -Ev '^python3-(sphinx|etcd|consul|kazoo|kubernetes)' \
-    && pg_dropcluster --stop ${PGVERSION} main \
+            | xargs apt-get install -y vim curl less jq locales haproxy sudo \
-    && pip install python-etcd==0.4.3 python-consul==0.6.1 --upgrade \
+                            python3-etcd python3-kazoo python3-pip busybox \
-
+    && pip3 install dumb-init \
-    # Clean up
+\
-    && apt-get remove -y python-pip python-setuptools \
+    # Cleanup all locales but en_US.UTF-8
+    && find /usr/share/i18n/charmaps/ -type f ! -name UTF-8.gz -delete \
+    && find /usr/share/i18n/locales/ -type f ! -name en_US ! -name en_GB ! -name i18n ! -name iso14651_t1 ! -name iso14651_t1_common ! -name 'translit_*' -delete \
+    && echo 'en_US.UTF-8 UTF-8' > /usr/share/i18n/SUPPORTED \
+\
+    # Make sure we have a en_US.UTF-8 locale available
+    && localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8 \
+\
+    # haproxy dummy config
+    && echo 'global\n        stats socket /run/haproxy/admin.sock mode 660 level admin' > /etc/haproxy/haproxy.cfg \
+\
+    # vim config
+    && echo 'syntax on\nfiletype plugin indent on\nset mouse-=a\nautocmd FileType yaml setlocal ts=2 sts=2 sw=2 expandtab' > /etc/vim/vimrc.local \
+\
+    # Prepare postgres/patroni/haproxy environment
+    && mkdir -p $PGHOME/.config/patroni /patroni /run/haproxy \
+    && ln -s ../../postgres0.yml $PGHOME/.config/patroni/patronictl.yaml \
+    && ln -s /patronictl.py /usr/local/bin/patronictl \
+    && sed -i "s|/var/lib/postgresql.*|$PGHOME:/bin/bash|" /etc/passwd \
+    && chown -R postgres:postgres /var/log \
+\
+    # Download etcd
+    && curl -sL https://github.com/coreos/etcd/releases/download/v${ETCDVERSION}/etcd-v${ETCDVERSION}-linux-amd64.tar.gz \
+            | tar xz -C /usr/local/bin --strip=1 --wildcards --no-anchored etcd etcdctl \
+\
+    # Download confd
+    && curl -sL https://github.com/kelseyhightower/confd/releases/download/v${CONFDVERSION}/confd-${CONFDVERSION}-linux-amd64 \
+            > /usr/local/bin/confd && chmod +x /usr/local/bin/confd \
+\
+    # Clean up all useless packages and some files
+    && apt-get purge -y --allow-remove-essential python3-pip gzip bzip2 util-linux e2fsprogs \
+                libmagic1 bsdmainutils login ncurses-bin libmagic-mgc e2fslibs bsdutils \
+                exim4-config gnupg-agent dirmngr libpython2.7-stdlib libpython2.7-minimal \
     && apt-get autoremove -y \
     && apt-get clean -y \
-    && rm -rf /var/lib/apt/lists/* /root/.cache
+    && rm -rf /var/lib/apt/lists/* \
+        /root/.cache \
+        /var/cache/debconf/* \
+        /etc/rc?.d \
+        /etc/systemd \
+        /docker-entrypoint* \
+        /sbin/pam* \
+        /sbin/swap* \
+        /sbin/unix* \
+        /usr/local/bin/gosu \
+        /usr/sbin/[acgipr]* \
+        /usr/sbin/*user* \
+        /usr/share/doc* \
+        /usr/share/man \
+        /usr/share/info \
+        /usr/share/i18n/locales/translit_hangul \
+        /usr/share/locale/?? \
+        /usr/share/locale/??_?? \
+        /usr/share/postgresql/*/man \
+        /usr/share/postgresql-common/pg_wrapper \
+        /usr/share/vim/vim80/doc \
+        /usr/share/vim/vim80/lang \
+        /usr/share/vim/vim80/tutor \
+#        /var/lib/dpkg/info/* \
+    && find /usr/bin -xtype l -delete \
+    && find /var/log -type f -exec truncate --size 0 {} \; \
+    && find /usr/lib/python3/dist-packages -name '*test*' | xargs rm -fr \
+    && find /lib/x86_64-linux-gnu/security -type f ! -name pam_env.so ! -name pam_permit.so ! -name pam_unix.so -delete
 
-ENV ETCDVERSION 3.0.15
+# perform compression if it is necessary
-RUN curl -L https://github.com/coreos/etcd/releases/download/v${ETCDVERSION}/etcd-v${ETCDVERSION}-linux-amd64.tar.gz \
+ARG COMPRESS
-    | tar xz -C /usr/local/bin --strip=1 --wildcards --no-anchored etcd etcdctl
+RUN if [ "$COMPRESS" = "true" ]; then \
+        set -ex \
+        # Allow certain sudo commands from postgres
+        && echo 'postgres ALL=(ALL) NOPASSWD: /bin/tar xpJf /a.tar.xz -C /, /bin/rm /a.tar.xz, /bin/ln -snf dash /bin/sh' >> /etc/sudoers \
+        && ln -snf busybox /bin/sh \
+        && files="/bin/sh /usr/bin/sudo /usr/lib/sudo/sudoers.so /lib/x86_64-linux-gnu/security/pam_*.so" \
+        && libs="$(ldd $files | awk '{print $3;}' | grep '^/' | sort -u) /lib/x86_64-linux-gnu/ld-linux-x86-64.so.* /lib/x86_64-linux-gnu/libnsl.so.* /lib/x86_64-linux-gnu/libnss_compat.so.*" \
+        && (echo /var/run $files $libs | tr ' ' '\n' && realpath $files $libs) | sort -u | sed 's/^\///' > /exclude \
+        && find /etc/alternatives -xtype l -delete \
+        && save_dirs="usr lib var bin sbin etc/ssl etc/init.d etc/alternatives etc/apt" \
+        && XZ_OPT=-e9v tar -X /exclude -cpJf a.tar.xz $save_dirs \
+        # we call "cat /exclude" to avoid including files from the $save_dirs that are also among
+        # the exceptions listed in the /exclude, as "uniq -u" eliminates all non-unique lines.
+        # By calling "cat /exclude" a second time we guarantee that there will be at least two lines
+        # for each exception and therefore they will be excluded from the output passed to 'rm'.
+        && /bin/busybox sh -c "(find $save_dirs -not -type d && cat /exclude /exclude && echo exclude) | sort | uniq -u | xargs /bin/busybox rm" \
+        && /bin/busybox --install -s \
+        && /bin/busybox sh -c "find $save_dirs -type d -depth -exec rmdir -p {} \; 2> /dev/null"; \
+    fi
 
-ENV CONFDVERSION 0.11.0
+FROM scratch
-RUN curl -L https://github.com/kelseyhightower/confd/releases/download/v${CONFDVERSION}/confd-${CONFDVERSION}-linux-amd64 > /usr/local/bin/confd \
+COPY --from=builder / /
-    && chmod +x /usr/local/bin/confd
 
-ADD patronictl.py patroni.py docker/entrypoint.sh /
+LABEL maintainer="Alexander Kukushkin <[email protected]>"
-ADD patroni /patroni/
-ADD extras/confd /etc/confd
-RUN ln -s /patronictl.py /usr/local/bin/patronictl
 
-### Setting up a simple script that will serve as an entrypoint
+ARG PG_MAJOR
-RUN mkdir /data/ && touch /pgpass /patroni.yml \
+ARG COMPRESS
-    && chown postgres:postgres -R /patroni/ /data/ /pgpass /patroni.yml /etc/haproxy /var/run/ /var/lib/ /var/log/ \
+ARG PGHOME
-    && echo 1 > /etc/zookeeper/conf/myid
+ARG PGDATA
+ARG LC_ALL
+ARG LANG
 
-EXPOSE 2379 5432 8008
+ARG PGBIN=/usr/lib/postgresql/$PG_MAJOR/bin
+
+ENV LC_ALL=$LC_ALL LANG=$LANG EDITOR=/usr/bin/editor
+ENV PGDATA=$PGDATA PATH=$PATH:$PGBIN
+
+COPY patroni /patroni/
+COPY extras/confd/conf.d/haproxy.toml /etc/confd/conf.d/
+COPY extras/confd/templates/haproxy.tmpl /etc/confd/templates/
+COPY patroni*.py docker/entrypoint.sh /
+COPY postgres?.yml $PGHOME/
+
+WORKDIR $PGHOME
+
+RUN sed -i 's/env python/&3/' /patroni*.py \
+    # "fix" patroni configs
+    && sed -i 's/^\(  connect_address:\|  - host\)/#&/' postgres?.yml \
+    && sed -i 's/^  listen: 127.0.0.1/  listen: 0.0.0.0/' postgres?.yml \
+    && sed -i "s|^\(  data_dir: \).*|\1$PGDATA|" postgres?.yml \
+    && sed -i "s|^#\(  bin_dir: \).*|\1$PGBIN|" postgres?.yml \
+    && sed -i 's/^  - encoding: UTF8/  - locale: en_US.UTF-8\n&/' postgres?.yml \
+    && sed -i 's/^\(scope\|name\|etcd\|  host\|  authentication\|  pg_hba\|  parameters\):/#&/' postgres?.yml \
+    && sed -i 's/^    \(replication\|superuser\|rewind\|unix_socket_directories\|\(\(  \)\{0,1\}\(username\|password\)\)\):/#&/' postgres?.yml \
+    && sed -i 's/^      parameters:/      pg_hba:\n      - local all all trust\n      - host replication all all md5\n      - host all all all md5\n&\n        max_connections: 100/'  postgres?.yml \
+    && if [ "$COMPRESS" = "true" ]; then chmod u+s /usr/bin/sudo; fi \
+    && chown -R postgres:postgres $PGHOME /run /etc/haproxy
 
-ENTRYPOINT ["/bin/bash", "/entrypoint.sh"]
 USER postgres
+
+ENTRYPOINT ["/bin/sh", "/entrypoint.sh"]

README.rst

@@ -1,27 +1,27 @@
 |Build Status| |Coverage Status|
 
 Patroni: A Template for PostgreSQL HA with ZooKeeper, etcd or Consul
-------------------------------------------------------------
+--------------------------------------------------------------------
 
 You can find a version of this documentation that is searchable and also easier to navigate at `patroni.readthedocs.io <https://patroni.readthedocs.io>`__.
 
 
 There are many ways to run high availability with PostgreSQL; for a list, see the `PostgreSQL Documentation <https://wiki.postgresql.org/wiki/Replication,_Clustering,_and_Connection_Pooling>`__.
 
-Patroni is a template for you to create your own customized, high-availability solution using Python and - for maximum accessibility - a distributed configuration store like `ZooKeeper <https://zookeeper.apache.org/>`__, `etcd <https://github.com/coreos/etcd>`__ or `Consul <https://github.com/hashicorp/consul>`__. Database engineers, DBAs, DevOps engineers, and SREs who are looking to quickly deploy HA PostgreSQL in the datacenter-or anywhere else-will hopefully find it useful.
+Patroni is a template for you to create your own customized, high-availability solution using Python and - for maximum accessibility - a distributed configuration store like `ZooKeeper <https://zookeeper.apache.org/>`__, `etcd <https://github.com/coreos/etcd>`__, `Consul <https://github.com/hashicorp/consul>`__ or `Kubernetes <https://kubernetes.io>`__. Database engineers, DBAs, DevOps engineers, and SREs who are looking to quickly deploy HA PostgreSQL in the datacenter-or anywhere else-will hopefully find it useful.
 
 We call Patroni a "template" because it is far from being a one-size-fits-all or plug-and-play replication system. It will have its own caveats. Use wisely.
 
-**Note to Kubernetes users**: We're currently developing Patroni to be as useful as possible for teams running Kubernetes on top of Google Compute Engine; Patroni can be the HA solution for Postgres in such an environment. To this end, there is a `Helm chart <https://github.com/kubernetes/charts/tree/master/incubator/patroni>`__ that uses Patroni and  `Spilo  <https://github.com/zalando/spilo/>`__ to provision a five-node PostgreSQL HA cluster in a Kubernetes+GCE environment. (The Helm chart deploys Spilo Docker images, not just "bare" Patroni.)
+**Note to Kubernetes users**: Patroni can run natively on top of Kubernetes. Take a look at the `Kubernetes <https://github.com/zalando/patroni/blob/master/docs/kubernetes.rst>`__ chapter of the Patroni documentation.
 
 .. contents::
     :local:
     :depth: 1
     :backlinks: none
 
-==============
+=================
 How Patroni Works
-==============
+=================
 
 Patroni originated as a fork of `Governor <https://github.com/compose/governor>`__, the project from Compose. It includes plenty of new features.
 
@@ -29,33 +29,97 @@ For an example of a Docker-based deployment with Patroni, see `Spilo <https://gi
 
 For additional background info, see:
 
+* `Elephants on Automatic: HA Clustered PostgreSQL with Helm <https://www.youtube.com/watch?v=CftcVhFMGSY>`_, talk by Josh Berkus and Oleksii Kliukin at KubeCon Berlin 2017
 * `PostgreSQL HA with Kubernetes and Patroni <https://www.youtube.com/watch?v=iruaCgeG7qs>`__, talk by Josh Berkus at KubeCon 2016 (video)
 * `Feb. 2016 Zalando Tech blog post <https://tech.zalando.de/blog/zalandos-patroni-a-template-for-high-availability-postgresql/>`__
 
-================
+==================
 Development Status
-================
+==================
 
 Patroni is in active development and accepts contributions. See our `Contributing <https://github.com/zalando/patroni/blob/master/docs/CONTRIBUTING.rst>`__ section below for more details.
 
 We report new releases information `here <https://github.com/zalando/patroni/releases>`__.
 
-===========================
+=========
-Technical Requirements/Installation
+Community
-===========================
+=========
 
-**For Mac**
+There are two places to connect with the Patroni community: `on github <https://github.com/zalando/patroni>`__, via Issues and PRs, and on channel #patroni in the `PostgreSQL Slack <https://postgres-slack.herokuapp.com/>`__.  If you're using Patroni, or just interested, please join us.
+
+===================================
+Technical Requirements/Installation
+===================================
+
+**Pre-requirements for Mac OS**
 
 To install requirements on a Mac, run the following:
 
 ::
 
     brew install postgresql etcd haproxy libyaml python
-    pip install psycopg2 pyyaml
 
-===================
+**Psycopg2**
+
+Starting from `psycopg2-2.8 <http://initd.org/psycopg/articles/2019/04/04/psycopg-28-released/>`__ the binary version of psycopg2 will no longer be installed by default. Installing it from the source code requires C compiler and postgres+python dev packages.
+Since in the python world it is not possible to specify dependency as ``psycopg2 OR psycopg2-binary`` you will have to decide how to install it.
+
+There are a few options available:
+
+1. Use the package manager from your distro
+
+::
+
+    sudo apt-get install python-psycopg2   # install python2 psycopg2 module on Debian/Ubuntu
+    sudo apt-get install python3-psycopg2  # install python3 psycopg2 module on Debian/Ubuntu
+    sudo yum install python-psycopg2       # install python2 psycopg2 on RedHat/Fedora/CentOS
+
+2. Install psycopg2 from the binary package
+
+::
+
+    pip install psycopg2-binary
+
+3. Install psycopg2 from source
+
+::
+
+    pip install psycopg2>=2.5.4
+
+**General installation for pip**
+
+Patroni can be installed with pip:
+
+::
+
+    pip install patroni[dependencies]
+
+where dependencies can be either empty, or consist of one or more of the following:
+
+etcd
+    `python-etcd` module in order to use Etcd as DCS
+consul
+    `python-consul` module in order to use Consul as DCS
+zookeeper
+    `kazoo` module in order to use Zookeeper as DCS
+exhibitor
+    `kazoo` module in order to use Exhibitor as DCS (same dependencies as for Zookeeper)
+kubernetes
+    `kubernetes` module in order to use Kubernetes as DCS in Patroni
+aws
+    `boto` in order to use AWS callbacks
+
+For example, the command in order to install Patroni together with dependencies for Etcd as a DCS and AWS callbacks is:
+
+::
+
+    pip install patroni[etcd,aws]
+
+Note that external tools to call in the replica creation or custom bootstap scripts (i.e. WAL-E) should be installed independently of Patroni.
+
+=======================
 Running and Configuring
-===================
+=======================
 
 To get started, do the following from different terminals:
 ::
@@ -79,9 +143,9 @@ run:
 
     > psql --host 127.0.0.1 --port 5000 postgres
 
-===============
+==================
 YAML Configuration
-===============
+==================
 
 Go `here <https://github.com/zalando/patroni/blob/master/docs/SETTINGS.rst>`__ for comprehensive information about settings for etcd, consul, and ZooKeeper. And for an example, see `postgres0.yml <https://github.com/zalando/patroni/blob/master/postgres0.yml>`__.
 
@@ -91,15 +155,15 @@ Environment Configuration
 
 Go `here <https://github.com/zalando/patroni/blob/master/docs/ENVIRONMENT.rst>`__ for comprehensive information about configuring(overriding) settings via environment variables.
 
-===============
+===================
 Replication Choices
-===============
+===================
 
 Patroni uses Postgres' streaming replication, which is asynchronous by default. Patroni's asynchronous replication configuration allows for ``maximum_lag_on_failover`` settings. This setting ensures failover will not occur if a follower is more than a certain number of bytes behind the leader. This setting should be increased or decreased based on business requirements. It's also possible to use synchronous replication for better durability guarantees. See `replication modes documentation <https://github.com/zalando/patroni/blob/master/docs/replication_modes.rst>`__ for details.
 
-===============================
+======================================
 Applications Should Not Use Superusers
-===============================
+======================================
 
 When connecting from an application, always use a non-superuser. Patroni requires access to the database to function properly. By using a superuser from an application, you can potentially use the entire connection pool, including the connections reserved for superusers, with the ``superuser_reserved_connections`` setting. If Patroni cannot access the Primary because the connection pool is full, behavior will be undesirable.
 

docker-compose.yml

@@ -1,58 +1,69 @@
 # docker compose file for running a 3-node PostgreSQL cluster
-# with etcd as the SIS
+# with 3-node etcd cluster as the DCS and one haproxy node
+version: "2"
 
-patroni_etcd:
+networks:
-    container_name: patroni_etcd
+    demo:
-    image: patroni
-    command: --etcd
 
-dbnode1:
+services:
-    image: patroni
+    etcd1:
-    hostname: dbnode1
+        image: patroni
-    links:
+        networks: [ demo ]
-        - patroni_etcd:patroni_etcd
+        env_file: docker/etcd.env
-    volumes:
+        container_name: demo-etcd1
-        - ./patroni:/patroni
+        hostname: etcd1
-    env_file: docker/patroni-secrets.env
+        command: etcd -name etcd1 -initial-advertise-peer-urls http://etcd1:2380
-    environment:
-        PATRONI_ETCD_HOST: patroni_etcd:2379
-        PATRONI_NAME: dbnode1
-        PATRONI_SCOPE: testcluster
 
-dbnode2:
+    etcd2:
-    image: patroni
+        image: patroni
-    hostname: dbnode2
+        networks: [ demo ]
-    links:
+        env_file: docker/etcd.env
-        - patroni_etcd:patroni_etcd
+        container_name: demo-etcd2
-    volumes:
+        hostname: etcd2
-        - ./patroni:/patroni
+        command: etcd -name etcd2 -initial-advertise-peer-urls http://etcd2:2380
-    env_file: docker/patroni-secrets.env
-    environment:
-        PATRONI_ETCD_HOST: patroni_etcd:2379
-        PATRONI_NAME: dbnode2
-        PATRONI_SCOPE: testcluster
 
-dbnode3:
+    etcd3:
-    image: patroni
+        image: patroni
-    hostname: dbnode3
+        networks: [ demo ]
-    links:
+        env_file: docker/etcd.env
-        - patroni_etcd:patroni_etcd
+        container_name: demo-etcd3
-    volumes:
+        hostname: etcd3
-        - ./patroni:/patroni
+        command: etcd -name etcd3 -initial-advertise-peer-urls http://etcd3:2380
-    env_file: docker/patroni-secrets.env
-    environment:
-        PATRONI_ETCD_HOST: patroni_etcd:2379
-        PATRONI_NAME: dbnode3
-        PATRONI_SCOPE: testcluster
 
-haproxy:
+    patroni1:
-    image: patroni
+        image: patroni
-    links:
+        networks: [ demo ]
-        - patroni_etcd:patroni_etcd
+        env_file: docker/patroni.env
-    ports:
+        hostname: patroni1
-        - "5000"
+        container_name: demo-patroni1
-        - "5001"
+        environment:
-    environment:
+            PATRONI_NAME: patroni1
-        PATRONI_ETCD_HOST: patroni_etcd:2379
+
-        PATRONI_SCOPE: testcluster
+    patroni2:
-    command: --confd
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: patroni2
+        container_name: demo-patroni2
+        environment:
+            PATRONI_NAME: patroni2
+
+    patroni3:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: patroni3
+        container_name: demo-patroni3
+        environment:
+            PATRONI_NAME: patroni3
+
+    haproxy:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: haproxy
+        container_name: demo-haproxy
+        ports:
+            - "5000:5000"
+            - "5001:5001"
+        command: haproxy

docker/README.md

@@ -1,47 +1,117 @@
 # Patroni Dockerfile
-You can run Patroni in a docker container using this Dockerfile, or by using one of the Docker image at
+You can run Patroni in a docker container using this Dockerfile
-
-    https://registry.opensource.zalan.do/v1/repositories/acid/patroni/tags
 
 This Dockerfile is meant in aiding development of Patroni and quick testing of features. It is not a production-worthy
 Dockerfile
 
+    docker build -t patroni .
+
 # Examples
 
 ## Standalone Patroni
 
-    docker run -d registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT
+    docker run -d patroni
 
-## Multiple Patroni's communicating with a standalone etcd inside Docker
+## Three-node Patroni cluster with three-node etcd cluster and one haproxy container using docker-compose
-
-Basically what you would do would be: 
-
-* Run 1 container which provides etcd
-
-        docker run -d <IMAGE> --etcd-only
-
-* Run n containers running Patroni, passing the `--etcd` option to the `docker run` command
-
-        docker run -d <IMAGE> --etcd=<IP FROM etcd CONTAINER:PORT>
-
-To automate this you can run the following script:
-
-    dev_patroni_cluster.sh [OPTIONS]
-
-    Options:
-
-        --image     IMAGE    The Docker image to use for the cluster
-        --members   INT      The number of members for the cluster
-        --name      NAME     The name of the new cluster
 
 Example session:
 
-    $ ./dev_patroni_cluster.sh --image registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT --members=2 --name=bravo
+    $ docker-compose up -d
-    The etcd container is 6be871a11cb373406ca5ea1c6b39e1.0-SNAPSHOTfdde9fb1d6177212d6ad0c0d1bd9b563, ip=172.17.1.24
+    Creating demo-haproxy ...
-    Started Patroni container 67e611f2eca7c40f9e6e0e24a4a8f2cba7e3e56d22a420e15ab9240a37a9d7a4, ip=172.17.1.25
+    Creating demo-patroni2 ...
-    Started Patroni container 47dd12ae635ab83b039f5889e250048b606ed5e48e3650b69e365e7e1d4acbcf, ip=172.17.1.26
+    Creating demo-patroni1 ...
+    Creating demo-patroni3 ...
+    Creating demo-etcd2 ...
+    Creating demo-etcd1 ...
+    Creating demo-etcd3 ...
+    Creating demo-haproxy
+    Creating demo-patroni2
+    Creating demo-patroni1
+    Creating demo-patroni3
+    Creating demo-etcd1
+    Creating demo-etcd2
+    Creating demo-etcd2 ... done
+
     $ docker ps
-    CONTAINER ID        IMAGE                                         COMMAND                CREATED             STATUS              PORTS                          NAMES
+    CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS                              NAMES
-    47dd12ae635a        registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT   "/bin/bash /entrypoi   10 seconds ago      Up 8 seconds        4001/tcp, 5432/tcp, 2380/tcp   bravo_OR64g8bx
+    5b7a90b4cfbf        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 27 seconds                                          demo-etcd2
-    67e611f2eca7        registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT   "/bin/bash /entrypoi   11 seconds ago      Up 10 seconds       2380/tcp, 4001/tcp, 5432/tcp   bravo_si9no8iz
+    e30eea5222f2        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 27 seconds                                          demo-etcd1
-    6be871a11cb3        registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT   "/bin/bash /entrypoi   12 seconds ago      Up 10 seconds       4001/tcp, 5432/tcp, 2380/tcp   bravo_etcd
+    83bcf3cb208f        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 27 seconds                                          demo-etcd3
+    922532c56e7d        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 28 seconds                                          demo-patroni3
+    14f875e445f3        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 28 seconds                                          demo-patroni2
+    110d1073b383        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 28 seconds                                          demo-patroni1
+    5af5e6e36028        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 28 seconds       0.0.0.0:5000-5001->5000-5001/tcp   demo-haproxy
+
+    $ docker logs demo-patroni1
+    2019-02-20 08:19:32,714 INFO: Failed to import patroni.dcs.consul
+    2019-02-20 08:19:32,737 INFO: Selected new etcd server http://etcd3:2379
+    2019-02-20 08:19:35,140 INFO: Lock owner: None; I am patroni1
+    2019-02-20 08:19:35,174 INFO: trying to bootstrap a new cluster
+    ...
+    2019-02-20 08:19:39,310 INFO: postmaster pid=37
+    2019-02-20 08:19:39.314 UTC [37] LOG:  listening on IPv4 address "0.0.0.0", port 5432
+    2019-02-20 08:19:39.321 UTC [37] LOG:  listening on Unix socket "/var/run/postgresql/.s.PGSQL.5432"
+    2019-02-20 08:19:39.353 UTC [39] LOG:  database system was shut down at 2019-02-20 08:19:36 UTC
+    2019-02-20 08:19:39.354 UTC [40] FATAL:  the database system is starting up
+    localhost:5432 - rejecting connections
+    2019-02-20 08:19:39.369 UTC [37] LOG:  database system is ready to accept connections
+    localhost:5432 - accepting connections
+    2019-02-20 08:19:39,383 INFO: establishing a new patroni connection to the postgres cluster
+    2019-02-20 08:19:39,408 INFO: running post_bootstrap
+    2019-02-20 08:19:39,432 WARNING: Could not activate Linux watchdog device: "Can't open watchdog device: [Errno 2] No such file or directory: '/dev/watchdog'"
+    2019-02-20 08:19:39,515 INFO: initialized a new cluster
+    2019-02-20 08:19:49,424 INFO: Lock owner: patroni1; I am patroni1
+    2019-02-20 08:19:49,447 INFO: Lock owner: patroni1; I am patroni1
+    2019-02-20 08:19:49,480 INFO: no action.  i am the leader with the lock
+    2019-02-20 08:19:59,422 INFO: Lock owner: patroni1; I am patroni1
+
+    $ docker exec -ti demo-patroni1 bash
+    postgres@patroni1:~$ patronictl list
+    +---------+----------+------------+--------+---------+----+-----------+
+    | Cluster |  Member  |    Host    |  Role  |  State  | TL | Lag in MB |
+    +---------+----------+------------+--------+---------+----+-----------+
+    |   demo  | patroni1 | 172.22.0.3 | Leader | running |  1 |         0 |
+    |   demo  | patroni2 | 172.22.0.7 |        | running |  1 |         0 |
+    |   demo  | patroni3 | 172.22.0.4 |        | running |  1 |         0 |
+    +---------+----------+------------+--------+---------+----+-----------+
+
+    postgres@patroni1:~$ etcdctl ls --recursive --sort -p /service/demo
+    /service/demo/config
+    /service/demo/initialize
+    /service/demo/leader
+    /service/demo/members/
+    /service/demo/members/patroni1
+    /service/demo/members/patroni2
+    /service/demo/members/patroni3
+    /service/demo/optime/
+    /service/demo/optime/leader
+
+    postgres@patroni1:~$ etcdctl member list
+    1bab629f01fa9065: name=etcd3 peerURLs=http://etcd3:2380 clientURLs=http://etcd3:2379 isLeader=false
+    8ecb6af518d241cc: name=etcd2 peerURLs=http://etcd2:2380 clientURLs=http://etcd2:2379 isLeader=true
+    b2e169fcb8a34028: name=etcd1 peerURLs=http://etcd1:2380 clientURLs=http://etcd1:2379 isLeader=false
+    postgres@patroni1:~$ exit
+
+    $ psql -h localhost -p 5000 -U postgres -W
+    Password: postgres
+    psql (11.2 (Ubuntu 11.2-1.pgdg18.04+1), server 10.7 (Debian 10.7-1.pgdg90+1))
+    Type "help" for help.
+
+    localhost/postgres=# select pg_is_in_recovery();
+     pg_is_in_recovery
+    ───────────────────
+     f
+    (1 row)
+
+    localhost/postgres=# \q
+
+    $ psql -h localhost -p 5001 -U postgres -W
+    Password: postgres
+    psql (11.2 (Ubuntu 11.2-1.pgdg18.04+1), server 10.7 (Debian 10.7-1.pgdg90+1))
+    Type "help" for help.
+
+    localhost/postgres=# select pg_is_in_recovery();
+     pg_is_in_recovery
+    ───────────────────
+     t
+    (1 row)

docker/dev_patroni_cluster.sh

@@ -1,104 +0,0 @@
-#!/bin/bash
-
-DOCKER_IMAGE="registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT"
-MEMBERS=3
-
-
-function usage()
-{
-    cat <<__EOF__
-Usage: $0 
-
-Options:
-
-    --image     IMAGE    The Docker image to use for the cluster
-    --members   INT      The number of members for the cluster
-    --name      NAME     The name of the new cluster
-
-Examples:
-
-    $0 --image ${DOCKER_IMAGE}
-    $0
-    $0 --image ${DOCKER_IMAGE} --members=2
-__EOF__
-}
-
-
-optspec=":-:"
-while getopts "$optspec" optchar; do
-    case "${optchar}" in
-        -)
-            case "${OPTARG}" in
-                help)
-                    usage
-                    exit 0
-                    ;;
-                name)
-                    PATRONI_SCOPE="${!OPTIND}"; OPTIND=$(( $OPTIND + 1 ))
-                    ;;
-                name=*)
-                    PATRONI_SCOPE="${OPTARG#*=}"
-                    ;;
-                image)
-                    DOCKER_IMAGE="${!OPTIND}"; OPTIND=$(( $OPTIND + 1 ))
-                    ;;
-                image=*)
-                    DOCKER_IMAGE="${OPTARG#*=}"
-                    ;;
-                members)
-                    MEMBERS="${!OPTIND}"; OPTIND=$(( $OPTIND + 1 ))
-                    ;;
-                members=*)
-                    MEMBERS="${OPTARG#*=}"
-                    ;;
-                *)
-                    if [ "$OPTERR" = 1 ] && [ "${optspec:0:1}" != ":" ]; then
-                        echo "Unknown option --${OPTARG}" >&2
-                    fi
-                    ;;
-            esac;;
-        *)
-            if [ "$OPTERR" != 1 ] || [ "${optspec:0:1}" = ":" ]; then
-                echo "Non-option argument: '-${OPTARG}'" >&2
-                usage
-                exit 1
-            fi
-            ;;
-    esac
-done
-
-if [ -z ${PATRONI_SCOPE} ]; then
-    PATRONI_SCOPE=$(cat /dev/urandom | LC_ALL=C tr -dc 'a-z0-9' | head -c 8)
-fi
-
-function docker_run()
-{
-    local name=$1
-    shift
-    container=$(docker run -d --name=$name $*)
-    container_ip=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' ${container})
-    echo "Started container ${name}, ip=${container_ip}"
-}
-
-
-ETCD_CONTAINER="${PATRONI_SCOPE}_etcd"
-docker_run ${ETCD_CONTAINER} ${DOCKER_IMAGE} --etcd
-
-DOCKER_ARGS="--link=${ETCD_CONTAINER}:${ETCD_CONTAINER} -e PATRONI_SCOPE=${PATRONI_SCOPE} -e PATRONI_ETCD_HOST=${ETCD_CONTAINER}:2379"
-PATRONI_ENV=$(sed 's/#.*//g' docker/patroni-secrets.env | sed -n 's/^PATRONI_.*$/-e &/p' | tr '\n' ' ')
-PATRONI_VOLUME="-v $(dirname $(dirname $(realpath $0)))/patroni:/patroni"
-
-for i in $(seq 1 "${MEMBERS}"); do
-    container_name=postgres${i}
-    docker_run "${PATRONI_SCOPE}_${container_name}" \
-        $PATRONI_VOLUME \
-        $DOCKER_ARGS \
-        $PATRONI_ENV \
-        -e PATRONI_NAME=${container_name} \
-        ${DOCKER_IMAGE}
-done
-
-docker_run "${PATRONI_SCOPE}_haproxy" \
-    -p=5000 -p=5001 \
-    $DOCKER_ARGS \
-    ${DOCKER_IMAGE} --confd

docker/entrypoint.sh

@@ -1,115 +1,60 @@
-#!/bin/bash
+#!/bin/sh
 
-function usage()
+if [ -f /a.tar.xz ]; then
-{
+    echo "decompressing image..."
-    cat <<__EOF__
+    sudo tar xpJf /a.tar.xz -C / > /dev/null 2>&1
-Usage: $0
+    sudo rm /a.tar.xz
+    sudo ln -snf dash /bin/sh
+fi
 
-Options:
+readonly PATRONI_SCOPE=${PATRONI_SCOPE:-batman}
+PATRONI_NAMESPACE=${PATRONI_NAMESPACE:-/service}
+readonly PATRONI_NAMESPACE=${PATRONI_NAMESPACE%/}
+readonly DOCKER_IP=$(hostname --ip-address)
 
-    --etcd      Do not run Patroni, run a standalone etcd
+case "$1" in
-    --confd     Do not run Patroni, run a standalone confd
+    haproxy)
-    --zookeeper Do not run Patroni, run a standalone zookeeper
+        haproxy -f /etc/haproxy/haproxy.cfg -p /var/run/haproxy.pid -D
-
+        CONFD="confd -prefix=$PATRONI_NAMESPACE/$PATRONI_SCOPE -interval=10 -backend"
-Examples:
+        if [ ! -z "$PATRONI_ZOOKEEPER_HOSTS" ]; then
-
+            while ! /usr/share/zookeeper/bin/zkCli.sh -server $PATRONI_ZOOKEEPER_HOSTS ls /; do
-    $0 --etcd
+                sleep 1
-    $0 --confd
+            done
-    $0 --zookeeper
+            exec dumb-init $CONFD zookeeper -node $PATRONI_ZOOKEEPER_HOSTS
-    $0
+        else
-__EOF__
+            while ! etcdctl cluster-health 2> /dev/null; do
-}
+                sleep 1
-
+            done
-DOCKER_IP=$(hostname --ip-address)
+            exec dumb-init $CONFD etcd -node $(echo $ETCDCTL_ENDPOINTS | sed 's/,/ -node /g')
-PATRONI_SCOPE=${PATRONI_SCOPE:-batman}
+        fi
-ETCD_ARGS="--data-dir /tmp/etcd.data -advertise-client-urls=http://${DOCKER_IP}:2379 -listen-client-urls=http://0.0.0.0:2379 -listen-peer-urls=http://0.0.0.0:2380"
+        ;;
-
+    etcd)
-optspec=":vh-:"
+        exec "$@" -advertise-client-urls http://$DOCKER_IP:2379
-while getopts "$optspec" optchar; do
+        ;;
-    case "${optchar}" in
+    zookeeper)
-        -)
+        exec /usr/share/zookeeper/bin/zkServer.sh start-foreground
-            case "${OPTARG}" in
+        ;;
-                confd)
+esac
-                    haproxy -f /etc/haproxy/haproxy.cfg -p /var/run/haproxy.pid -D
-                    CONFD="confd -prefix=${PATRONI_NAMESPACE:-/service}/$PATRONI_SCOPE -interval=10 -backend"
-                    if [ ! -z ${PATRONI_ZOOKEEPER_HOSTS} ]; then
-                        while ! /usr/share/zookeeper/bin/zkCli.sh -server ${PATRONI_ZOOKEEPER_HOSTS} ls /; do
-                            sleep 1
-                        done
-                        exec $CONFD zookeeper -node ${PATRONI_ZOOKEEPER_HOSTS}
-                    else
-                        while ! curl -s ${PATRONI_ETCD_HOST}/v2/members | jq -r '.members[0].clientURLs[0]' | grep -q http; do
-                            sleep 1
-                        done
-                        exec $CONFD etcd -node $PATRONI_ETCD_HOST
-                    fi
-                    ;;
-                etcd)
-                    exec etcd $ETCD_ARGS
-                    ;;
-                zookeeper)
-                    exec /usr/share/zookeeper/bin/zkServer.sh start-foreground
-                    ;;
-                cheat)
-                    CHEAT=1
-                    ;;
-                help)
-                    usage
-                    exit 0
-                    ;;
-                *)
-                    if [ "$OPTERR" = 1 ] && [ "${optspec:0:1}" != ":" ]; then
-                        echo "Unknown option --${OPTARG}" >&2
-                    fi
-                    ;;
-            esac;;
-        *)
-            if [ "$OPTERR" != 1 ] || [ "${optspec:0:1}" = ":" ]; then
-                echo "Non-option argument: '-${OPTARG}'" >&2
-                usage
-                exit 1
-            fi
-            ;;
-    esac
-done
 
 ## We start an etcd
-if [[ -z ${PATRONI_ETCD_HOST} && -z ${PATRONI_ZOOKEEPER_HOSTS} ]]; then
+if [ -z "$PATRONI_ETCD_HOSTS" ] && [ -z "$PATRONI_ZOOKEEPER_HOSTS" ]; then
-    etcd $ETCD_ARGS > /var/log/etcd.log 2> /var/log/etcd.err &
+    export PATRONI_ETCD_URL="http://127.0.0.1:2379"
-    export PATRONI_ETCD_HOST="127.0.0.1:2379"
+    etcd --data-dir /tmp/etcd.data -advertise-client-urls=$PATRONI_ETCD_URL -listen-client-urls=http://0.0.0.0:2379 > /var/log/etcd.log 2> /var/log/etcd.err &
 fi
 
 export PATRONI_SCOPE
-export PATRONI_NAME="${PATRONI_NAME:-${HOSTNAME}}"
+export PATRONI_NAMESPACE
-export PATRONI_RESTAPI_CONNECT_ADDRESS="${DOCKER_IP}:8008"
+export PATRONI_NAME="${PATRONI_NAME:-$(hostname)}"
+export PATRONI_RESTAPI_CONNECT_ADDRESS="$DOCKER_IP:8008"
 export PATRONI_RESTAPI_LISTEN="0.0.0.0:8008"
-export PATRONI_admin_PASSWORD="${PATRONI_admin_PASSWORD:=admin}"
+export PATRONI_admin_PASSWORD="${PATRONI_admin_PASSWORD:-admin}"
 export PATRONI_admin_OPTIONS="${PATRONI_admin_OPTIONS:-createdb, createrole}"
-export PATRONI_POSTGRESQL_CONNECT_ADDRESS="${DOCKER_IP}:5432"
+export PATRONI_POSTGRESQL_CONNECT_ADDRESS="$DOCKER_IP:5432"
 export PATRONI_POSTGRESQL_LISTEN="0.0.0.0:5432"
-export PATRONI_POSTGRESQL_DATA_DIR="data/${PATRONI_SCOPE}"
+export PATRONI_POSTGRESQL_DATA_DIR="${PATRONI_POSTGRESQL_DATA_DIR:-$PGDATA}"
 export PATRONI_REPLICATION_USERNAME="${PATRONI_REPLICATION_USERNAME:-replicator}"
-export PATRONI_REPLICATION_PASSWORD="${PATRONI_REPLICATION_PASSWORD:-abcd}"
+export PATRONI_REPLICATION_PASSWORD="${PATRONI_REPLICATION_PASSWORD:-replicate}"
 export PATRONI_SUPERUSER_USERNAME="${PATRONI_SUPERUSER_USERNAME:-postgres}"
 export PATRONI_SUPERUSER_PASSWORD="${PATRONI_SUPERUSER_PASSWORD:-postgres}"
-export PATRONI_POSTGRESQL_PGPASS="$HOME/.pgpass"
 
-cat > /patroni.yml <<__EOF__
+exec python3 /patroni.py postgres0.yml
-bootstrap:
-  dcs:
-    postgresql:
-      use_pg_rewind: true
-
-  pg_hba:
-  - host all all 0.0.0.0/0 md5
-  - host replication replicator ${DOCKER_IP}/16    md5
-__EOF__
-
-mkdir -p "$HOME/.config/patroni"
-[ -h "$HOME/.config/patroni/patronictl.yaml" ] || ln -s /patroni.yml "$HOME/.config/patroni/patronictl.yaml"
-
-[ -z $CHEAT ] && exec python /patroni.py /patroni.yml
-
-while true; do
-    sleep 60
-done

docker/etcd.env

@@ -0,0 +1,5 @@
+ETCD_LISTEN_PEER_URLS=http://0.0.0.0:2380
+ETCD_LISTEN_CLIENT_URLS=http://0.0.0.0:2379
+ETCD_INITIAL_CLUSTER=etcd1=http://etcd1:2380,etcd2=http://etcd2:2380,etcd3=http://etcd3:2380
+ETCD_INITIAL_CLUSTER_STATE=new
+ETCD_INITIAL_CLUSTER_TOKEN=tutorial

docker/patroni.env

@@ -1,3 +1,6 @@
+PATRONI_SCOPE=demo
+PATRONI_ETCD_HOSTS='etcd1:2379','etcd2:2379','etcd3:2379'
+
 PATRONI_RESTAPI_USERNAME=admin
 PATRONI_RESTAPI_PASSWORD=admin
 PATRONI_SUPERUSER_USERNAME=postgres
@@ -6,3 +9,6 @@ PATRONI_REPLICATION_USERNAME=replicator
 PATRONI_REPLICATION_PASSWORD=replicate
 PATRONI_admin_PASSWORD=admin
 PATRONI_admin_OPTIONS=createdb,createrole
+
+# for etcdctl
+ETCDCTL_ENDPOINTS=http://etcd1:2379,http://etcd2:2379,http://etcd3:2379

docs/CONTRIBUTING.rst

@@ -5,6 +5,11 @@ Contributing guidelines
 
 Wanna contribute to Patroni? Yay - here is how!
 
+Chatting
+--------
+
+Just want to chat with other Patroni users?  Looking for interactive troubleshooting help? Join us on channel #patroni in the `PostgreSQL Slack <https://postgres-slack.herokuapp.com/>`__.
+
 Reporting issues
 ----------------
 

docs/ENVIRONMENT.rst

@@ -1,6 +1,5 @@
 .. _environment:
 
-==================================
 Environment Configuration Settings
 ==================================
 
@@ -12,6 +11,16 @@ Global/Universal
 -  **PATRONI\_NAME**: name of the node where the current instance of Patroni is running. Must be unique for the cluster.
 -  **PATRONI\_NAMESPACE**: path within the configuration store where Patroni will keep information about the cluster. Default value: "/service"
 -  **PATRONI\_SCOPE**: cluster name
+-  **PATRONI\_LOG\_LEVEL**: sets the general logging level. Default value is **INFO** (see `the docs for Python logging <https://docs.python.org/3.6/library/logging.html#levels>`_)
+-  **PATRONI\_LOG\_FORMAT**: sets the log formatting string. Default value is **%(asctime)s %(levelname)s: %(message)s** (see `the LogRecord attributes <https://docs.python.org/3.6/library/logging.html#logrecord-attributes>`_)
+-  **PATRONI\_LOG\_DATEFORMAT**: sets the datetime formatting string. (see the `formatTime() documentation <https://docs.python.org/3.6/library/logging.html#logging.Formatter.formatTime>`_)
+-  **PATRONI\_LOG\_MAX\_QUEUE\_SIZE**: Patroni is using two-step logging. Log records are written into the in-memory queue and there is a separate thread which pulls them from the queue and writes to stderr or file. The maximum size of the internal queue is limited by default by **1000** records, which is enough to keep logs for the past 1h20m.
+-  **PATRONI\_LOG\_DIR**: Directory to write application logs to. The directory must exist and be writable by the user executing Patroni. If you set this env variable, the application will retain 4 25MB logs by default. You can tune those retention values with `PATRONI_LOG_FILE_NUM` and `PATRONI_LOG_FILE_SIZE` (see below).
+-  **PATRONI\_LOG\_FILE\_NUM**: The number of application logs to retain.
+-  **PATRONI\_LOG\_FILE\_SIZE**: Size of patroni.log file (in bytes) that triggers a log rolling.
+-  **PATRONI\_LOG\_LOGGERS**: Redefine logging level per python module. Example ``PATRONI_LOG_LOGGERS="{patroni.postmaster: WARNING, urllib3: DEBUG}"``
+-  **PATRONI\_DEBUG\_MODE**: When set to a non-empty value, makes Patroni run in the special debug mode that enables stepping with a debugger through some execution paths within Patroni `
+Example ``PATRONI_DEBUG_MODE="on"``
 
 Bootstrap configuration
 -----------------------
@@ -25,15 +34,34 @@ Example: defining ``PATRONI_admin_PASSWORD=strongpasswd`` and ``PATRONI_admin_OP
 Consul
 ------
 -  **PATRONI\_CONSUL\_HOST**: the host:port for the Consul endpoint.
+-  **PATRONI\_CONSUL\_URL**: url for the Consul, in format: http(s)://host:port
+-  **PATRONI\_CONSUL\_PORT**: (optional) Consul port
+-  **PATRONI\_CONSUL\_SCHEME**: (optional) **http** or **https**, defaults to **http**
+-  **PATRONI\_CONSUL\_TOKEN**: (optional) ACL token
+-  **PATRONI\_CONSUL\_VERIFY**: (optional) whether to verify the SSL certificate for HTTPS requests
+-  **PATRONI\_CONSUL\_CACERT**: (optional) The ca certificate. If present it will enable validation.
+-  **PATRONI\_CONSUL\_CERT**: (optional) File with the client certificate
+-  **PATRONI\_CONSUL\_KEY**: (optional) File with the client key. Can be empty if the key is part of certificate.
+-  **PATRONI\_CONSUL\_DC**: (optional) Datacenter to communicate with. By default the datacenter of the host is used.
+-  **PATRONI\_CONSUL\_CONSISTENCY**: (optional) Select consul consistency mode. Possible values are ``default``, ``consistent``, or ``stale`` (more details in `consul API reference <https://www.consul.io/api/features/consistency.html/>`__)
+-  **PATRONI\_CONSUL\_CHECKS**: (optional) list of Consul health checks used for the session. If not specified Consul will use "serfHealth" in additional to the TTL based check created by Patroni. Additional checks, in particular the "serfHealth", may cause the leader lock to expire faster than in `ttl` seconds when the leader instance becomes unavailable.
+-  **PATRONI\_CONSUL\_REGISTER\_SERVICE**: (optional) whether or not to register a service with the name defined by the scope parameter and the tag master, replica or standby-leader depending on the node's role. Defaults to **false**
+-  **PATRONI\_CONSUL\_SERVICE\_CHECK\_INTERVAL**: (optional) how often to perform health check against registered url
 
 Etcd
 ----
--  **PATRONI\_ETCD\_HOST**: the host:port for the etcd endpoint.
+
--  **PATRONI\_ETCD\_URL**: url for the etcd, in format: http(s)://(username:password@)host:port
 -  **PATRONI\_ETCD\_PROXY**: proxy url for the etcd. If you are connecting to the etcd using proxy, use this parameter instead of **PATRONI\_ETCD\_URL**
+-  **PATRONI\_ETCD\_URL**: url for the etcd, in format: http(s)://(username:password@)host:port
+-  **PATRONI\_ETCD\_HOSTS**: list of etcd endpoints in format 'host1:port1','host2:port2',etc...
+-  **PATRONI\_ETCD\_USE\_PROXIES**: If this parameter is set to true, Patroni will consider **hosts** as a list of proxies and will not perform a topology discovery of etcd cluster but stick to a fixed list of **hosts**.
+-  **PATRONI\_ETCD\_PROTOCOL**: http or https, if not specified http is used. If the **url** or **proxy** is specified - will take protocol from them.
+-  **PATRONI\_ETCD\_HOST**: the host:port for the etcd endpoint.
 -  **PATRONI\_ETCD\_SRV**: Domain to search the SRV record(s) for cluster autodiscovery.
--  **PATRONI\_ETCD\_CACERT**: The ca certificate. If pressent it will enable validation.
+-  **PATRONI\_ETCD\_USERNAME**: username for etcd authentication.
--  **PATRONI\_ETCD\_CERT**: File with the client certificate
+-  **PATRONI\_ETCD\_PASSWORD**: password for etcd authentication.
+-  **PATRONI\_ETCD\_CACERT**: The ca certificate. If present it will enable validation.
+-  **PATRONI\_ETCD\_CERT**: File with the client certificate.
 -  **PATRONI\_ETCD\_KEY**: File with the client key. Can be empty if the key is part of certificate.
 
 Exhibitor
@@ -41,20 +69,35 @@ Exhibitor
 -  **PATRONI\_EXHIBITOR\_HOSTS**: initial list of Exhibitor (ZooKeeper) nodes in format: 'host1,host2,etc...'. This list updates automatically whenever the Exhibitor (ZooKeeper) cluster topology changes.
 -  **PATRONI\_EXHIBITOR\_PORT**: Exhibitor port.
 
+.. _kubernetes_environment:
+
+Kubernetes
+----------
+-  **PATRONI\_KUBERNETES\_NAMESPACE**: (optional) Kubernetes namespace where the Patroni pod is running. Default value is `default`.
+-  **PATRONI\_KUBERNETES\_LABELS**: Labels in format ``{label1: value1, label2: value2}``. These labels will be used to find existing objects (Pods and either Endpoints or ConfigMaps) associated with the current cluster. Also Patroni will set them on every object (Endpoint or ConfigMap) it creates.
+-  **PATRONI\_KUBERNETES\_SCOPE\_LABEL**: (optional) name of the label containing cluster name. Default value is `cluster-name`.
+-  **PATRONI\_KUBERNETES\_ROLE\_LABEL**: (optional) name of the label containing Postgres role (`master` or `replica`). Patroni will set this label on the pod it is running in. Default value is `role`.
+-  **PATRONI\_KUBERNETES\_USE\_ENDPOINTS**: (optional) if set to true, Patroni will use Endpoints instead of ConfigMaps to run leader elections and keep cluster state.
+-  **PATRONI\_KUBERNETES\_POD\_IP**: (optional) IP address of the pod Patroni is running in. This value is required when `PATRONI_KUBERNETES_USE_ENDPOINTS` is enabled and is used to populate the leader endpoint subsets when the pod's PostgreSQL is promoted.
+-  **PATRONI\_KUBERNETES\_PORTS**: (optional) if the Service object has the name for the port, the same name must appear in the Endpoint object, otherwise service won't work. For example, if your service is defined as ``{Kind: Service, spec: {ports: [{name: postgresql, port: 5432, targetPort: 5432}]}}``, then you have to set ``PATRONI_KUBERNETES_PORTS='{[{"name": "postgresql", "port": 5432}]}'`` and Patroni will use it for updating subsets of the leader Endpoint. This parameter is used only if `PATRONI_KUBERNETES_USE_ENDPOINTS` is set.
+
 PostgreSQL
 ----------
 -  **PATRONI\_POSTGRESQL\_LISTEN**: IP address + port that Postgres listens to. Multiple comma-separated addresses are permitted, as long as the port component is appended after to the last one with a colon, i.e. ``listen: 127.0.0.1,127.0.0.2:5432``. Patroni will use the first address from this list to establish local connections to the PostgreSQL node.
 -  **PATRONI\_POSTGRESQL\_CONNECT\_ADDRESS**: IP address + port through which Postgres is accessible from other nodes and applications.
 -  **PATRONI\_POSTGRESQL\_DATA\_DIR**: The location of the Postgres data directory, either existing or to be initialized by Patroni.
+-  **PATRONI\_POSTGRESQL\_CONFIG\_DIR**: The location of the Postgres configuration directory, defaults to the data directory. Must be writable by Patroni.
 -  **PATRONI\_POSTGRESQL\_BIN_DIR**: Path to PostgreSQL binaries. (pg_ctl, pg_rewind, pg_basebackup, postgres) The  default value is an empty string meaning that PATH environment variable will be used to find the executables.
 -  **PATRONI\_POSTGRESQL\_PGPASS**: path to the `.pgpass <https://www.postgresql.org/docs/current/static/libpq-pgpass.html>`__ password file. Patroni creates this file before executing pg\_basebackup and under some other circumstances. The location must be writable by Patroni.
 -  **PATRONI\_REPLICATION\_USERNAME**: replication username; the user will be created during initialization. Replicas will use this user to access master via streaming replication
 -  **PATRONI\_REPLICATION\_PASSWORD**: replication password; the user will be created during initialization.
 -  **PATRONI\_SUPERUSER\_USERNAME**: name for the superuser, set during initialization (initdb) and later used by Patroni to connect to the postgres. Also this user is used by pg_rewind.
 -  **PATRONI\_SUPERUSER\_PASSWORD**: password for the superuser, set during initialization (initdb).
+-  **PATRONI\_REWIND\_USERNAME**: name for the user for ``pg_rewind``; the user will be created during initialization of postgres 11+ and all necessary `permissions <https://www.postgresql.org/docs/11/app-pgrewind.html#id-1.9.5.8.8>`__ will be granted.
+-  **PATRONI\_REWIND\_PASSWORD**: password for the user for ``pg_rewind``; the user will be created during initialization.
 
 REST API
--------- 
+--------
 -  **PATRONI\_RESTAPI\_CONNECT\_ADDRESS**: IP address and port to access the REST API.
 -  **PATRONI\_RESTAPI\_LISTEN**: IP address and port that Patroni will listen to, to provide health-check information for HAProxy.
 -  **PATRONI\_RESTAPI\_USERNAME**: Basic-auth username to protect unsafe REST API endpoints.

docs/README.rst

@@ -1,8 +1,8 @@
 .. _readme:
 
-=================
+============
-How Patroni Works
+Introduction
-=================
+============
 
 Patroni originated as a fork of `Governor <https://github.com/compose/governor>`__, the project from Compose. It includes plenty of new features.
 
@@ -13,30 +13,96 @@ For additional background info, see:
 * `PostgreSQL HA with Kubernetes and Patroni <https://www.youtube.com/watch?v=iruaCgeG7qs>`__, talk by Josh Berkus at KubeCon 2016 (video)
 * `Feb. 2016 Zalando Tech blog post <https://tech.zalando.de/blog/zalandos-patroni-a-template-for-high-availability-postgresql/>`__
 
-==================
+
 Development Status
-==================
+------------------
 
 Patroni is in active development and accepts contributions. See our :ref:`Contributing <contributing>` section below for more details.
 
 We report new releases information :ref:`here <releases>`.
 
-===================================
-Technical Requirements/Installation
-===================================
 
-**For Mac**
+Technical Requirements/Installation
+-----------------------------------
+
+**Pre-requirements for Mac OS**
 
 To install requirements on a Mac, run the following:
 
 ::
 
     brew install postgresql etcd haproxy libyaml python
-    pip install psycopg2 pyyaml
 
-=======================
+.. _psycopg2_install_options:
+
+**Psycopg2**
+
+Starting from `psycopg2-2.8 <http://initd.org/psycopg/articles/2019/04/04/psycopg-28-released/>`__ the binary version of psycopg2 will no longer be installed by default. Installing it from the source code requires C compiler and postgres+python dev packages.
+Since in the python world it is not possible to specify dependency as ``psycopg2 OR psycopg2-binary`` you will have to decide how to install it.
+
+There are a few options available:
+
+1. Use the package manager from your distro
+
+::
+
+    sudo apt-get install python-psycopg2   # install python2 psycopg2 module on Debian/Ubuntu
+    sudo apt-get install python3-psycopg2  # install python3 psycopg2 module on Debian/Ubuntu
+    sudo yum install python-psycopg2       # install python2 psycopg2 on RedHat/Fedora/CentOS
+
+2. Install psycopg2 from the binary package
+
+::
+
+    pip install psycopg2-binary
+
+3. Install psycopg2 from source
+
+::
+
+    pip install psycopg2>=2.5.4
+
+**General installation for pip**
+
+Patroni can be installed with pip:
+
+::
+
+    pip install patroni[dependencies]
+
+where dependencies can be either empty, or consist of one or more of the following:
+
+etcd
+    `python-etcd` module in order to use Etcd as DCS
+consul
+    `python-consul` module in order to use Consul as DCS
+zookeeper
+    `kazoo` module in order to use Zookeeper as DCS
+exhibitor
+    `kazoo` module in order to use Exhibitor as DCS (same dependencies as for Zookeeper)
+kubernetes
+    `kubernetes` module in order to use Kubernetes as DCS in Patroni
+aws
+    `boto` in order to use AWS callbacks
+
+For example, the command in order to install Patroni together with dependencies for Etcd as a DCS and AWS callbacks is:
+
+::
+
+    pip install patroni[etcd,aws]
+
+Note that external tools to call in the replica creation or custom bootstrap scripts (i.e. WAL-E) should be installed
+independently of Patroni.
+
+
+.. _running_configuring:
+
 Running and Configuring
-=======================
+-----------------------
+
+The following section assumes Patroni repository as being cloned from https://github.com/zalando/patroni. Namely, you
+will need example configuration files `postgres0.yml` and `postgres1.yml`. If you installed Patroni with pip, you can
+obtain those files from the git repository and replace `./patroni.py` below with `patroni` command.
 
 To get started, do the following from different terminals:
 ::
@@ -60,27 +126,27 @@ run:
 
     > psql --host 127.0.0.1 --port 5000 postgres
 
-==================
+
 YAML Configuration
-==================
+------------------
 
 Go :ref:`here <settings>` for comprehensive information about settings for etcd, consul, and ZooKeeper. And for an example, see `postgres0.yml <https://github.com/zalando/patroni/blob/master/postgres0.yml>`__.
 
-=========================
+
 Environment Configuration
-=========================
+-------------------------
 
 Go :ref:`here <environment>` for comprehensive information about configuring(overriding) settings via environment variables.
 
-===================
+
 Replication Choices
-===================
+-------------------
 
 Patroni uses Postgres' streaming replication, which is asynchronous by default. Patroni's asynchronous replication configuration allows for ``maximum_lag_on_failover`` settings. This setting ensures failover will not occur if a follower is more than a certain number of bytes behind the leader. This setting should be increased or decreased based on business requirements. It's also possible to use synchronous replication for better durability guarantees. See :ref:`replication modes documentation <replication_modes>` for details.
 
-======================================
+
 Applications Should Not Use Superusers
-======================================
+--------------------------------------
 
 When connecting from an application, always use a non-superuser. Patroni requires access to the database to function properly. By using a superuser from an application, you can potentially use the entire connection pool, including the connections reserved for superusers, with the ``superuser_reserved_connections`` setting. If Patroni cannot access the Primary because the connection pool is full, behavior will be undesirable.
 

docs/SETTINGS.rst

@@ -10,19 +10,53 @@ Global/Universal
 -  **namespace**: path within the configuration store where Patroni will keep information about the cluster. Default value: "/service"
 -  **scope**: cluster name
 
+Log
+---
+-  **level**: sets the general logging level. Default value is **INFO** (see `the docs for Python logging <https://docs.python.org/3.6/library/logging.html#levels>`_)
+-  **format**: sets the log formatting string. Default value is **%(asctime)s %(levelname)s: %(message)s** (see `the LogRecord attributes <https://docs.python.org/3.6/library/logging.html#logrecord-attributes>`_)
+-  **dateformat**: sets the datetime formatting string. (see the `formatTime() documentation <https://docs.python.org/3.6/library/logging.html#logging.Formatter.formatTime>`_)
+-  **max\_queue\_size**: Patroni is using two-step logging. Log records are written into the in-memory queue and there is a separate thread which pulls them from the queue and writes to stderr or file. The maximum size of the internal queue is limited by default by **1000** records, which is enough to keep logs for the past 1h20m.
+-  **dir**: Directory to write application logs to. The directory must exist and be writable by the user executing Patroni. If you set this value, the application will retain 4 25MB logs by default. You can tune those retention values with `file_num` and `file_size` (see below).
+-  **file\_num**: The number of application logs to retain.
+-  **file\_size**: Size of patroni.log file (in bytes) that triggers a log rolling.
+-  **loggers**: This section allows redefining logging level per python module
+    -  **patroni.postmaster: WARNING**
+    -  **urllib3: DEBUG**
+
+.. _bootstrap_settings:
+
 Bootstrap configuration
 -----------------------
 -  **dcs**: This section will be written into `/<namespace>/<scope>/config` of a given configuration store after initializing of new cluster. This is the global configuration for the cluster. If you want to change some parameters for all cluster nodes - just do it in DCS (or via Patroni API) and all nodes will apply this configuration.
     -  **loop\_wait**: the number of seconds the loop will sleep. Default value: 10
     -  **ttl**: the TTL to acquire the leader lock. Think of it as the length of time before initiation of the automatic failover process. Default value: 30
+    -  **retry\_timeout**: timeout for DCS and PostgreSQL operation retries. DCS or network issues shorter than this will not cause Patroni to demote the leader. Default value: 10
     -  **maximum\_lag\_on\_failover**: the maximum bytes a follower may lag to be able to participate in leader election.
     -  **master\_start\_timeout**: the amount of time a master is allowed to recover from failures before failover is triggered. Default is 300 seconds. When set to 0 failover is done immediately after a crash is detected if possible. When using asynchronous replication a failover can cause lost transactions. Best worst case failover time for master failure is: loop\_wait + master\_start\_timeout + loop\_wait, unless master\_start\_timeout is zero, in which case it's just loop\_wait. Set the value according to your durability/availability tradeoff.
-    -  **synchronous\_mode**: turns on synchronous replication mode. In this mode a replica will be chosen as synchronous and only the latest leader and synchronous replica are able to participate in leader election. Synchronous mode makes sure that succesfully committed transactions will not be lost at failover, at the cost of losing availability for writes when Patroni cannot ensure transaction durability. See `replication modes documentation <https://github.com/zalando/patroni/blob/master/docs/replication_modes.rst>`__ for details.
+    -  **synchronous\_mode**: turns on synchronous replication mode. In this mode a replica will be chosen as synchronous and only the latest leader and synchronous replica are able to participate in leader election. Synchronous mode makes sure that successfully committed transactions will not be lost at failover, at the cost of losing availability for writes when Patroni cannot ensure transaction durability. See :ref:`replication modes documentation <replication_modes>` for details.
+    -  **synchronous\_mode\_strict**: prevents disabling synchronous replication if no synchronous replicas are available, blocking all client writes to the master. See :ref:`replication modes documentation <replication_modes>` for details.
     -  **postgresql**:
-        -  **use\_pg\_rewind**:whether or not to use pg_rewind
+        -  **use\_pg\_rewind**: whether or not to use pg_rewind
         -  **use\_slots**: whether or not to use replication_slots. Must be False for PostgreSQL 9.3. You should comment out max_replication_slots before it becomes ineligible for leader status.
-        -  **recovery\_conf**: additional configuration settings written to recovery.conf when configuring follower. 
+        -  **recovery\_conf**: additional configuration settings written to recovery.conf when configuring follower.
         -  **parameters**: list of configuration settings for Postgres. Many of these are required for replication to work.
+    -  **standby\_cluster**: if this section is defined, we want to bootstrap a standby cluster.
+        -  **host**: an address of remote master
+        -  **port**: a port of remote master
+        -  **primary\_slot\_name**: which slot on the remote master to use for replication. This parameter is optional, the default value is derived from the instance name (see function `slot_name_from_member_name`).
+        -  **create\_replica\_methods**: an ordered list of methods that can be used to bootstrap standby leader from the remote master, can be different from the list defined in :ref:`postgresql_settings`
+        -  **restore\_command**: command to restore WAL records from the remote master to standby leader, can be different from the list defined in :ref:`postgresql_settings`
+        -  **archive\_cleanup\_command**: cleanup command for standby leader
+        -  **recovery\_min\_apply\_delay**: how long to wait before actually apply WAL records on a standby leader
+    -  **slots**: define permanent replication slots. These slots will be preserved during switchover/failover. Patroni will try to create slots before opening connections to the cluster.
+        -  **my_slot_name**: the name of replication slot. It is the responsibility of the operator to make sure that there are no clashes in names between replication slots automatically created by Patroni for members and permanent replication slots.
+            -  **type**: slot type. Could be ``physical`` or ``logical``. If the slot is logical, you have to additionally define ``database`` and ``plugin``.
+            -  **database**: the database name where logical slots should be created.
+            -  **plugin**: the plugin name for the logical slot.
+-  **method**: custom script to use for bootstrapping this cluster.
+   See :ref:`custom bootstrap methods documentation <custom_bootstrap>` for details.
+   When ``initdb`` is specified revert to the default ``initdb`` command. ``initdb`` is also triggered when no ``method``
+   parameter is present in the configuration file.
 -  **initdb**: List options to be passed on to initdb.
         -  **- data-checksums**: Must be enabled when pg_rewind is needed on 9.3.
         -  **- encoding: UTF8**: default encoding for new databases.
@@ -30,30 +64,50 @@ Bootstrap configuration
 -  **pg\_hba**: list of lines that you should add to pg\_hba.conf.
         -  **- host all all 0.0.0.0/0 md5**.
         -  **- host replication replicator 127.0.0.1/32 md5**: A line like this is required for replication.
--  **users**: Some additional users users which needs to be created after initializing new cluster
+-  **users**: Some additional users which need to be created after initializing new cluster
     -  **admin**: the name of user
         -  **password: zalando**:
         -  **options**: list of options for CREATE USER statement
             -  **- createrole**
             -  **- createdb**
--  **post_init**: An additional script that will be executed after initializing the cluster. The script receives a connection string URL (with the cluster superuser as a user name). The PGPASSFILE variable is set to the location of pgpass file.
+-  **post\_bootstrap** or **post\_init**: An additional script that will be executed after initializing the cluster. The script receives a connection string URL (with the cluster superuser as a user name). The PGPASSFILE variable is set to the location of pgpass file.
+
+.. _consul_settings:
 
 Consul
 ------
--  **host**: the host:port for the Consul endpoint.
+Most of the parameters are optional, but you have to specify one of the **host** or **url**
+
+-  **host**: the host:port for the Consul endpoint, in format: http(s)://host:port
+-  **url**: url for the Consul endpoint
+-  **port**: (optional) Consul port
+-  **scheme**: (optional) **http** or **https**, defaults to **http**
+-  **token**: (optional) ACL token
+-  **verify**: (optional) whether to verify the SSL certificate for HTTPS requests
+-  **cacert**: (optional) The ca certificate. If present it will enable validation.
+-  **cert**: (optional) file with the client certificate
+-  **key**: (optional) file with the client key. Can be empty if the key is part of **cert**.
+-  **dc**: (optional) Datacenter to communicate with. By default the datacenter of the host is used.
+-  **consistency**: (optional) Select consul consistency mode. Possible values are ``default``, ``consistent``, or ``stale`` (more details in `consul API reference <https://www.consul.io/api/features/consistency.html/>`__)
+-  **checks**: (optional) list of Consul health checks used for the session. If not specified Consul will use "serfHealth" in additional to the TTL based check created by Patroni. Additional checks, in particular the "serfHealth", may cause the leader lock to expire faster than in `ttl` seconds when the leader instance becomes unavailable
+-  **register\_service**: (optional) whether or not to register a service with the name defined by the scope parameter and the tag master, replica or standby-leader depending on the node's role. Defaults to **false**
+-  **service\_check\_interval**: (optional) how often to perform health check against registered url
 
 Etcd
 ----
-Most of the parameters are optional, but you have to specify one of the **host**, **url**, **proxy** or **srv**
+Most of the parameters are optional, but you have to specify one of the **host**, **hosts**, **url**, **proxy** or **srv**
+
 -  **host**: the host:port for the etcd endpoint.
+-  **hosts**: list of etcd endpoint in format host1:port1,host2:port2,etc... Could be a comma separated string or an actual yaml list.
+-  **use\_proxies**: If this parameter is set to true, Patroni will consider **hosts** as a list of proxies and will not perform a topology discovery of etcd cluster.
 -  **url**: url for the etcd
 -  **proxy**: proxy url for the etcd. If you are connecting to the etcd using proxy, use this parameter instead of **url**
 -  **srv**: Domain to search the SRV record(s) for cluster autodiscovery.
 -  **protocol**: (optional) http or https, if not specified http is used. If the **url** or **proxy** is specified - will take protocol from them.
--  **username**: (optional) username for etcd authentication
+-  **username**: (optional) username for etcd authentication.
 -  **password**: (optional) password for etcd authentication.
--  **cacert**: (optional) The ca certificate. If pressent it will enable validation.
+-  **cacert**: (optional) The ca certificate. If present it will enable validation.
--  **cert**: (optional) file with the client certificate
+-  **cert**: (optional) file with the client certificate.
 -  **key**: (optional) file with the client key. Can be empty if the key is part of **cert**.
 
 Exhibitor
@@ -62,6 +116,18 @@ Exhibitor
 -  **poll\_interval**: how often the list of ZooKeeper and Exhibitor nodes should be updated from Exhibitor
 -  **port**: Exhibitor port.
 
+.. _kubernetes_settings:
+
+Kubernetes
+----------
+-  **namespace**: (optional) Kubernetes namespace where Patroni pod is running. Default value is `default`.
+-  **labels**: Labels in format ``{label1: value1, label2: value2}``. These labels will be used to find existing objects (Pods and either Endpoints or ConfigMaps) associated with the current cluster. Also Patroni will set them on every object (Endpoint or ConfigMap) it creates.
+-  **scope\_label**: (optional) name of the label containing cluster name. Default value is `cluster-name`.
+-  **role\_label**: (optional) name of the label containing role (master or replica). Patroni will set this label on the pod it runs in. Default value is ``role``.
+-  **use\_endpoints**: (optional) if set to true, Patroni will use Endpoints instead of ConfigMaps to run leader elections and keep cluster state.
+-  **pod\_ip**: (optional) IP address of the pod Patroni is running in. This value is required when `use_endpoints` is enabled and is used to populate the leader endpoint subsets when the pod's PostgreSQL is promoted.
+-  **ports**: (optional) if the Service object has the name for the port, the same name must appear in the Endpoint object, otherwise service won't work. For example, if your service is defined as ``{Kind: Service, spec: {ports: [{name: postgresql, port: 5432, targetPort: 5432}]}}``, then you have to set ``kubernetes.ports: {[{"name": "postgresql", "port": 5432}]}`` and Patroni will use it for updating subsets of the leader Endpoint. This parameter is used only if `kubernetes.use_endpoints` is set.
+
 .. _postgresql_settings:
 
 PostgreSQL
@@ -73,30 +139,46 @@ PostgreSQL
     -  **replication**:
         -  **username**: replication username; the user will be created during initialization. Replicas will use this user to access master via streaming replication
         -  **password**: replication password; the user will be created during initialization.
+    -  **rewind**:
+        -  **username**: name for the user for ``pg_rewind``; the user will be created during initialization of postgres 11+ and all necessary `permissions <https://www.postgresql.org/docs/11/app-pgrewind.html#id-1.9.5.8.8>`__ will be granted.
+        -  **password**: password for the user for ``pg_rewind``; the user will be created during initialization.
 -  **callbacks**: callback scripts to run on certain actions. Patroni will pass the action, role and cluster name. (See scripts/aws.py as an example of how to write them.)
         -  **on\_reload**: run this script when configuration reload is triggered.
-        -  **on\_restart**: run this script when the cluster restarts.
+        -  **on\_restart**: run this script when the postgres restarts (without changing role).
-        -  **on\_role\_change**: run this script when the cluster is being promoted or demoted.
+        -  **on\_role\_change**: run this script when the postgres is being promoted or demoted.
-        -  **on\_start**: run this script when the cluster starts.
+        -  **on\_start**: run this script when the postgres starts.
-        -  **on\_stop**: run this script when the cluster stops.
+        -  **on\_stop**: run this script when the postgres stops.
 -  **connect\_address**: IP address + port through which Postgres is accessible from other nodes and applications.
--  **create\_replica\_methods**: an ordered list of the create methods for turning a Patroni node into a new replica. "basebackup" is the default method; other methods are assumed to refer to scripts, each of which is configured as its own config item.
+-  **create\_replica\_methods**: an ordered list of the create methods for turning a Patroni node into a new replica.
--  **data\_dir**: The location of the Postgres data directory, either existing or to be initialized by Patroni.
+   "basebackup" is the default method; other methods are assumed to refer to scripts, each of which is configured as its
--  **bin\_dir**: Path to PostgreSQL binaries. (pg_ctl, pg_rewind, pg_basebackup, postgres) The  default value is an empty string meaning that PATH environment variable will be used to find the executables.
+   own config item. See :ref:`custom replica creation methods documentation <custom_replica_creation>` for further explanation.
+-  **data\_dir**: The location of the Postgres data directory, either :ref:`existing <existing_data>` or to be initialized by Patroni.
+-  **config\_dir**: The location of the Postgres configuration directory, defaults to the data directory. Must be writable by Patroni.
+-  **bin\_dir**: Path to PostgreSQL binaries (pg_ctl, pg_rewind, pg_basebackup, postgres). The default value is an empty string meaning that PATH environment variable will be used to find the executables.
 -  **listen**: IP address + port that Postgres listens to; must be accessible from other nodes in the cluster, if you're using streaming replication. Multiple comma-separated addresses are permitted, as long as the port component is appended after to the last one with a colon, i.e. ``listen: 127.0.0.1,127.0.0.2:5432``. Patroni will use the first address from this list to establish local connections to the PostgreSQL node.
+-  **use\_unix\_socket**: specifies that Patroni should prefer to use unix sockets to connect to the cluster. Default value is ``false``. If ``unix_socket_directories`` is defined, Patroni will use the first suitable value from it to connect to the cluster and fallback to tcp if nothing is suitable. If ``unix_socket_directories`` is not specified in ``postgresql.parameters``, Patroni will assume that the default value should be used and omit ``host`` from the connection parameters.
 -  **pgpass**: path to the `.pgpass <https://www.postgresql.org/docs/current/static/libpq-pgpass.html>`__ password file. Patroni creates this file before executing pg\_basebackup, the post_init script and under some other circumstances. The location must be writable by Patroni.
 -  **recovery\_conf**: additional configuration settings written to recovery.conf when configuring follower.
--  **custom_conf** : path to an optional custom ``postgresql.conf`` file, that will be used in place of ``postgresql.base.conf``. The file must exist on all cluster nodes, be readable by PostgreSQL and will be included from its location on the real ``postgresql.conf``. Note that Patroni will not monitor this file for changes, nor backup it. However, its settings can still be overriden by Patroni's own configuration facilities - see `dynamic configuration <https://github.com/zalando/patroni/blob/master/docs/dynamic_configuration.rst>`__ for details.
+-  **custom\_conf** : path to an optional custom ``postgresql.conf`` file, that will be used in place of ``postgresql.base.conf``. The file must exist on all cluster nodes, be readable by PostgreSQL and will be included from its location on the real ``postgresql.conf``. Note that Patroni will not monitor this file for changes, nor backup it. However, its settings can still be overridden by Patroni's own configuration facilities - see :ref:`dynamic configuration <dynamic_configuration>` for details.
 -  **parameters**: list of configuration settings for Postgres. Many of these are required for replication to work.
+-  **pg\_hba**: list of lines that Patroni will use to generate ``pg_hba.conf``. This parameter has higher priority than ``bootstrap.pg_hba``. Together with :ref:`dynamic configuration <dynamic_configuration>` it simplifies management of ``pg_hba.conf``.
+        -  **- host all all 0.0.0.0/0 md5**.
+        -  **- host replication replicator 127.0.0.1/32 md5**: A line like this is required for replication.
+-  **pg\_ident**: list of lines that Patroni will use to generate ``pg_ident.conf``. Together with :ref:`dynamic configuration <dynamic_configuration>` it simplifies management of ``pg_ident.conf``.
+        -  **- mapname1 systemname1 pguser1**.
+        -  **- mapname1 systemname2 pguser2**.
 -  **pg\_ctl\_timeout**: How long should pg_ctl wait when doing ``start``, ``stop`` or ``restart``. Default value is 60 seconds.
 -  **use\_pg\_rewind**: try to use pg\_rewind on the former leader when it joins cluster as a replica.
--  **remove\_data\_directory\_on\_rewind\_failure**: If this option is enabled, Patroni will remove postgres data directory and recreate replica. Otherwise it will try to follow the new leader. Default value is **false**.
+-  **remove\_data\_directory\_on\_rewind\_failure**: If this option is enabled, Patroni will remove the PostgreSQL data directory and recreate the replica. Otherwise it will try to follow the new leader. Default value is **false**.
--  **replica\_method** for each create_replica_method other than basebackup, you would add a configuration section of the same name. At a minimum, this should include "command" with a full path to the actual script to be executed.  Other configuration parameters will be passed along to the script in the form "parameter=value".
+-  **remove\_data\_directory\_on\_diverged\_timelines**: Patroni will remove the PostgreSQL data directory and recreate the replica if it notices that timelines are diverging and the former master can not start streaming from the new master. This option is useful when ``pg_rewind`` can not be used. Default value is **false**.
+-  **replica\_method**: for each create_replica_methods other than basebackup, you would add a configuration section of the same name. At a minimum, this should include "command" with a full path to the actual script to be executed. Other configuration parameters will be passed along to the script in the form "parameter=value".
 
 REST API
--------- 
+--------
--  **connect\_address**: IP address and port to access the REST API.
+-  **connect\_address**: IP address (or hostname) and port, to access the Patroni's REST API. All the members of the cluster must be able to connect to this address, so unless the Patroni setup is intended for a demo inside the localhost, this address must be a non "localhost" or loopback addres (ie: "localhost" or "127.0.0.1"). It can serve as a endpoint for HTTP health checks (read below about the "listen" REST API parameter), and also for user queries (either directly or via the REST API), as well as for the health checks done by the cluster members during leader elections (for example, to determine whether the master is still running, or if there is a node which has a WAL position that is ahead of the one doing the query; etc.) The connect_address is put in the member key in DCS, making it possible to translate the member name into the address to connect to its REST API.
--  **listen**: IP address and port that Patroni will listen to, to provide health-check information for HAProxy.
+
+-  **listen**: IP address (or hostname) and port that Patroni will listen to for the REST API - to provide also the same health checks and cluster messaging between the participating nodes, as described above. to provide health-check information for HAProxy (or any other load balancer capable of doing a HTTP "OPTION" or "GET" checks).
+
 -  **Optional**:
         -  **authentication**:
             -  **username**: Basic-auth username to protect unsafe REST API endpoints.
@@ -105,6 +187,21 @@ REST API
         -  **certfile**: Specifies the file with the certificate in the PEM format. If the certfile is not specified or is left empty, the API server will work without SSL.
         -  **keyfile**: Specifies the file with the secret key in the PEM format.
 
+.. _patronictl_settings:
+
+CTL
+---
+-  **Optional**:
+    -  **insecure**: Allow connections to REST API without verifying SSL certs.
+    -  **cacert**: Specifices the file with the CA_BUNDLE file or directory with certificates of trusted CAs to use while verifying REST API SSL certs.
+    -  **certfile**: Specifies the file with the certificate in the PEM format to use while verifying REST API SSL certs. If not provided patronictl will use the value provided for REST API "certfile" parameter.
+
 ZooKeeper
 ----------
 -  **hosts**: list of ZooKeeper cluster members in format: ['host1:port1', 'host2:port2', 'etc...'].
+
+Watchdog
+--------
+- **mode**: ``off``, ``automatic`` or ``required``. When ``off`` watchdog is disabled. When ``automatic`` watchdog will be used if available, but ignored if it is not. When ``required`` the node will not become a leader unless watchdog can be successfully enabled.
+- **device**: Path to watchdog device. Defaults to ``/dev/watchdog``.
+- **safety_margin**: Number of seconds of safety margin between watchdog triggering and leader key expiration.

docs/conf.py

@@ -18,9 +18,11 @@
 # documentation root, use os.path.abspath to make it absolute, like shown here.
 #
 import os
-# import sys
-# sys.path.insert(0, os.path.abspath('.'))
 
+import sys
+sys.path.insert(0, os.path.abspath('..'))
+
+from patroni.version import __version__
 
 # -- General configuration ------------------------------------------------
 
@@ -51,7 +53,7 @@ master_doc = 'index'
 
 # General information about the project.
 project = 'Patroni'
-copyright = '2016, Zalando SE'
+copyright = '2015 Compose, Zalando SE'
 author = 'Zalando SE'
 
 # The version info for the project you're documenting, acts as replacement for
@@ -59,9 +61,9 @@ author = 'Zalando SE'
 # built documents.
 #
 # The short X.Y version.
-version = '1.2'
+version = __version__[:__version__.rfind('.')]
 # The full version, including alpha/beta/rc tags.
-release = '1.2.2'
+release = __version__
 
 # The language for content autogenerated by Sphinx. Refer to documentation
 # for a list of supported languages.

docs/dynamic_configuration.rst

@@ -16,9 +16,9 @@ Patroni configuration is stored in the DCS (Distributed Configuration Store). Th
 	These options are defined in the configuration file and take precedence over dynamic configuration.
 	patroni.yml could be changed and reload in runtime (without restart of Patroni) by sending SIGHUP to the Patroni process or by performing ``POST /reload`` REST-API request.
 
-- Environment :ref:`configuration <environment>` .
+- Environment :ref:`configuration <environment>`.
 	It is possible to set/override some of the "Local" configuration parameters with environment variables.
-	Environment configuration is very useful when you are running in a dynamic environment and you don't know some of the parameters in advance (for example it's not possible to know you external IP address when you are running inside ``docker``).
+	Environment configuration is very useful when you are running in a dynamic environment and you don't know some of the parameters in advance (for example it's not possible to know your external IP address when you are running inside ``docker``).
 
 Some of the PostgreSQL parameters must hold the same values on the master and the replicas. For those, values set either in the local patroni configuration files or via the environment variables take no effect. To alter or set their values one must change the shared configuration in the DCS. Below is the actual list of such parameters together with the default values:
 
@@ -42,7 +42,7 @@ There are some other Postgres parameters controlled by Patroni:
 
 - listen_addresses - is set either from ``postgresql.listen`` or from ``PATRONI_POSTGRESQL_LISTEN`` environment variable
 - port - is set either from ``postgresql.listen`` or from ``PATRONI_POSTGRESQL_LISTEN`` environment variable
-- cluster_name - is set either from ``scope`` or from ``PATRRONI_SCOPE`` environment variable
+- cluster_name - is set either from ``scope`` or from ``PATRONI_SCOPE`` environment variable
 - hot_standby: on
 
 To be on the safe side parameters from the above lists are not written into ``postgresql.conf``, but passed as a list of arguments to the ``pg_ctl start`` which gives them the highest precedence, even above `ALTER SYSTEM <https://www.postgresql.org/docs/current/static/sql-altersystem.html>`__
@@ -76,6 +76,7 @@ Also, the following Patroni configuration options can be changed only dynamicall
 - loop_wait: 10
 - retry_timeouts: 10
 - maximum_lag_on_failover: 1048576
+- check_timeline: false
 - postgresql.use_slots: true
 
 Upon changing these options, Patroni will read the relevant section of the configuration stored in DCS and change its

docs/existing_data.rst

@@ -0,0 +1,36 @@
+.. _existing_data:
+
+Convert a Standalone to a Patroni Cluster
+=========================================
+
+This section describes the process for converting a standalone PostgreSQL instance into a Patroni cluster.
+
+To deploy a Patroni cluster without using a pre-existing PostgreSQL instance, see :ref:`Running and Configuring <running_configuring>` instead.
+
+Procedure
+---------
+
+A Patroni cluster can be started with a data directory from a single-node PostgreSQL database. This is achieved by following closely these steps:
+
+1. Manually start PostgreSQL daemon
+2. Create Patroni superuser and replication users as defined in the :ref:`authentication <postgresql_settings>` section of the Patroni configuration. If this user is created in SQL, the following queries achieve this:
+
+.. code-block:: sql
+
+   CREATE USER $PATRONI_SUPERUSER_USERNAME WITH SUPERUSER ENCRYPTED PASSWORD '$PATRONI_SUPERUSER_PASSWORD';
+   CREATE USER $PATRONI_REPLICATION_USERNAME WITH REPLICATION ENCRYPTED PASSWORD '$PATRONI_REPLICATION_PASSWORD';
+
+3. Start Patroni (e.g. ``patroni /etc/patroni/patroni.yml``). It automatically detects that PostgreSQL daemon is already running but its configuration might be out-of-date.
+4. Ask Patroni to restart the node with ``patronictl restart cluster-name node-name``. This step is only required if PostgreSQL configuration is out-of-date.
+
+
+FAQ
+---
+
+- During Patroni startup, Patroni complains that it cannot bind to the PostgreSQL port.
+
+  You need to verify ``listen_addresses`` and ``port`` in ``postgresql.conf`` and ``postgresql.listen`` in ``patroni.yml``. Don't forget that ``pg_hba.conf`` should allow such access.
+
+- After asking Patroni to restart the node, PostgreSQL displays the error message ``could not open configuration file "/etc/postgresql/10/main/pg_hba.conf": No such file or directory``
+
+  It can mean various things depending on how you manage PostgreSQL configuration. If you specified `postgresql.config_dir`, Patroni generates the ``pg_hba.conf`` based on the settings in the :ref:`bootstrap <bootstrap_settings>` section only when it bootstraps a new cluster. In this scenario the ``PGDATA`` was not empty, therefore no bootstrap happened. This file must exist beforehand.

docs/index.rst

@@ -6,11 +6,11 @@
 Introduction
 ============
 
-Patroni is a template for you to create your own customized, high-availability solution using Python and - for maximum accessibility - a distributed configuration store like `ZooKeeper <https://zookeeper.apache.org/>`__, `etcd <https://github.com/coreos/etcd>`__ or `Consul <https://github.com/hashicorp/consul>`__. Database engineers, DBAs, DevOps engineers, and SREs who are looking to quickly deploy HA PostgreSQL in the datacenter-or anywhere else-will hopefully find it useful.
+Patroni is a template for you to create your own customized, high-availability solution using Python and - for maximum accessibility - a distributed configuration store like `ZooKeeper <https://zookeeper.apache.org/>`__, `etcd <https://github.com/coreos/etcd>`__, `Consul <https://github.com/hashicorp/consul>`__ or `Kubernetes <https://kubernetes.io>`__. Database engineers, DBAs, DevOps engineers, and SREs who are looking to quickly deploy HA PostgreSQL in the datacenter-or anywhere else-will hopefully find it useful.
 
 We call Patroni a "template" because it is far from being a one-size-fits-all or plug-and-play replication system. It will have its own caveats. Use wisely. There are many ways to run high availability with PostgreSQL; for a list, see the `PostgreSQL Documentation <https://wiki.postgresql.org/wiki/Replication,_Clustering,_and_Connection_Pooling>`__.
 
-**Note to Kubernetes users**: We're currently developing Patroni to be as useful as possible for teams running Kubernetes on top of Google Compute Engine; Patroni can be the HA solution for Postgres in such an environment. To this end, we've created a `Helm Chart <https://github.com/kubernetes/charts/tree/master/incubator/patroni>`__ that enables you to deploy a five-node Patroni cluster using a Kubernetes PetSet.
+**Note to Kubernetes users**: Patroni can run natively on top of Kubernetes. Take a look at the :ref:`Kubernetes <kubernetes>` chapter of the Patroni documentation.
 
 
 .. toctree::
@@ -21,8 +21,11 @@ We call Patroni a "template" because it is far from being a one-size-fits-all or
    dynamic_configuration
    ENVIRONMENT
    SETTINGS
+   replica_bootstrap
    replication_modes
    pause
+   kubernetes
+   watchdog
    releases
    CONTRIBUTING
 
@@ -32,5 +35,3 @@ Indices and tables
 * :ref:`genindex`
 * :ref:`modindex`
 * :ref:`search`
-
-

docs/kubernetes.rst

@@ -0,0 +1,53 @@
+.. _kubernetes:
+
+Using Patroni with Kubernetes
+=============================
+
+Patroni can use Kubernetes objects in order to store the state of the cluster and manage the leader key. That makes it
+capable of operating Postgres in Kubernetes environment without any consistency store, namely, one doesn't
+need to run an extra Etcd deployment. There are two different type of Kubernetes objects Patroni can use to store the
+leader and the configuration keys, they are configured with the `kubernetes.use_endpoints` or `PATRONI_KUBERNETES_USE_ENDPOINTS`
+environment variable.
+
+Use Endpoints
+-------------
+
+Despite the fact that this is the recommended mode, it is turned off by default for compatibility reasons. When it is on, Patroni stores
+the cluster configuration and the leader key in the `metadata: annotations` fields of the respective `Endpoints` it creates.
+Changing the leader is safer than when using `ConfigMaps`, since both the annotations, containing the leader information, and the actual addresses
+pointing to the running leader pod are updated simultaneously in one go.
+
+Use ConfigMaps
+--------------
+
+In this mode, Patroni will create ConfigMaps instead of Endpoints and store keys inside meta-data of those ConfigMaps.
+Changing the leader takes at least two updates, one to the leader ConfigMap and another to the respective Endpoint.
+
+There are two ways to direct the traffic to the Postgres master:
+
+- use the `callback script <https://github.com/zalando/patroni/blob/master/kubernetes/callback.py>`_ provided by Patroni
+- configure the Kubernetes Postgres service to use the label selector with the `role_label` (configured in patroni configuration).
+
+Note that in some cases, for instance, when running on OpenShift, there is no alternative to using ConfigMaps.
+
+Configuration
+-------------
+
+Patroni Kubernetes :ref:`settings <kubernetes_settings>` and :ref:`environment variables <kubernetes_environment>` are described in the general chapters of the documentation.
+
+Examples
+--------
+
+- The `kubernetes <https://github.com/zalando/patroni/tree/master/kubernetes>`__ folder of the Patroni repository contains
+  examples of the Docker image, the Kubernetes manifest and the callback script in order to test Patroni Kubernetes setup.
+  Note that in the current state it will not be able to use PersistentVolumes because of permission issues.
+
+- You can find the full-featured Docker image that can use Persistent Volumes in the
+  `Spilo Project <https://github.com/zalando/spilo>`_.
+
+- There is also a `Helm chart <https://github.com/kubernetes/charts/tree/master/incubator/patroni>`_
+  to deploy the Spilo image configured with Patroni running using Kubernetes.
+
+- In order to run your database clusters at scale using Patroni and Spilo, take a look at the
+  `postgres-operator <https://github.com/zalando-incubator/postgres-operator>`_ project. It implements the operator pattern
+  to manage Spilo clusters.

docs/pause.rst

@@ -25,7 +25,7 @@ When Patroni runs in a paused mode, it does not change the state of PostgreSQL,
 
 - If there is no leader lock in the cluster, the running master acquires the lock. If there is more than one master node, then the first master to acquire the lock wins. If there are no masters altogether, Patroni does not try to promote any replicas. There is an exception in this rule: if there is no leader lock because the old master has demoted itself due to the manual promotion, then only the candidate node mentioned in the promotion request may take the leader lock. When the new leader lock is granted (i.e. after promoting a replica manually), Patroni makes sure the replicas that were streaming from the previous leader will switch to the new one.
 
-- When Postgres is stopped, Patroni does not try to start it. When Patroni is stopped, it does not to stop Postgres instance it is managing.
+- When Postgres is stopped, Patroni does not try to start it. When Patroni is stopped, it does not try to stop the Postgres instance it is managing.
 
 User guide
 ----------

docs/releases.rst

@@ -3,6 +3,933 @@
 Release notes
 =============
 
+Version 1.6.0
+-------------
+
+This version adds compatibility with PostgreSQL 12, makes is possible to run pg_rewind without superuser on PostgreSQL 11 and newer, and enables IPv6 support.
+
+
+**New features**
+
+- Psycopg2 was removed from requirements and must be installed independently (Alexander Kukushkin)
+
+  Starting from 2.8.0 ``psycopg2`` was split into two different packages, ``psycopg2``, and ``psycopg2-binary``, which could be installed at the same time into the same place on the filesystem. In order to decrease dependency hell problem, we let a user choose how to install it. There are a few options available, please consult the :ref:`documentation <psycopg2_install_options>`.
+
+- Compatibility with PostgreSQL 12 (Alexander Kukushkin)
+
+  Starting from PostgreSQL 12 there is no ``recovery.conf`` anymore and all former recovery parameters are converted into `GUC <https://www.enterprisedb.com/blog/what-is-a-guc-variable>`_. In order to protect from ``ALTER SYSTEM SET primary_conninfo`` or similar, Patroni will parse ``postgresql.auto.conf`` and remove all standby and recovery parameters from there. Patroni config remains backward compatible. For example despite ``restore_command`` being a GUC, one can still specify it in the ``postgresql.recovery_conf.restore_command`` section and Patroni will write it into ``postgresql.conf`` for PostgreSQL 12.
+
+- Make it possible to use ``pg_rewind`` without superuser on PostgreSQL 11 and newer (Alexander Kukushkin)
+
+  If you want to use this feature please define ``username`` and ``password`` in the ``postgresql.authentication.rewind`` section of Patroni configuration file. For an already existing cluster you will have to create the user manually and ``GRANT EXECUTE`` permission on a few functions. You can find more details in the PostgreSQL `documentation <https://www.postgresql.org/docs/11/app-pgrewind.html#id-1.9.5.8.8>`__.
+
+- Do a smart comparison of actual and desired ``primary_conninfo`` values on replicas (Alexander Kukushkin)
+
+  It might help to avoid replica restart when you are converting an already existing primary-standby cluster to one managed by Patroni
+
+- IPv6 support (Alexander Kukushkin)
+
+  There were two major issues. Patroni REST API service was listening only on ``0.0.0.0`` and IPv6 IP addresses used in the ``api_url`` and ``conn_url`` were not properly quoted.
+
+- Kerberos support (Ajith Vilas, Alexander Kukushkin)
+
+  It makes possible using Kerberos authentication between Postgres nodes instead of defining passwords in Patroni configuration file
+
+- Manage ``pg_ident.conf`` (Alexander Kukushkin)
+
+  This functionality works similarly to ``pg_hba.conf``: if the ``postgresql.pg_ident`` is defined in the config file or DCS, Patroni will write its value to ``pg_ident.conf``, however, if ``postgresql.parameters.ident_file`` is defined, Patroni will assume that ``pg_ident`` is managed from outside and not update the file.
+
+
+  **Improvements in REST API**
+
+- Added ``/health`` endpoint (Wilfried Roset)
+
+  It will return an HTTP status code only if PostgreSQL is running
+
+- Added ``/read-only`` and ``/read-write`` endpoints (Julien Riou)
+
+  The ``/read-only`` endpoint enables reads balanced across replicas and the primary. The ``/read-write`` endpoint is an alias for ``/primary``, ``/leader`` and ``/master``.
+
+- Use ``SSLContext`` to wrap the REST API socket (Julien Riou)
+
+  Usage of ``ssl.wrap_socket()`` is deprecated and was still allowing soon-to-be-deprecated protocols like TLS 1.1.
+
+
+**Logging improvements**
+
+- Two-step logging (Alexander Kukushkin)
+
+  All log messages are first written into the in-memory queue and later they are asynchronously flushed into the stderr or file from a separate thread. The maximum queue size is limited (configurable). If the limit is reached, Patroni will start losing logs, which is still better than blocking the HA loop.
+
+- Enable debug logging for GET/OPTIONS API calls together with latency (Jan Tomsa)
+
+  It will help with debugging of health-checks performed by HAProxy, Consul or other tooling that decides which node is the primary/replica.
+
+- Log exceptions caught in Retry (Daniel Kucera)
+
+  Log the final exception when either the number of attempts or the timeout were reached. It will hopefully help to debug some issues when communication to DCS fails.
+
+
+**Improvements in patronictl**
+
+- Enhance dialogues for scheduled switchover and restart (Rafia Sabih)
+
+  Previously dialogues did not take into account scheduled actions and therefore were misleading.
+
+- Check if config file exists (Wilfried Roset)
+
+  Be verbose about configuration file when the given filename does not exists, instead of ignoring silently (which can lead to misunderstanding).
+
+- Add fallback value for ``EDITOR`` (Wilfried Roset)
+
+  When the ``EDITOR`` environment variable was not defined, ``patronictl edit-config`` was failing with `PatroniCtlException`. The new strategy is to try ``editor`` and than ``vi``, which should be available on most systems.
+
+
+**Improvements in Consul support**
+
+- Allow to specify Consul consistency mode (Jan Tomsa)
+
+  You can read more about consistency mode `here <https://www.consul.io/api/features/consistency.html>`__.
+
+- Reload Consul config on SIGHUP (Cameron Daniel, Alexander Kukushkin)
+
+  It is especially useful when somebody is changing the value of ``token``.
+
+
+**Bugfixes**
+
+- Fix corner case in switchover/failover (Sharoon Thomas)
+
+  The variable ``scheduled_at`` may be undefined if REST API is not accessible and we are using DCS as a fallback.
+
+- Open trust to localhost in ``pg_hba.conf`` during custom bootstrap (Alexander Kukushkin)
+
+  Previously it was open only to unix_socket, which was causing a lot of errors: ``FATAL:  no pg_hba.conf entry for replication connection from host "127.0.0.1", user "replicator"``
+
+- Consider synchronous node as healthy even when the former leader is ahead (Alexander Kukushkin)
+
+  If the primary loses access to the DCS, it restarts Postgres in read-only, but it might happen that other nodes can still access the old primary via the REST API. Such a situation was causing the synchronous standby not to promote because the old primary was reporting WAL position ahead of the synchronous standby.
+
+- Standby cluster bugfixes (Alexander Kukushkin)
+
+  Make it possible to bootstrap a replica in a standby cluster when the standby_leader is not accessible and a few other minor fixes.
+
+
+Version 1.5.6
+-------------
+
+**New features**
+
+- Support work with etcd cluster via set of proxies (Alexander Kukushkin)
+
+  It might happen that etcd cluster is not accessible directly but via set of proxies. In this case Patroni will not perform etcd topology discovery but just round-robin via proxy hosts. Behavior is controlled by `etcd.use_proxies`.
+
+- Changed callbacks behavior when role on the node is changed (Alexander)
+
+  If the role was changed from `master` or `standby_leader` to `replica` or from `replica` to `standby_leader`, `on_restart` callback will not be called anymore in favor of `on_role_change` callback.
+
+- Change the way how we start postgres (Alexander)
+
+  Use `multiprocessing.Process` instead of executing itself and `multiprocessing.Pipe` to transmit the postmaster pid to the Patroni process. Before that we were using pipes, what was leaving postmaster process with stdin closed.
+
+**Bug fixes**
+
+- Fix role returned by REST API for the standby leader (Alexander)
+
+  It was incorrectly returning `replica` instead of `standby_leader`
+
+- Wait for callback end if it could not be killed (Julien Tachoires)
+
+  Patroni doesn't have enough privileges to terminate the callback script running under `sudo` what was cancelling the new callback. If the running script could not be killed, Patroni will wait until it finishes and then run the next callback.
+
+- Reduce lock time taken by dcs.get_cluster method (Alexander)
+
+  Due to the lock being held DCS slowness was affecting the REST API health checks causing false positives.
+
+- Improve cleaning of PGDATA when `pg_wal`/`pg_xlog` is a symlink (Julien)
+
+  In this case Patroni will explicitly remove files from the target directory.
+
+- Remove unnecessary usage of os.path.relpath (Ants Aasma)
+
+  It depends on being able to resolve the working directory, what will fail if Patroni is started in a directory that is later unlinked from the filesystem.
+
+- Do not enforce ssl version when communicating with Etcd (Alexander)
+
+  For some unknown reason python3-etcd on debian and ubuntu are not based on the latest version of the package and therefore it enforces TLSv1 which is not supported by Etcd v3. We solved this problem on Patroni side.
+
+Version 1.5.5
+-------------
+
+This version introduces the possibility of automatic reinit of the former master, improves patronictl list output and fixes a number of bugs.
+
+**New features**
+
+- Add support of `PATRONI_ETCD_PROTOCOL`, `PATRONI_ETCD_USERNAME` and `PATRONI_ETCD_PASSWORD` environment variables (Étienne M)
+
+  Before it was possible to configure them only in the config file or as a part of `PATRONI_ETCD_URL`, which is not always convenient.
+
+- Make it possible to automatically reinit the former master (Alexander Kukushkin)
+
+  If the pg_rewind is disabled or can't be used, the former master could fail to start as a new replica due to diverged timelines. In this case, the only way to fix it is wiping the data directory and reinitializing. This behavior could be changed by setting `postgresql.remove_data_directory_on_diverged_timelines`. When it is set, Patroni will wipe the data directory and reinitialize the former master automatically.
+
+- Show information about timelines in patronictl list (Alexander)
+
+  It helps to detect stale replicas. In addition to that, `Host` will include ':{port}' if the port value isn't default or there is more than one member running on the same host.
+
+- Create a headless service associated with the $SCOPE-config endpoint (Alexander)
+
+  The "config" endpoint keeps information about the cluster-wide Patroni and Postgres configuration, history file, and last but the most important, it holds the `initialize` key. When the Kubernetes master node is restarted or upgraded, it removes endpoints without services. The headless service will prevent it from being removed.
+
+**Bug fixes**
+
+- Adjust the read timeout for the leader watch blocking query (Alexander)
+
+  According to the Consul documentation, the actual response timeout is increased by a small random amount of additional wait time added to the supplied maximum wait time to spread out the wake up time of any concurrent requests. It adds up to `wait / 16` additional time to the maximum duration. In our case we are adding `wait / 15` or 1 second depending on what is bigger.
+
+- Always use replication=1 when connecting via replication protocol to the postgres (Alexander)
+
+  Starting from Postgres 10 the line in the pg_hba.conf with database=replication doesn't accept connections with the parameter replication=database.
+
+- Don't write primary_conninfo into recovery.conf for wal-only standby cluster (Alexander)
+
+  Despite not having neither `host` nor `port` defined in the `standby_cluster` config, Patroni was putting the `primary_conninfo` into the `recovery.conf`, which is useless and generating a lot of errors.
+
+
+Version 1.5.4
+-------------
+
+This version implements flexible logging and fixes a number of bugs.
+
+**New features**
+
+- Improvements in logging infrastructure (Alexander Kukushkin, Lucas Capistrant, Alexander Anikin)
+
+  Logging configuration could be configured not only from environment variables but also from Patroni config file. It makes it possible to change logging configuration in runtime by updating config and doing reload or sending SIGHUP to the Patroni process. By default Patroni writes logs to stderr, but now it becomes possible to write logs directly into the file and rotate when it reaches a certain size. In addition to that added support of custom dateformat and the possibility to fine-tune log level for each python module.
+
+- Make it possible to take into account the current timeline during leader elections (Alexander Kukushkin)
+
+  It could happen that the node is considering itself as a healthiest one although it is currently not on the latest known timeline. In some cases we want to avoid promoting of such node, which could be achieved by setting `check_timeline` parameter to `true` (default behavior remains unchanged).
+
+- Relaxed requirements on superuser credentials
+
+  Libpq allows opening connections without explicitly specifying neither username nor password. Depending on situation it relies either on pgpass file or trust authentication method in pg_hba.conf. Since pg_rewind is also using libpq, it will work the same way.
+
+- Implemented possibility to configure Consul Service registration and check interval via environment variables (Alexander Kukushkin)
+
+  Registration of service in Consul was added in the 1.5.0, but so far it was only possible to turn it on via patroni.yaml.
+
+**Stability Improvements**
+
+- Set archive_mode to off during the custom bootstrap (Alexander Kukushkin)
+
+  We want to avoid archiving wals and history files until the cluster is fully functional.  It really helps if the custom bootstrap involves pg_upgrade.
+
+- Apply five seconds backoff when loading global config on start (Alexander Kukushkin)
+
+  It helps to avoid hammering DCS when Patroni just starting up.
+
+- Reduce amount of error messages generated on shutdown (Alexander Kukushkin)
+
+  They were harmless but rather annoying and sometimes scary.
+
+- Explicitly secure rw perms for recovery.conf at creation time (Lucas)
+
+  We don't want anybody except patroni/postgres user reading this file, because it contains replication user and password.
+
+- Redirect HTTPServer exceptions to logger (Julien Riou)
+
+  By default, such exceptions were logged on standard output messing with regular logs.
+
+**Bug fixes**
+
+- Removed stderr pipe to stdout on pg_ctl process (Cody Coons)
+
+  Inheriting stderr from the main Patroni process allows all Postgres logs to be seen along with all patroni logs. This is very useful in a container environment as Patroni and Postgres logs may be consumed using standard tools (docker logs, kubectl, etc). In addition to that, this change fixes a bug with Patroni not being able to catch postmaster pid when postgres writing some warnings into stderr.
+
+- Set Consul service check deregister timeout in Go time format (Pavel Kirillov)
+
+  Without explicitly mentioned time unit registration was failing.
+
+- Relax checks of standby_cluster cluster configuration (Dmitry Dolgov, Alexander Kukushkin)
+
+  It was accepting only strings as valid values and therefore it was not possible to specify the port as integer and create_replica_methods as a list.
+
+Version 1.5.3
+-------------
+
+Compatibility and bugfix release.
+
+- Improve stability when running with python3 against zookeeper (Alexander Kukushkin)
+
+  Change of `loop_wait` was causing Patroni to disconnect from zookeeper and never reconnect back.
+
+- Fix broken compatibility with postgres 9.3 (Alexander)
+
+  When opening a replication connection we should specify replication=1, beacuse 9.3 does not understand replication='database'
+
+- Make sure we refresh Consul session at least once per HA loop and improve handling of consul sessions exceptions (Alexander)
+
+  Restart of local consul agent invalidates all sessions related to the node. Not calling session refresh on time and not doing proper handling of session errors was causing demote of the primary.
+
+Version 1.5.2
+-------------
+
+Compatibility and bugfix release.
+
+- Compatibility with kazoo-2.6.0 (Alexander Kukushkin)
+
+  In order to make sure that requests are performed with an appropriate timeout, Patroni redefines create_connection method from python-kazoo module. The last release of kazoo slightly changed the way how create_connection method is called.
+
+- Fix Patroni crash when Consul cluster loses the leader (Alexander)
+
+  The crash was happening due to incorrect implementation of touch_member method, it should return boolean and not raise any exceptions.
+
+Version 1.5.1
+-------------
+
+This version implements support of permanent replication slots, adds support of pgBackRest and fixes number of bugs.
+
+**New features**
+
+- Permanent replication slots (Alexander Kukushkin)
+
+  Permanent replication slots are preserved on failover/switchover, that is, Patroni on the new primary will create configured replication slots right after doing promote. Slots could be configured with the help of `patronictl edit-config`. The initial configuration could be also done in the :ref:`bootstrap.dcs <settings>`.
+
+- Add pgbackrest support (Yogesh Sharma)
+
+  pgBackrest can restore in existing $PGDATA folder, this allows speedy restore as files which have not changed since last backup are skipped, to support this feature new parameter `keep_data` has been introduced. See :ref:`replica creation method <custom_replica_creation>` section for additional examples.
+
+**Bug fixes**
+
+- A few bugfixes in the "standby cluster" workflow (Alexander)
+
+  Please see https://github.com/zalando/patroni/pull/823 for more details.
+
+- Fix REST API health check when cluster management is paused and DCS is not accessible (Alexander)
+
+  Regression was introduced in https://github.com/zalando/patroni/commit/90cf930036a9d5249265af15d2b787ec7517cf57
+
+Version 1.5.0
+-------------
+
+This version enables Patroni HA cluster to operate in a standby mode, introduces experimental support for running on Windows, and provides a new configuration parameter to register PostgreSQL service in Consul.
+
+**New features**
+
+- Standby cluster (Dmitry Dolgov)
+
+  One or more Patroni nodes can form a standby cluster that runs alongside the primary one (i.e. in another datacenter) and consists of standby nodes that replicate from the master in the primary cluster. All PostgreSQL nodes in the standby cluster are replicas; one of those replicas elects itself to replicate directly from the remote master, while the others replicate from it in a cascading manner. More detailed description of this feature and some configuration examples can be found at :ref:`here <standby_cluster>`.
+
+- Register Services in Consul (Pavel Kirillov, Alexander Kukushkin)
+
+  If `register_service` parameter in the consul :ref:`configuration <consul_settings>` is enabled, the node will register a service with the name `scope` and the tag `master`, `replica` or `standby-leader`.
+
+- Experimental Windows support (Pavel Golub)
+
+  From now on it is possible to run Patroni on Windows, although Windows support is brand-new and hasn't received as much real-world testing as its Linux counterpart. We welcome your feedback!
+
+**Improvements in patronictl**
+
+- Add patronictl -k/--insecure flag and support for restapi cert (Wilfried Roset)
+
+  In the past if the REST API was protected by the self-signed certificates `patronictl` would fail to verify them. There was no way to  disable that verification. It is now possible to configure `patronictl` to skip the certificate verification altogether or provide CA and client certificates in the :ref:`ctl: <patronictl_settings>` section of configuration.
+
+- Exclude members with nofailover tag from patronictl switchover/failover output (Alexander Anikin)
+
+  Previously, those members were incorrectly proposed as candidates when performing interactive switchover or failover via patronictl.
+
+**Stability improvements**
+
+- Avoid parsing non-key-value output lines of pg_controldata (Alexander Anikin)
+
+  Under certain circuimstances pg_controldata outputs lines without a colon character. That would trigger an error in Patroni code that parsed pg_controldata output, hiding the actual problem; often such lines are emitted in a warning shown by pg_controldata before the regular output, i.e. when the binary major version does not match the one of the PostgreSQL data directory.
+
+- Add member name to the error message during the leader election (Jan Mussler)
+
+  During the leader election, Patroni connects to all known members of the cluster and requests their status. Such status is written to the Patroni log and includes the name of the member. Previously, if the member was not accessible, the error message did not indicate its name, containing only the URL.
+
+- Immediately reserve the WAL position upon creation of the replication slot (Alexander Kukushkin)
+
+  Starting from 9.6, `pg_create_physical_replication_slot` function provides an additional boolean parameter `immediately_reserve`. When it is set to `false`, which is also the default, the slot doesn't reserve the WAL position until it receives the first client connection, potentially losing some segments required by the client in a time window between the slot creation and the intiial client connection.
+
+- Fix bug in strict synchronous replication (Alexander Kukushkin)
+
+  When running with `synchronous_mode_strict: true`, in some cases Patroni puts `*` into the `synchronous_standby_names`, changing the sync state for most of the replication connections to `potential`. Previously, Patroni couldn't pick a synchronous candidate under such curcuimstances, as it only considered those with the state `async`.
+
+
+Version 1.4.6
+-------------
+
+**Bug fixes and stability improvements**
+
+This release fixes a critical issue with Patroni API /master endpoint returning 200 for the non-master node. This is a
+reporting issue, no actual split-brain, but under certain circumstances clients might be directed to the read-only node.
+
+- Reset is_leader status on demote (Alexander Kukushkin, Oleksii Kliukin)
+
+  Make sure demoted cluster member stops responding with code 200 on the /master API call.
+
+- Add new "cluster_unlocked" field to the API output (Dmitry Dolgov)
+
+  This field indicates whether the cluster has the master running. It can be used when it is not possible to query any
+  other node but one of the replicas.
+
+Version 1.4.5
+-------------
+
+**New features**
+
+- Improve logging when applying new postgres configuration (Don Seiler)
+
+  Patroni logs changed parameter names and values.
+
+- Python 3.7 compatibility (Christoph Berg)
+
+  async is a reserved keyword in python3.7
+
+- Set state to "stopped" in the DCS when a member is shut down (Tony Sorrentino)
+
+  This shows the member state as "stopped" in "patronictl list" command.
+
+- Improve the message logged when stale postmaster.pid matches a running process (Ants Aasma)
+
+  The previous one was beyond confusing.
+
+- Implement patronictl reload functionality (Don Seiler)
+
+  Before that it was only possible to reload configuration by either calling REST API or by sending SIGHUP signal to the Patroni process.
+
+- Take and apply some parameters from controldata when starting as a replica (Alexander Kukushkin)
+
+  The value of `max_connections` and some other parameters set in the global configuration may be lower than the one actually used by the primary; when this happens, the replica cannot start and should be fixed manually. Patroni takes care of that now by reading and applying the value from  `pg_controldata`, starting postgres and setting `pending_restart` flag.
+
+- If set, use LD_LIBRARY_PATH when starting postgres (Chris Fraser)
+
+  When starting up Postgres, Patroni was passing along PATH, LC_ALL and LANG env vars if they are set. Now it is doing the same with LD_LIBRARY_PATH. It should help if somebody installed PostgreSQL to non-standard place.
+
+- Rename create_replica_method to create_replica_methods (Dmitry Dolgov)
+
+  To make it clear that it's actually an array. The old name is still supported for backward compatibility.
+
+**Bug fixes and stability improvements**
+
+- Fix condition for the replica start due to pg_rewind in paused state (Oleksii  Kliukin)
+
+  Avoid starting the replica that had already executed pg_rewind before.
+
+- Respond 200 to the master health-check only if update_lock has been successful (Alexander)
+
+  Prevent Patroni from reporting itself a master on the former (demoted) master if DCS is partitioned.
+
+- Fix compatibility with the new consul module (Alexander)
+
+  Starting from v1.1.0 python-consul changed internal API and started using `list` instead of `dict` to pass query parameters.
+
+- Catch exceptions from Patroni REST API thread during shutdown (Alexander)
+
+  Those uncaught exceptions kept PostgreSQL running at shutdown.
+
+- Do crash recovery only when Postgres runs as the master (Alexander)
+
+  Require `pg_controldata` to report  'in production' or 'shutting down' or 'in crash recovery'. In all other cases no crash recovery is necessary.
+
+- Improve handling of configuration errors (Henning Jacobs, Alexander)
+
+  It is possible to change a lot of parameters in runtime (including `restapi.listen`) by updating Patroni config file and sending SIGHUP to Patroni process. This fix eliminates obscure exceptions from the 'restapi' thread when some of the parameters receive invalid values.
+
+
+Version 1.4.4
+-------------
+
+**Stability improvements**
+
+- Fix race condition in poll_failover_result (Alexander Kukushkin)
+
+  It didn't affect directly neither failover nor switchover, but in some rare cases it was reporting success too early, when the former leader released the lock, producing a 'Failed over to "None"' instead of 'Failed over to "desired-node"' message.
+
+- Treat Postgres parameter names as case insensitive (Alexander)
+
+  Most of the Postgres parameters have snake_case names, but there are three exceptions from this rule: DateStyle, IntervalStyle and TimeZone. Postgres accepts those parameters when written in a different case (e.g. timezone = 'some/tzn'); however, Patroni was unable to find case-insensitive matches of those parameter names in pg_settings and ignored such parameters as a result.
+
+- Abort start if attaching to running postgres and cluster not initialized (Alexander)
+
+  Patroni can attach itself to an already running Postgres instance. It is imperative to start running Patroni on the master node before getting to the replicas.
+
+- Fix behavior of patronictl scaffold (Alexander)
+
+  Pass dict object to touch_member instead of json encoded string, DCS implementation will take care of encoding it.
+
+- Don't demote master if failed to update leader key in pause (Alexander)
+
+  During maintenance a DCS may start failing write requests while continuing to responds to read ones. In that case, Patroni used to put the Postgres master node to a read-only mode after failing to update the leader lock in DCS.
+
+- Sync replication slots when Patroni notices a new postmaster process (Alexander)
+
+  If Postgres has been restarted, Patroni has to make sure that list of replication slots matches its expectations.
+
+- Verify sysid and sync replication slots after coming out of pause (Alexander)
+
+  During the `maintenance` mode it may happen that data directory was completely rewritten and therefore we have to make sure that `Database system identifier` still belongs to our cluster and replication slots are in sync with Patroni expectations.
+
+- Fix a possible failure to start not running Postgres on a data directory with postmaster lock file present (Alexander)
+
+  Detect reuse of PID from the postmaster lock file. More likely to hit such problem if you run Patroni and Postgres in the docker container.
+
+- Improve protection of DCS being accidentally wiped (Alexander)
+
+  Patroni has a lot of logic in place to prevent failover in such case; it can also restore all keys back; however, until this change an accidental removal of /config key was switching off pause mode for 1 cycle of HA loop.
+
+- Do not exit when encountering invalid system ID (Oleksii Kliukin)
+
+  Do not exit when the cluster system ID is empty or the one that doesn't pass the validation check. In that case, the cluster most likely needs a reinit; mention it in the result message. Avoid terminating Patroni, as otherwise reinit cannot happen.
+
+**Compatibility with Kubernetes 1.10+**
+
+- Added check for empty subsets (Cody Coons)
+
+  Kubernetes 1.10.0+ started returning `Endpoints.subsets` set to `None` instead of `[]`.
+
+**Bootstrap improvements**
+
+- Make deleting recovery.conf optional (Brad Nicholson)
+
+  If `bootstrap.<custom_bootstrap_method_name>.keep_existing_recovery_conf` is defined and set to ``True``, Patroni will not remove the existing ``recovery.conf`` file. This is useful when bootstrapping from a backup with tools like pgBackRest that generate the appropriate `recovery.conf` for you.
+
+- Allow options to the basebackup built-in method (Oleksii)
+
+  It is now possible to supply options to the built-in basebackup method by defining the `basebackup` section in the configuration, similar to how those are defined for custom replica creation methods. The difference is in the format accepted by the `basebackup` section: since pg_basebackup accepts both `--key=value` and `--key` options, the contents of the section could be either a dictionary of key-value pairs, or a list of either one-element dictionaries or just keys (for the options that don't accept values). See :ref:`replica creation method <custom_replica_creation>` section for additional examples.
+
+
+Version 1.4.3
+-------------
+
+**Improvements in logging**
+
+- Make log level configurable from environment variables (Andy Newton, Keyvan Hedayati)
+
+  `PATRONI_LOGLEVEL` - sets the general logging level
+  `PATRONI_REQUESTS_LOGLEVEL` - sets the logging level for all HTTP requests e.g. Kubernetes API calls
+  See `the docs for Python logging <https://docs.python.org/3.6/library/logging.html#levels>` to get the names of possible log levels
+
+**Stability improvements and bug fixes**
+
+- Don't rediscover etcd cluster topology when watch timed out (Alexander Kukushkin)
+
+  If we have only one host in etcd configuration and exactly this host is not accessible, Patroni was starting discovery of cluster topology and never succeeding. Instead it should just switch to the next available node.
+
+- Write content of bootstrap.pg_hba into a pg_hba.conf after custom bootstrap (Alexander)
+
+  Now it behaves similarly to the usual bootstrap with `initdb`
+
+- Single user mode was waiting for user input and never finish (Alexander)
+
+  Regression was introduced in https://github.com/zalando/patroni/pull/576
+
+
+Version 1.4.2
+-------------
+
+**Improvements in patronictl**
+
+- Rename scheduled failover to scheduled switchover (Alexander Kukushkin)
+
+  Failover and switchover functions were separated in version 1.4, but `patronictl list` was still reporting `Scheduled failover` instead of `Scheduled switchover`.
+
+- Show information about pending restarts (Alexander)
+
+  In order to apply some configuration changes sometimes it is necessary to restart postgres. Patroni was already giving a hint about that in the REST API and when writing node status into DCS, but there were no easy way to display it.
+
+- Make show-config to work with cluster_name from config file (Alexander)
+
+  It works similar to the `patronictl edit-config`
+
+**Stability improvements**
+
+- Avoid calling pg_controldata during bootstrap (Alexander)
+
+  During initdb or custom bootstrap there is a time window when pgdata is not empty but pg_controldata has not been written yet. In such case pg_controldata call was failing with error messages.
+
+- Handle exceptions raised from psutil (Alexander)
+
+  cmdline is read and parsed every time when `cmdline()` method is called. It could happen that the process being examined
+  has already disappeared, in that case `NoSuchProcess` is raised.
+
+**Kubernetes support improvements**
+
+- Don't swallow errors from k8s API (Alexander)
+
+  A call to Kubernetes API could fail for a different number of reasons. In some cases such call should be retried, in some other cases we should log the error message and the exception stack trace. The change here will help debug Kubernetes permission issues.
+
+- Update Kubernetes example Dockerfile to install Patroni from the master branch (Maciej Szulik)
+
+  Before that it was using `feature/k8s`, which became outdated.
+
+- Add proper RBAC to run patroni on k8s (Maciej)
+
+  Add the Service account that is assigned to the pods of the cluster, the role that holds only the necessary permissions, and the rolebinding that connects the Service account and the Role.
+
+
+Version 1.4.1
+-------------
+
+**Fixes in patronictl**
+
+- Don't show current leader in suggested list of members to failover to. (Alexander Kukushkin)
+
+  patronictl failover could still work when there is leader in the cluster and it should be excluded from the list of member where it is possible to failover to.
+
+- Make patronictl switchover compatible with the old Patroni api (Alexander)
+
+  In case if POST /switchover REST API call has failed with status code 501 it will do it once again, but to /failover endpoint.
+
+
+Version 1.4
+-----------
+
+This version adds support for using Kubernetes as a DCS, allowing to run Patroni as a cloud-native agent in Kubernetes without any additional deployments of Etcd, Zookeeper or Consul.
+
+**Upgrade notice**
+
+Installing Patroni via pip will no longer bring in dependencies for (such as libraries for Etcd, Zookeper, Consul or Kubernetes, or support for AWS). In order to enable them one need to list them in pip install command explicitely, for instance `pip install patroni[etcd,kubernetes]`.
+
+**Kubernetes support**
+
+Implement Kubernetes-based DCS. The endpoints meta-data is used in order to store the configuration and the leader key. The meta-data field inside the pods definition is used to store the member-related data.
+In addition to using Endpoints, Patroni supports ConfigMaps. You can find more information about this feature in the :ref:`Kubernetes chapter of the documentation <kubernetes>`
+
+**Stability improvements**
+
+- Factor out postmaster process into a separate object (Ants Aasma)
+
+  This object identifies a running postmaster process via pid and start time and simplifies detection (and resolution) of situations when the postmaster was restarted behind our back or when postgres directory disappeared from the file system.
+
+- Minimize the amount of SELECT's issued by Patroni on every loop of HA cylce (Alexander Kukushkin)
+
+  On every iteration of HA loop Patroni needs to know recovery status and absolute wal position. From now on Patroni will run only single SELECT to get this information instead of two on the replica and three on the master.
+
+- Remove leader key on shutdown only when we have the lock (Ants)
+
+  Unconditional removal was generating unnecessary and missleading exceptions.
+
+**Improvements in patronictl**
+
+- Add version command to patronictl (Ants)
+
+  It will show the version of installed Patroni and versions of running Patroni instances (if the cluster name is specified).
+
+- Make optional specifying cluster_name argument for some of patronictl commands (Alexander, Ants)
+
+  It will work if patronictl is using usual Patroni configuration file with the ``scope`` defined.
+
+- Show information about scheduled switchover and maintenance mode (Alexander)
+
+  Before that it was possible to get this information only from Patroni logs or directly from DCS.
+
+- Improve ``patronictl reinit`` (Alexander)
+
+  Sometimes ``patronictl reinit`` refused to proceed when Patroni was busy with other actions, namely trying to start postgres. `patronictl` didn't provide any commands to cancel such long running actions and the only (dangerous) workarond was removing a data directory manually. The new implementation of `reinit` forcefully cancells other long-running actions before proceeding with reinit.
+
+- Implement ``--wait`` flag in ``patronictl pause`` and ``patronictl resume`` (Alexander)
+
+  It will make ``patronictl`` wait until the requested action is acknowledged by all nodes in the cluster.
+  Such behaviour is achieved by exposing the ``pause`` flag for every node in DCS and via the REST API.
+
+- Rename ``patronictl failover`` into ``patronictl switchover`` (Alexander)
+
+  The previous ``failover`` was actually only capable of doing a switchover; it refused to proceed in a cluster without the leader.
+
+- Alter the behavior of ``patronictl failover`` (Alexander)
+
+  It will work even if there is no leader, but in that case you will have to explicitely specify a node which should become the new leader.
+
+**Expose information about timeline and history**
+
+- Expose current timeline in DCS and via API (Alexander)
+
+  Store information about the current timeline for each member of the cluster. This information is accessible via the API and is stored in the DCS
+
+- Store promotion history in the /history key in DCS (Alexander)
+
+  In addition, store the timeline history enriched with the timestamp of the corresponding promotion in the /history key in DCS and update it with each promote.
+
+**Add endpoints for getting synchronous and asynchronous replicas**
+
+- Add new /sync and /async endpoints (Alexander, Oleksii Kliukin)
+
+ Those endpoints (also accessible as /synchronous and /asynchronous) return 200 only for synchronous and asynchornous replicas correspondingly (exclusing those marked as `noloadbalance`).
+
+**Allow multiple hosts for Etcd**
+
+- Add a new `hosts` parameter to Etcd configuration (Alexander)
+
+  This parameter should contain the initial list of hosts that will be used to discover and populate the list of the running etcd cluster members. If for some reason during work this list of discovered hosts is exhausted (no available hosts from that list), Patroni will return to the initial list from the `hosts` parameter.
+
+
+Version 1.3.6
+-------------
+
+**Stability improvements**
+
+- Verify process start time when checking if postgres is running. (Ants Aasma)
+
+  After a crash that doesn't clean up postmaster.pid there could be a new process with the same pid, resulting in a false positive for is_running(), which will lead to all kinds of bad behavior.
+
+- Shutdown postgresql before bootstrap when we lost data directory (ainlolcat)
+
+  When data directory on the master is forcefully removed, postgres process can still stay alive for some time and prevent the replica created in place of that former master from starting or replicating.
+  The fix makes Patroni cache the postmaster pid and its start time and let it terminate the old postmaster in case it is still running after the corresponding data directory has been removed.
+
+- Perform crash recovery in a single user mode if postgres master dies (Alexander Kukushkin)
+
+  It is unsafe to start immediately as a standby and not possible to run ``pg_rewind`` if postgres hasn't been shut down cleanly.
+  The single user crash recovery only kicks in if ``pg_rewind`` is enabled or there is no master at the moment.
+
+**Consul improvements**
+
+- Make it possible to provide datacenter configuration for Consul (Vilius Okockis, Alexander)
+
+  Before that Patroni was always communicating with datacenter of the host it runs on.
+
+- Always send a token in X-Consul-Token http header (Alexander)
+
+  If ``consul.token`` is defined in Patroni configuration, we will always send it in the 'X-Consul-Token' http header.
+  python-consul module tries to be "consistent" with Consul REST API, which doesn't accept token as a query parameter for `session API <https://www.consul.io/api/session.html>`__, but it still works with 'X-Consul-Token' header.
+
+- Adjust session TTL if supplied value is smaller than the minimum possible (Stas Fomin, Alexander)
+
+  It could happen that the TTL provided in the Patroni configuration is smaller than the minimum one supported by Consul. In that case, Consul agent fails to create a new session.
+  Without a session Patroni cannot create member and leader keys in the Consul KV store, resulting in an unhealthy cluster.
+
+**Other improvements**
+
+- Define custom log format via environment variable ``PATRONI_LOGFORMAT`` (Stas)
+
+  Allow disabling timestamps and other similar fields in Patroni logs if they are already added by the system logger (usually when Patroni runs as a service).
+
+Version 1.3.5
+-------------
+
+**Bugfix**
+
+- Set role to 'uninitialized' if data directory was removed (Alexander Kukushkin)
+
+  If the node was running as a master it was preventing from failover.
+
+**Stability improvement**
+
+- Try to run postmaster in a single-user mode if we tried and failed to start postgres (Alexander)
+
+  Usually such problem happens when node running as a master was terminated and timelines were diverged.
+  If ``recovery.conf`` has ``restore_command`` defined, there are really high chances that postgres will abort startup and leave controldata unchanged.
+  It makes impossible to use ``pg_rewind``, which requires a clean shutdown.
+
+**Consul improvements**
+
+- Make it possible to specify health checks when creating session (Alexander)
+
+  If not specified, Consul will use "serfHealth". From one side it allows fast detection of isolated master, but from another side it makes it impossible for Patroni to tolerate short network lags.
+
+**Bugfix**
+
+- Fix watchdog on Python 3 (Ants Aasma)
+
+  A misunderstanding of the ioctl() call interface. If mutable=False then fcntl.ioctl() actually returns the arg buffer back.
+  This accidentally worked on Python2 because int and str comparison did not return an error.
+  Error reporting is actually done by raising IOError on Python2 and OSError on Python3.
+
+Version 1.3.4
+-------------
+
+**Different Consul improvements**
+
+- Pass the consul token as a header (Andrew Colin Kissa)
+
+  Headers are now the prefered way to pass the token to the consul `API <https://www.consul.io/api/index.html#authentication>`__.
+
+
+- Advanced configuration for Consul (Alexander Kukushkin)
+
+  possibility to specify ``scheme``, ``token``, client and ca certificates :ref:`details <consul_settings>`.
+
+- compatibility with python-consul-0.7.1 and above (Alexander)
+
+  new python-consul module has changed signature of some methods
+
+- "Could not take out TTL lock" message was never logged (Alexander)
+
+  Not a critical bug, but lack of proper logging complicates investigation in case of problems.
+
+
+**Quote synchronous_standby_names using quote_ident**
+
+- When writing ``synchronous_standby_names`` into the ``postgresql.conf`` its value must be quoted (Alexander)
+
+  If it is not quoted properly, PostgreSQL will effectively disable synchronous replication and continue to work.
+
+
+**Different bugfixes around pause state, mostly related to watchdog** (Alexander)
+
+- Do not send keepalives if watchdog is not active
+- Avoid activating watchdog in a pause mode
+- Set correct postgres state in pause mode
+- Do not try to run queries from API if postgres is stopped
+
+
+Version 1.3.3
+-------------
+
+**Bugfixes**
+
+- synchronous replication was disabled shortly after promotion even when synchronous_mode_strict was turned on (Alexander Kukushkin)
+- create empty ``pg_ident.conf`` file if it is missing after restoring from the backup (Alexander)
+- open access in ``pg_hba.conf`` to all databases, not only postgres (Franco Bellagamba)
+
+
+Version 1.3.2
+-------------
+
+**Bugfix**
+
+- patronictl edit-config didn't work with ZooKeeper (Alexander Kukushkin)
+
+
+Version 1.3.1
+-------------
+
+**Bugfix**
+
+- failover via API was broken due to change in ``_MemberStatus`` (Alexander Kukushkin)
+
+
+Version 1.3
+-----------
+
+Version 1.3 adds custom bootstrap possibility, significantly improves support for pg_rewind, enhances the
+synchronous mode support, adds configuration editing to patronictl and implements watchdog support on Linux.
+In addition, this is the first version to work correctly with PostgreSQL 10.
+
+**Upgrade notice**
+
+There are no known compatibility issues with the new version of Patroni. Configuration from version 1.2 should work
+without any changes. It is possible to upgrade by installing new packages and either  restarting Patroni (will cause
+PostgreSQL restart), or by putting Patroni into a :ref:`pause mode <pause>` first and then restarting Patroni on all
+nodes in the cluster (Patroni in a pause mode will not attempt to stop/start PostgreSQL), resuming from the pause mode
+at the end.
+
+**Custom bootstrap**
+
+- Make the process of bootstrapping the cluster configurable (Alexander Kukushkin)
+
+  Allow custom bootstrap scripts instead of ``initdb`` when initializing the very first node in the cluster.
+  The bootstrap command receives the name of the cluster and the path to the data directory. The resulting cluster can
+  be configured to perform recovery, making it possible to bootstrap from a backup and do point in time recovery. Refer
+  to the :ref:`documentaton page <custom_bootstrap>` for more detailed description of this feature.
+
+**Smarter pg_rewind support**
+
+-  Decide on whether to run pg_rewind by looking at the timeline differences from the current master (Alexander)
+
+   Previously, Patroni had a fixed set of conditions to trigger pg_rewind, namely when starting a former master, when
+   doing a switchover to the designated node for every other node in the cluster or when there is a replica with the
+   nofailover tag. All those cases have in common a chance that some replica may be ahead of the new master. In some cases,
+   pg_rewind did nothing, in some other ones it was not running when necessary. Instead of relying on this limited list
+   of rules make Patroni compare the master and the replica WAL positions (using the streaming replication protocol)
+   in order to reliably decide if rewind is necessary for the replica.
+
+**Synchronous replication mode strict**
+
+-  Enhance synchronous replication support by adding the strict mode (James Sewell, Alexander)
+
+   Normally, when ``synchronous_mode`` is enabled and there are no replicas attached to the master, Patroni will disable
+   synchronous replication in order to keep the master available for writes. The ``synchronous_mode_strict`` option
+   changes that, when it is set Patroni will not disable the synchronous replication in a lack of replicas, effectively
+   blocking all clients writing data to the master. In addition to the synchronous mode guarantee of preventing any data
+   loss due to automatic failover, the strict mode ensures that each write is either durably stored on two nodes or not
+   happening altogether if there is only one node in the cluster.
+
+**Configuration editing with patronictl**
+
+- Add configuration editing to patronictl (Ants Aasma, Alexander)
+
+  Add the ability to patronictl of editing dynamic cluster configuration stored in DCS. Support either specifying the
+  parameter/values from the command-line, invoking the $EDITOR, or applying configuration from the yaml file.
+
+**Linux watchdog support**
+
+- Implement watchdog support for Linux (Ants)
+
+  Support Linux software watchdog in order to reboot the node where Patroni is not running or not responding (e.g because
+  of the high load) The Linux software watchdog reboots the non-responsive node. It is possible to configure the watchdog
+  device to use (`/dev/watchdog` by default) and the mode (on, automatic, off) from the watchdog section of the Patroni
+  configuration. You can get more information from the :ref:`watchdog documentation <watchdog>`.
+
+**Add support for PostgreSQL 10**
+
+- Patroni is compatible with all beta versions of PostgreSQL 10 released so far and we expect it to be compatible with
+  the PostgreSQL 10 when it will be released.
+
+**PostgreSQL-related minor improvements**
+
+- Define pg_hba.conf via the Patroni configuration file or the dynamic configuration in DCS (Alexander)
+
+  Allow to define the contents of ``pg_hba.conf`` in the ``pg_hba`` sub-section of the ``postgresql`` section of the
+  configuration. This simplifies managing ``pg_hba.conf`` on multiple nodes, as one needs to define it only ones in DCS
+  instead of logging to every node, changing it manually and reload the configuration.
+
+  When defined, the contents of this section will replace the current ``pg_hba.conf`` completely. Patroni ignores it
+  if ``hba_file`` PostgreSQL parameter is set.
+
+- Support connecting via a UNIX socket to the local PostgreSQL cluster (Alexander)
+
+  Add the ``use_unix_socket`` option to the ``postgresql`` section of Patroni configuration. When set to true and the
+  PostgreSQL ``unix_socket_directories`` option is not empty, enables Patroni to use the first value from it to connect
+  to the local PostgreSQL cluster. If ``unix_socket_directories`` is not defined, Patroni will assume its default value
+  and omit the ``host`` parameter in the PostgreSQL connection string altogether.
+
+- Support change of superuser and replication credentials on reload (Alexander)
+
+- Support storing of configuration files outside of PostgreSQL data directory (@jouir)
+
+  Add the new configuration ``postgresql`` configuration directive ``config_dir``.
+  It defaults to the data directory and must be writable by Patroni.
+
+**Bug fixes and stability improvements**
+
+- Handle EtcdEventIndexCleared and EtcdWatcherCleared exceptions (Alexander)
+
+  Faster recovery when the watch operation is ended by Etcd by avoiding useless retries.
+
+- Remove error spinning on Etcd failure and reduce log spam (Ants)
+
+  Avoid immediate retrying and emitting stack traces in the log on the second and subsequent Etcd connection failures.
+
+- Export locale variables when forking PostgreSQL processes (Oleksii Kliukin)
+
+  Avoid the `postmaster became multithreaded during startup` fatal error on non-English locales for PostgreSQL built with NLS.
+
+- Extra checks when dropping the replication slot (Alexander)
+
+  In some cases Patroni is prevented from dropping the replication slot by the WAL sender.
+
+- Truncate the replication slot name to 63  (NAMEDATALEN - 1) characters to comply with PostgreSQL naming rules (Nick Scott)
+
+- Fix a race condition resulting in extra connections being opened to the PostgreSQL cluster from Patroni (Alexander)
+
+- Release the leader key when the node restarts with an empty data directory (Alex Kerney)
+
+- Set asynchronous executor busy when running bootstrap without a leader (Alexander)
+
+  Failure to do so could have resulted in errors stating the node belonged to a different cluster, as Patroni proceeded with
+  the normal business while being bootstrapped by a bootstrap method that doesn't require a leader to be present in the
+  cluster.
+
+- Improve WAL-E replica creation method (Joar Wandborg, Alexander).
+
+  - Use csv.DictReader when parsing WAL-E base backup, accepting ISO dates with space-delimited date and time.
+  - Support fetching current WAL position from the replica to estimate the amount of WAL to restore. Previously, the code used to call system information functions that were available only on the master node.
+
+
 Version 1.2
 -----------
 
@@ -391,4 +1318,4 @@ This release adds support for *cascading replication* and simplifies Patroni man
 
   The tests can be launched manually using the *behave* command. They are also launched automatically for pull requests and after commits.
 
-  Releases notes for some older versions can be found on `project's github page <https://github.com/zalando/patroni/releases>`__.
+  Release notes for some older versions can be found on `project's github page <https://github.com/zalando/patroni/releases>`__.

docs/replica_bootstrap.rst

@@ -0,0 +1,207 @@
+Replica imaging and bootstrap
+=============================
+
+Patroni allows customizing creation of a new replica. It also supports defining what happens when the new empty cluster
+is being bootstrapped. The distinction between two is well defined: Patroni creates replicas only if the ``initialize``
+key is present in DCS for the cluster. If there is no ``initialize`` key - Patroni calls bootstrap exclusively on the
+first node that takes the initialize key lock.
+
+.. _custom_bootstrap:
+
+Bootstrap
+---------
+
+PostgreSQL provides ``initdb`` command to initialize a new cluster and Patroni calls it by default. In certain cases,
+particularly when creating a new cluster as a copy of an existing one, it is necessary to replace a built-in method with
+custom actions. Patroni supports executing user-defined scripts to bootstrap new clusters, supplying some required
+arguments to them, i.e. the name of the cluster and the path to the data directory. This is configured in the
+``bootstrap`` section of the Patroni configuration. For example:
+
+.. code:: YAML
+
+    bootstrap:
+        method: <custom_bootstrap_method_name>
+        <custom_bootstrap_method_name>:
+            command: <path_to_custom_bootstrap_script> [param1 [, ...]]
+            keep_existing_recovery_conf: False
+            recovery_conf:
+                recovery_target_action: promote
+                recovery_target_timeline: latest
+                restore_command: <method_specific_restore_command>
+
+
+Each bootstrap method must define at least a ``name`` and a ``command``. A special ``initdb`` method is available to trigger
+the default behavior, in which case ``method`` parameter can be omitted altogether. The ``command`` can be specified using either
+an absolute path, or the one relative to the ``patroni`` command location. In addition to the fixed parameters defined
+in the configuration files, Patroni supplies two cluster-specific ones:
+
+--scope
+    Name of the cluster to be bootstrapped
+--datadir
+    Path to the data directory of the cluster instance to be bootstrapped
+
+If the bootstrap script returns 0, Patroni tries to configure and start the PostgreSQL instance produced by it. If any
+of the intermediate steps fail, or the script returns a non-zero value, Patroni assumes that the bootstrap has failed,
+cleans up after itself and releases the initialize lock to give another node the opportunity to bootstrap.
+
+If a ``recovery_conf`` block is defined in the same section as the custom bootstrap method, Patroni will generate a
+``recovery.conf`` before starting the newly bootstrapped instance. Typically, such recovery.conf should contain at least
+one of the ``recovery_target_*`` parameters, together with the ``recovery_target_timeline`` set to ``promote``.
+
+If ``keep_existing_recovery_conf`` is defined and set to ``True``, Patroni will not remove the existing ``recovery.conf`` file if it exists.
+This is useful when bootstrapping from a backup with tools like pgBackRest that generate the appropriate ``recovery.conf`` for you.
+
+ .. note:: Bootstrap methods are neither chained, nor fallen-back to the default one in case the primary one fails
+
+
+.. _custom_replica_creation:
+
+Building replicas
+-----------------
+
+Patroni uses tried and proven ``pg_basebackup`` in order to create new replicas. One downside of it is that it requires
+a running master node. Another one is the lack of 'on-the-fly' compression for the backup data and no built-in cleanup
+for outdated backup files. Some people prefer other backup solutions, such as ``WAL-E``, ``pgBackRest``, ``Barman`` and
+others, or simply roll their own scripts. In order to accommodate all those use-cases Patroni supports running custom
+scripts to clone a new replica. Those are configured in the ``postgresql`` configuration block:
+
+.. code:: YAML
+
+    postgresql:
+        create_replica_methods:
+            - <method name>
+        <method name>:
+            command: <command name>
+            keep_data: True
+            no_params: True
+            no_master: 1
+
+example: wal_e
+
+.. code:: YAML
+
+    postgresql:
+        create_replica_methods:
+            - wal_e
+            - basebackup
+        wal_e:
+            command: patroni_wale_restore
+            no_master: 1
+            envdir: {{WALE_ENV_DIR}}
+            use_iam: 1
+        basebackup:
+            max-rate: '100M'
+
+example: pgbackrest
+
+.. code:: YAML
+
+    postgresql:
+        create_replica_methods:
+            - pgbackrest
+            - basebackup
+        pgbackrest:
+            command: /usr/bin/pgbackrest --stanza=<scope> --delta restore
+            keep_data: True
+            no_params: True
+        basebackup:
+            max-rate: '100M'
+
+
+The ``create_replica_methods`` defines available replica creation methods and the order of executing them. Patroni will
+stop on the first one that returns 0. Each method should define a separate section in the configuration file, listing the command
+to execute and any custom parameters that should be passed to that command. All parameters will be passed in a
+``--name=value`` format. Besides user-defined parameters, Patroni supplies a couple of cluster-specific ones:
+
+--scope
+    Which cluster this replica belongs to
+--datadir
+    Path to the data directory of the replica
+--role
+    Always 'replica'
+--connstring
+    Connection string to connect to the cluster member to clone from (master or other replica). The user in the
+    connection string can execute SQL and replication protocol commands.
+
+A special ``no_master`` parameter, if defined, allows Patroni to call the replica creation method even if there is no
+running master or replicas. In that case, an empty string will be passed in a connection string. This is useful for
+restoring the formerly running cluster from the binary backup.
+
+A special ``keep_data`` parameter, if defined, will instuct Patroni to  not clean PGDATA folder before calling restore.
+
+A special ``no_params`` parameter, if defined, restricts passing parameters to custom command.
+
+A ``basebackup`` method is a special case: it will be used if
+``create_replica_methods`` is empty, although it is possible
+to list it explicitly among the ``create_replica_methods`` methods. This method initializes a new replica with the
+``pg_basebackup``, the base backup is taken from the master unless there are replicas with ``clonefrom`` tag, in which case one
+of such replicas will be used as the origin for pg_basebackup. It works without any configuration; however, it is
+possible to specify a ``basebackup`` configuration section. Same rules as with the other method configuration apply,
+namely, only long (with --) options should be specified there. Not all parameters make sense, if you override a connection
+string or provide an option to created tar-ed or compressed base backups, patroni won't be able to make a replica out
+of it. There is no validation performed on the names or values of the parameters passed to the ``basebackup`` section.
+You can specify basebackup parameters as either a map (key-value pairs) or a list of elements, where each element
+could be either a key-value pair or a single key (for options that does not receive any values, for instance, ``--verbose``).
+Consider those 2 examples:
+
+.. code:: YAML
+
+    postgresql:
+        basebackup:
+            max-rate: '100M'
+            checkpoint: 'fast'
+
+and
+
+.. code:: YAML
+
+    postgresql:
+        basebackup:
+            - verbose
+            - max-rate: '100M'
+
+If all replica creation methods fail, Patroni will try again all methods in order during the next event loop cycle.
+
+.. _standby_cluster:
+
+Standby cluster
+---------------
+
+Another available option is to run a "standby cluster", that contains only of
+standby nodes replicating from some remote master. This type of clusters has:
+
+* "standby leader", that behaves pretty much like a regular cluster leader,
+  except it replicates from a remote master.
+
+* cascade replicas, that are replicating from standby leader.
+
+Standby leader holds and updates a leader lock in DCS. If the leader lock
+expires, cascade replicas will perform an election to choose another leader
+from the standbys.
+
+For the sake of flexibility, you can specify methods of creating a replica and
+recovery WAL records when a cluster is in the "standby mode" by providing
+`create_replica_methods` key in `standby_cluster` section. It is distinct from
+creating replicas, when cluster is detached and functions as a normal cluster,
+which is controlled by `create_replica_methods` in `postgresql` section. Both
+"standby" and "normal" `create_replica_methods` reference  keys in `postgresql`
+section.
+
+To configure such cluster you need to specify the section ``standby_cluster``
+in a patroni configuration:
+
+.. code:: YAML
+
+    bootstrap:
+        dcs:
+            standby_cluster:
+                host: 1.2.3.4
+                port: 5432
+                primary_slot_name: patroni
+                create_replica_methods:
+                - basebackup
+
+Note, that these options will be applied only once during cluster bootstrap,
+and the only way to change them afterwards is through DCS.
+
+If you use replication slots on the standby cluster, you must also create the corresponding replication slot on the primary cluster.  It will not be done automatically by the standby cluster implementation.  You can use Patroni's permenant replication slots feature on the primary cluster to maintain a replication slot with the same name as ``primary_slot_name``, or its default value if ``primary_slot_name`` is not provided.

docs/replication_modes.rst

@@ -9,9 +9,11 @@ Patroni uses PostgreSQL streaming replication. For more information about stream
 Asynchronous mode durability
 ----------------------------
 
-In asynchronous mode the cluster is allowed to lose some committed transactions to ensure availability. When master server fails or becomes unavailable for any other reason Patroni will automatically promote a sufficiently healthy standby to master. Any transactions that have not been replicated to that standby remain in a "forked timeline" on the master, and are effectively unrecoverable [1]_.
+In asynchronous mode the cluster is allowed to lose some committed transactions to ensure availability. When the primary server fails or becomes unavailable for any other reason Patroni will automatically promote a sufficiently healthy standby to primary. Any transactions that have not been replicated to that standby remain in a "forked timeline" on the primary, and are effectively unrecoverable [1]_.
 
-The amount of transactions that can be lost is controlled via ``maximum_lag_on_failover`` parameter. Because master transaction log position is not sampled in real time, in reality the amount of lost data on failover is worst case bounded by  ``maximum_lag_on_failover`` bytes of transaction log plus the amount that is written in the last ``ttl`` seconds (``loop_wait``/2 seconds in the average case). However typical steady state replication delay is well under a second.
+The amount of transactions that can be lost is controlled via ``maximum_lag_on_failover`` parameter. Because the primary transaction log position is not sampled in real time, in reality the amount of lost data on failover is worst case bounded by  ``maximum_lag_on_failover`` bytes of transaction log plus the amount that is written in the last ``ttl`` seconds (``loop_wait``/2 seconds in the average case). However typical steady state replication delay is well under a second.
+
+By default, when running leader elections, Patroni does not take into account the current timeline of replicas, what in some cases could be undesirable behavior. You can prevent the node not having the same timeline as a former master become the new leader by changing the value of ``check_timeline`` parameter to ``true``.
 
 PostgreSQL synchronous replication
 ----------------------------------
@@ -20,7 +22,7 @@ You can use Postgres's `synchronous replication <http://www.postgresql.org/docs/
 
 In hosted datacenter environments (like AWS, Rackspace, or any network you do not control), synchronous replication significantly increases the variability of write performance. If followers become inaccessible from the leader, the leader effectively becomes read-only.
 
-To enable a simple synchronous replication test, add the follow lines to the ``parameters`` section of your YAML configuration files:
+To enable a simple synchronous replication test, add the following lines to the ``parameters`` section of your YAML configuration files:
 
 .. code:: YAML
 
@@ -29,28 +31,37 @@ To enable a simple synchronous replication test, add the follow lines to the ``p
 
 When using PostgreSQL synchronous replication, use at least three Postgres data nodes to ensure write availability if one host fails.
 
-Using PostgreSQL synchronous replication does not guarantee zero lost transactions under all circumstances. When master and standby that is currently acting as synchronous fail simultaneously a third node that might not contain all transactions will be promoted.
+Using PostgreSQL synchronous replication does not guarantee zero lost transactions under all circumstances. When the primary and the secondary that is currently acting as a synchronous replica fail simultaneously a third node that might not contain all transactions will be promoted.
 
 .. _synchronous_mode:
 
 Synchronous mode
 ----------------
 
-For use cases where losing committed transactions is not permissible you can turn on Patronis ``synchronous_mode``. When ``synchronous_mode`` is turned on Patroni will not promote a standby unless it is certain that the standby contains all transactions that may have returned a successful commit status to client [2]_. This means that the system may be unavailable for writes even though some servers are available. System administrators can still use manual failover commmands to promote a standby even if it results in transaction loss.
+For use cases where losing committed transactions is not permissible you can turn on Patroni's ``synchronous_mode``. When ``synchronous_mode`` is turned on Patroni will not promote a standby unless it is certain that the standby contains all transactions that may have returned a successful commit status to client [2]_. This means that the system may be unavailable for writes even though some servers are available. System administrators can still use manual failover commands to promote a standby even if it results in transaction loss.
 
-Turning on ``synchronous_mode`` does not guarantee multi node durability of commits under all circumstances. When no suitable standby is available, master server will still accept writes, but does not guarantee their replication. When the master fails in this mode no standby will be promote.  When the host that used to be master comes back it will get promoted automatically, unless system administrator performed a manual failover. This behavior makes synchronous mode usable with 2 node clusters.
+Turning on ``synchronous_mode`` does not guarantee multi node durability of commits under all circumstances. When no suitable standby is available, primary server will still accept writes, but does not guarantee their replication. When the primary fails in this mode no standby will be promoted. When the host that used to be the primary comes back it will get promoted automatically, unless system administrator performed a manual failover. This behavior makes synchronous mode usable with 2 node clusters.
 
-When ``synchronous_mode`` is on and a standby crashes, commits will block until next iteration of Patroni runs and switches master to standalone mode (worst case delay for writes ``ttl`` seconds, average case ``loop_wait``/2 seconds). Manually shutting down or restarting a standby will not cause a commit service interruption. Standby will signal the master to release itself from synchronous standby duties before PostgreSQL shutdown is initiated.
+When ``synchronous_mode`` is on and a standby crashes, commits will block until next iteration of Patroni runs and switches the primary to standalone mode (worst case delay for writes ``ttl`` seconds, average case ``loop_wait``/2 seconds). Manually shutting down or restarting a standby will not cause a commit service interruption. Standby will signal the primary to release itself from synchronous standby duties before PostgreSQL shutdown is initiated.
+
+When it is absolutely necessary to guarantee that each write is stored durably
+on at least two nodes, enable ``synchronous_mode_strict`` in addition to the
+``synchronous_node``. This parameter prevents Patroni from switching off the
+synchronous replication on the primary when no synchronous standby candidates
+are available. As a downside, the primary is not be available for writes
+(unless the Postgres transaction explicitly turns of ``synchronous_mode``),
+blocking all client write requests until at least one synchronous replica comes
+up.
 
 You can ensure that a standby never becomes the synchronous standby by setting ``nosync`` tag to true. This is recommended to set for standbys that are behind slow network connections and would cause performance degradation when becoming a synchronous standby.
 
-Synchronous mode can be switched on and off via Patroni REST interface. See `dynamic configuration <https://github.com/zalando/patroni/blob/master/docs/dynamic_configuration.rst>`__ for instructions.
+Synchronous mode can be switched on and off via Patroni REST interface. See :ref:`dynamic configuration <dynamic_configuration>` for instructions.
 
 
 Synchronous mode implementation
 -------------------------------
 
-When in synchronous mode Patroni maintains synchronization state in the DCS, containing the latest master and current synchronous standby. This state is updated with strict ordering constraints to ensure the following invariants:
+When in synchronous mode Patroni maintains synchronization state in the DCS, containing the latest primary and current synchronous standby. This state is updated with strict ordering constraints to ensure the following invariants:
 
 - A node must be marked as the latest leader whenever it can accept write transactions. Patroni crashing or PostgreSQL not shutting down can cause violations of this invariant.
 
@@ -60,9 +71,9 @@ When in synchronous mode Patroni maintains synchronization state in the DCS, con
 
 Patroni will only ever assign one standby to ``synchronous_standby_names`` because with multiple candidates it is not possible to know which node was acting as synchronous during the failure.
 
-On each HA loop iteration Patroni re-evaluates synchronous standby choice. If the current synchronous standby is connected and has not requested its synchronous status to be removed it remains picked. Otherwise the cluster member avaiable for sync that is furthest ahead in replication is picked.
+On each HA loop iteration Patroni re-evaluates synchronous standby choice. If the current synchronous standby is connected and has not requested its synchronous status to be removed it remains picked. Otherwise the cluster member available for sync that is furthest ahead in replication is picked.
 
 
-.. [1] The data is still there, but recovering it requires a manual recovery effort by data recovery specialists. When Patroni is allowed to rewind with ``use_pg_rewind`` the forked timeline will be automatically erased to rejoin the failed master with the cluster.
+.. [1] The data is still there, but recovering it requires a manual recovery effort by data recovery specialists. When Patroni is allowed to rewind with ``use_pg_rewind`` the forked timeline will be automatically erased to rejoin the failed primary with the cluster.
 
-.. [2] Clients can change the behavior per transaction using PostgreSQL's ``synchronous_commit`` setting. Transactions with ``synchronous_commit`` values of ``off`` and ``local`` may be lost on fail over, but will not be blocked by replication delays.
+.. [2] Clients can change the behavior per transaction using PostgreSQL's ``synchronous_commit`` setting. Transactions with ``synchronous_commit`` values of ``off`` and ``local`` may be lost on fail over, but will not be blocked by replication delays.

docs/watchdog.rst

@@ -0,0 +1,39 @@
+.. _watchdog:
+
+Watchdog support
+================
+
+Having multiple PostgreSQL servers running as master can result in transactions lost due to diverging timelines. This situation is also called a split-brain problem. To avoid split-brain Patroni needs to ensure PostgreSQL will not accept any transaction commits after leader key expires in the DCS. Under normal circumstances Patroni will try to achieve this by stopping PostgreSQL when leader lock update fails for any reason. However, this may fail to happen due to various reasons:
+
+- Patroni has crashed due to a bug, out-of-memory condition or by being accidentally killed by a system administrator.
+
+- Shutting down PostgreSQL is too slow.
+
+- Patroni does not get to run due to high load on the system, the VM being paused by the hypervisor, or other infrastructure issues.
+
+To guarantee correct behavior under these conditions Patroni supports watchdog devices. Watchdog devices are software or hardware mechanisms that will reset the whole system when they do not get a keepalive heartbeat within a specified timeframe. This adds an additional layer of fail safe in case usual Patroni split-brain protection mechanisms fail.
+
+Patroni will try to activate the watchdog before promoting PostgreSQL to master. If watchdog activation fails and watchdog mode is ``required`` then the node will refuse to become master. When deciding to participate in leader election Patroni will also check that watchdog configuration will allow it to become leader at all. After demoting PostgreSQL (for example due to a manual failover) Patroni will disable the watchdog again. Watchdog will also be disabled while Patroni is in paused state.
+
+By default Patroni will set up the watchdog to expire 5 seconds before TTL expires. With the default setup of ``loop_wait=10`` and ``ttl=30`` this gives HA loop at least 15 seconds (``ttl`` - ``safety_margin`` - ``loop_wait``) to complete before the system gets forcefully reset. By default accessing DCS is configured to time out after 10 seconds. This means that when DCS is unavailable, for example due to network issues, Patroni and PostgreSQL will have at least 5 seconds (``ttl`` - ``safety_margin`` - ``loop_wait`` - ``retry_timeout``) to come to a state where all client connections are terminated.
+
+Safety margin is the amount of time that Patroni reserves for time between leader key update and watchdog keepalive. Patroni will try to send a keepalive immediately after confirmation of leader key update. If Patroni process is suspended for extended amount of time at exactly the right moment the keepalive may be delayed for more than the safety margin without triggering the watchdog. This results in a window of time where watchdog will not trigger before leader key expiration, invalidating the guarantee. To be absolutely sure that watchdog will trigger under all circumstances set up the watchdog to expire after half of TTL by setting ``safety_margin`` to -1 to set watchdog timeout to ``ttl // 2``. If you need this guarantee you probably should increase ``ttl`` and/or reduce ``loop_wait`` and ``retry_timeout``.
+
+Currently watchdogs are only supported using Linux watchdog device interface.
+
+Setting up software watchdog on Linux
+-------------------------------------
+
+Default Patroni configuration will try to use ``/dev/watchdog`` on Linux if it is accessible to Patroni. For most use cases using software watchdog built into the Linux kernel is secure enough.
+
+To enable software watchdog issue the following commands as root before starting Patroni:
+
+.. code-block:: bash
+
+    modprobe softdog
+    # Replace postgres with the user you will be running patroni under
+    chown postgres /dev/watchdog
+
+For testing it may be helpful to disable rebooting by adding ``soft_noboot=1`` to the modprobe command line. In this case the watchdog will just log a line in kernel ring buffer, visible via `dmesg`.
+
+Patroni will log information about the watchdog when it is successfully enabled.

extras/README.md

@@ -1,11 +1,11 @@
 ### confd
 
-`confd` directory contains haproxy template files for the [confd](https://github.com/kelseyhightower/confd) -- lightweight configuration management tool
+`confd` directory contains haproxy and pgbouncer template files for the [confd](https://github.com/kelseyhightower/confd) -- lightweight configuration management tool
 You need to copy content of `confd` directory into /etcd/confd and run confd service:
 ```bash
-$ confd -prefix=/service/$PATRONI_SCOPE -backend etcd -node $PATRONI_ETCD_HOST -interval=10
+$ confd -prefix=/service/$PATRONI_SCOPE -backend etcd -node $PATRONI_ETCD_URL -interval=10
 ```
-It will periodically update haproxy.cfg with the actual list of Patroni nodes from `etcd` and "reload" haproxy when it is necessary.
+It will periodically update haproxy.cfg and pgbouncer.ini with the actual list of Patroni nodes from `etcd` and "reload" haproxy and pgbouncer.ini when it is necessary.
 
 
 ### startup-scripts

extras/confd/conf.d/pgbouncer.toml

@@ -0,0 +1,12 @@
+[template]
+prefix = "/service/batman"
+owner = "postgres"
+mode = "0644"
+src = "pgbouncer.tmpl"
+dest = "/etc/pgbouncer/pgbouncer.ini"
+
+reload_cmd = "systemctl reload pgbouncer"
+
+keys = [
+    "/members/","/leader"
+]

extras/confd/templates/haproxy.tmpl

@@ -10,19 +10,23 @@ defaults
 	timeout server 30m
 	timeout check 5s
 
-frontend master_postgresql
+listen stats
+	mode http
+	bind *:7000
+	stats enable
+	stats uri /
+
+listen master
 	bind *:5000
-	default_backend backend_master
-
-frontend replicas_postgresql
-	bind *:5001
-	default_backend backend_replicas
-
-backend backend_master
 	option httpchk OPTIONS /master
+	http-check expect status 200
+	default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
 {{range gets "/members/*"}}	server {{base .Key}} {{$data := json .Value}}{{base (replace (index (split $data.conn_url "/") 2) "@" "/" -1)}} maxconn 100 check port {{index (split (index (split $data.api_url "/") 2) ":") 1}}
 {{end}}
-backend backend_replicas
+listen replicas
+	bind *:5001
 	option httpchk OPTIONS /replica
+	http-check expect status 200
+	default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
 {{range gets "/members/*"}}	server {{base .Key}} {{$data := json .Value}}{{base (replace (index (split $data.conn_url "/") 2) "@" "/" -1)}} maxconn 100 check port {{index (split (index (split $data.api_url "/") 2) ":") 1}}
 {{end}}

extras/confd/templates/pgbouncer.tmpl

@@ -0,0 +1,17 @@
+[databases]
+{{with get "/leader"}}{{$leader := .Value}}{{$leadkey := printf "/members/%s" $leader}}{{with get $leadkey}}{{$data := json .Value}}{{$hostport := base (replace (index (split $data.conn_url "/") 2) "@" "/" -1)}}{{ $host := base (index (split $hostport ":") 0)}}{{ $port := base (index (split $hostport ":") 1)}}* = host={{ $host }} port={{ $port }} pool_size=10{{end}}{{end}}
+
+[pgbouncer]
+logfile = /var/log/postgresql/pgbouncer.log
+pidfile = /var/run/postgresql/pgbouncer.pid
+listen_addr = *
+listen_port = 6432
+unix_socket_dir = /var/run/postgresql
+auth_type = trust
+auth_file = /etc/pgbouncer/userlist.txt
+auth_hba_file = /etc/pgbouncer/pg_hba.txt
+admin_users = pgbouncer
+stats_users = pgbouncer
+pool_mode = session
+max_client_conn = 100
+default_pool_size = 20

extras/startup-scripts/README.md

@@ -11,3 +11,12 @@ Upstart job for Ubuntu 12.04 or 14.04.  Requires Upstart > 1.4. Intended for sys
 
 ### patroni.service
 Systemd service file, to be copied to /etc/systemd/system/patroni.service, tested on Centos 7.1 with Patroni installed from pip. 
+
+### patroni
+Init.d service file for Debian-like distributions. Copy it to /etc/init.d/, make executable:
+```chmod 755 /etc/init.d/patroni``` and run with ```service patroni start```, or make it starting on boot with ```update-rc.d patroni defaults```. Also you might edit some configuration variables in it:
+PATRONI for patroni.py location
+CONF for configuration file
+LOGFILE for log (script creates it if does not exist)
+
+Note. If you have several versions of Postgres installed, please add to POSTGRES_VERSION the release number which you wish to run. Script uses this value to append PATH environment with correct path to Postgres bin.

extras/startup-scripts/patroni

@@ -0,0 +1,144 @@
+#!/bin/sh
+#
+### BEGIN INIT INFO
+# Provides:          patroni
+# Required-Start:    $remote_fs $syslog
+# Required-Stop:     $remote_fs $syslog
+# Default-Start:     2 3 4 5
+# Default-Stop:      0 1 6
+# Short-Description: Patroni init script
+# Description:       Runners to orchestrate a high-availability PostgreSQL
+### END INIT INFO
+
+### BEGIN USER CONFIGURATION
+
+CONF="/etc/patroni/postgres.yml"
+LOGFILE="/var/log/patroni.log"
+USER="postgres"
+GROUP="postgres"
+
+NAME=patroni
+PATRONI="/opt/patroni/$NAME.py"
+PIDFILE="/var/run/$NAME.pid"
+
+# Set this parameter, if you have several Postgres versions installed
+# POSTGRES_VERSION="9.4"
+POSTGRES_VERSION=""
+
+### END USER CONFIGURATION
+
+. /lib/lsb/init-functions
+
+# Loading this library for get_versions() function
+if test ! -e /usr/share/postgresql-common/init.d-functions; then
+  log_failure_msg "Probably postgresql-common does not installed."
+  exit 1
+else
+  . /usr/share/postgresql-common/init.d-functions
+fi
+
+# Is there Patroni executable?
+if test ! -e $PATRONI; then
+  log_failure_msg "Patroni executable $PATRONI does not exist."
+  exit 1
+fi
+
+# Is there Patroni configuration file?
+if test ! -e $CONF; then
+  log_failure_msg "Patroni configuration file $CONF does not exist."
+  exit 1
+fi
+
+# Create logfile if doesn't exist
+if test ! -e $LOGFILE; then
+  log_action_msg "Creating logfile for Patroni..."
+  touch $LOGFILE
+  chown $USER:$GROUP $LOGFILE
+fi
+
+prepare_pgpath() {
+  if [ "$POSTGRES_VERSION" != "" ]; then
+    if [ -x /usr/lib/postgresql/$POSTGRES_VERSION/bin/pg_ctl ]; then
+      PGPATH="/usr/lib/postgresql/$POSTGRES_VERSION/bin"
+    else
+      log_failure_msg "Postgres version incorrect, check POSTGRES_VERSION variable."
+      exit 0
+    fi
+  else
+    get_versions
+    if echo $versions | grep -q -e "\s"; then
+      log_warning_msg "You have several Postgres versions installed. Please, use POSTGRES_VERSION to define correct environment."
+    else
+      versions=`echo $versions | sed -e 's/^[ \t]*//'`
+      PGPATH="/usr/lib/postgresql/$versions/bin"
+    fi
+  fi
+}
+
+get_pid() {
+  if test -e $PIDFILE; then
+    PID=`cat $PIDFILE`
+    CHILDPID=`ps --ppid $PID -o %p --no-headers`
+  else
+    log_failure_msg "Could not find PID file. Patroni probably down."
+    exit 1
+  fi
+}
+
+
+case "$1" in
+  start)
+    prepare_pgpath
+    PGPATH=$PATH:$PGPATH
+    log_success_msg "Starting Patroni\n"
+    exec start-stop-daemon --start --quiet \
+        --background \
+        --pidfile $PIDFILE --make-pidfile \
+        --chuid $USER:$GROUP \
+        --chdir `eval echo ~$USER` \
+        --exec $PATRONI \
+        --startas /bin/sh -- \
+        -c "/usr/bin/env PATH=$PGPATH /usr/bin/python $PATRONI $CONF >> $LOGFILE 2>&1"
+    ;;
+
+  stop)
+    log_success_msg "Stopping Patroni"
+    get_pid
+    start-stop-daemon --stop --pid $CHILDPID
+    start-stop-daemon --stop --pidfile $PIDFILE --remove-pidfile --quiet
+    ;;
+
+  reload)
+    log_success_msg "Reloading Patroni configuration"
+    get_pid
+    kill -HUP $CHILDPID
+    ;;
+
+  status)
+    get_pid
+    if start-stop-daemon -T --pid $CHILDPID; then
+      log_success_msg "Patroni is running\n"
+      exit 0
+    else
+      log_warning_msg "Patroni in not running\n"
+    fi
+    ;;
+
+  restart)
+    $0 stop
+    $0 start
+    ;;
+
+  *)
+    echo "Usage: /etc/init.d/$NAME {start|stop|restart|reload|status}"
+    exit 1
+    ;;
+esac
+
+if [ $? -eq 0 ]; then
+        echo .
+        exit 0
+else
+        echo " failed"
+        exit 1
+fi

extras/startup-scripts/patroni.service

@@ -11,17 +11,31 @@ Type=simple
 User=postgres
 Group=postgres
 
+# Read in configuration file if it exists, otherwise proceed
+EnvironmentFile=-/etc/patroni_env.conf
+
+WorkingDirectory=~
+
 # Where to send early-startup messages from the server
 # This is normally controlled by the global default set by systemd
-# StandardOutput=syslog
+#StandardOutput=syslog
 
+# Pre-commands to start watchdog device
+# Uncomment if watchdog is part of your patroni setup
+#ExecStartPre=-/usr/bin/sudo /sbin/modprobe softdog
+#ExecStartPre=-/usr/bin/sudo /bin/chown postgres /dev/watchdog
+
+# Start the patroni process
 ExecStart=/bin/patroni /etc/patroni.yml
 
+# Send HUP to reload from patroni.yml
+ExecReload=/bin/kill -s HUP $MAINPID
+
 # only kill the patroni process, not it's children, so it will gracefully stop postgres
 KillMode=process
 
 # Give a reasonable amount of time for the server to start up/shut down
-TimeoutSec=10
+TimeoutSec=30
 
 # Do not restart the service if it crashes, we want to manually inspect database on failure
 Restart=no

features/backup_create.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+
+while getopts ":-:" optchar; do
+    [[ "${optchar}" == "-" ]] || continue
+    case "${OPTARG}" in
+        datadir=* )
+            PGDATA=${OPTARG#*=}
+            ;;
+        dbname=* )
+            DBNAME=${OPTARG#*=}
+            ;;
+        walmethod=* )
+            WALMETHOD=${OPTARG#*=}
+            ;;
+    esac
+done
+
+[[ -z $PGDATA || -z $DBNAME || -z $WALMETHOD ]] && exit 1
+
+[[ $WALMETHOD != "none" ]] && WALMETHOD="-X $WALMETHOD" || WALMETHOD=""
+
+exec pg_basebackup -D $PGDATA $WALMETHOD -c fast -d $DBNAME

features/backup_restore.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+set -x
+
+while getopts ":-:" optchar; do
+    [[ "${optchar}" == "-" ]] || continue
+    case "${OPTARG}" in
+        datadir=* )
+            PGDATA=${OPTARG#*=}
+            ;;
+        sourcedir=* )
+            SOURCE=${OPTARG#*=}
+            ;;
+    esac
+done
+
+[[ -z $PGDATA || -z $SOURCE ]] && exit 1
+
+mkdir -p $(dirname $PGDATA)
+
+exec cp -af $SOURCE $PGDATA

features/basic_replication.feature

@@ -14,16 +14,28 @@ Feature: basic replication
     Then table foo is present on postgres2 after 20 seconds
 
   Scenario: check restart of sync replica
-    Given I run patronictl.py restart batman postgres2 --force
+    Given I shut down postgres2
-    And "sync" key in DCS has sync_standby=postgres1 after 2 seconds
+    Then "sync" key in DCS has sync_standby=postgres1 after 5 seconds
-    And I run patronictl.py restart batman postgres1 --force
+    When I start postgres2
-    Then I receive a response returncode 0
+    And I shut down postgres1
-    And "sync" key in DCS has sync_standby=postgres2 after 10 seconds
+    Then "sync" key in DCS has sync_standby=postgres2 after 10 seconds
+    When I start postgres1
+    And "members/postgres1" key in DCS has state=running after 10 seconds
+    And I sleep for 2 seconds
+    When I issue a GET request to http://127.0.0.1:8010/sync
+    Then I receive a response code 200
+    When I issue a GET request to http://127.0.0.1:8009/async
+    Then I receive a response code 200
 
   Scenario: check the basic failover in synchronous mode
-    When I kill postgres0
+    Given I run patronictl.py pause batman
-    Then postgres2 role is the primary after 22 seconds
+    Then I receive a response returncode 0
-    When I issue a PATCH request to http://127.0.0.1:8009/config with {"synchronous_mode": null, "master_start_timeout": 0}
+    When I sleep for 2 seconds
+    And I shut down postgres0
+    And I run patronictl.py resume batman
+    Then I receive a response returncode 0
+    And postgres2 role is the primary after 24 seconds
+    When I issue a PATCH request to http://127.0.0.1:8010/config with {"synchronous_mode": null, "master_start_timeout": 0}
     Then I receive a response code 200
     When I add the table bar to postgres2
     Then table bar is present on postgres1 after 20 seconds
@@ -34,7 +46,8 @@ Feature: basic replication
     And postgres1 role is the primary after 10 seconds
 
   Scenario: check rejoin of the former master with pg_rewind
-    Given I start postgres0
+    Given I add the table splitbrain to postgres0
+    And I start postgres0
     Then postgres0 role is the secondary after 20 seconds
     When I add the table buz to postgres1
     Then table buz is present on postgres0 after 20 seconds

features/callback.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+
+[[ "$3" == "master" ]] || exit
+
+PGPASSWORD=zalando psql -h localhost -U postgres -p $1 -w -tAc "SELECT slot_name FROM pg_replication_slots WHERE slot_type = 'logical'" >> data/postgres0/label

features/custom_bootstrap.feature

@@ -0,0 +1,17 @@
+Feature: custom bootstrap
+    We should check that patroni can bootstrap a new cluster from a backup
+
+Scenario: clone existing cluster using pg_basebackup
+    Given I start postgres0
+    Then postgres0 is a leader after 10 seconds
+    When I add the table foo to postgres0
+    And I start postgres1 in a cluster batman1 as a clone of postgres0
+    Then postgres1 is a leader of batman1 after 10 seconds
+    Then table foo is present on postgres1 after 10 seconds
+
+Scenario: make a backup and do a restore into a new cluster
+    Given I add the table bar to postgres1
+    And I do a backup of postgres1
+    When I start postgres2 in a cluster batman2 from backup
+    Then postgres2 is a leader of batman2 after 10 seconds
+    And table bar is present on postgres2 after 10 seconds

features/environment.py

@@ -1,14 +1,20 @@
 import abc
 import consul
+import datetime
 import etcd
 import kazoo.client
 import kazoo.exceptions
 import os
+import psutil
 import psycopg2
+import json
 import shutil
+import signal
 import six
 import subprocess
+import sys
 import tempfile
+import threading
 import time
 import yaml
 
@@ -74,18 +80,29 @@ class AbstractController(object):
         if self._log:
             self._log.close()
 
+    def cancel_background(self):
+        pass
+
 
 class PatroniController(AbstractController):
-    __PORT = 5440
+    __PORT = 5360
     PATRONI_CONFIG = '{}.yml'
     """ starts and stops individual patronis"""
 
-    def __init__(self, context, name, work_directory, output_dir, tags=None):
+    def __init__(self, context, name, work_directory, output_dir, custom_config=None):
         super(PatroniController, self).__init__(context, 'patroni_' + name, work_directory, output_dir)
         PatroniController.__PORT += 1
         self._data_dir = os.path.join(work_directory, 'data', name)
         self._connstring = None
-        self._config = self._make_patroni_test_config(name, tags)
+        if custom_config and 'watchdog' in custom_config:
+            self.watchdog = WatchdogMonitor(name, work_directory, output_dir)
+            custom_config['watchdog'] = {'driver': 'testing', 'device': self.watchdog.fifo_path, 'mode': 'required'}
+        else:
+            self.watchdog = None
+
+        self._scope = (custom_config or {}).get('scope', 'batman')
+        self._config = self._make_patroni_test_config(name, custom_config)
+        self._closables = []
 
         self._conn = None
         self._curs = None
@@ -94,62 +111,97 @@ class PatroniController(AbstractController):
         with open(os.path.join(self._data_dir, 'label'), 'w') as f:
             f.write(content)
 
-    def read_label(self):
+    def read_label(self, label):
         try:
-            with open(os.path.join(self._data_dir, 'label'), 'r') as f:
+            with open(os.path.join(self._data_dir, label), 'r') as f:
                 return f.read().strip()
         except IOError:
             return None
 
-    def add_tag_to_config(self, tag, value):
+    @staticmethod
+    def recursive_update(dst, src):
+        for k, v in src.items():
+            if k in dst and isinstance(dst[k], dict):
+                PatroniController.recursive_update(dst[k], v)
+            else:
+                dst[k] = v
+
+    def update_config(self, custom_config):
         with open(self._config) as r:
             config = yaml.safe_load(r)
-            config['tags']['tag'] = value
+            self.recursive_update(config, custom_config)
             with open(self._config, 'w') as w:
                 yaml.safe_dump(config, w, default_flow_style=False)
+        self._scope = config.get('scope', 'batman')
+
+    def add_tag_to_config(self, tag, value):
+        self.update_config({'tags': {tag: value}})
 
     def _start(self):
-        return subprocess.Popen(['coverage', 'run', '--source=patroni', '-p', 'patroni.py', self._config],
+        if self.watchdog:
+            self.watchdog.start()
+        if isinstance(self._context.dcs_ctl, KubernetesController):
+            self._context.dcs_ctl.create_pod(self._name[8:], self._scope)
+            os.environ['PATRONI_KUBERNETES_POD_IP'] = '10.0.0.' + self._name[-1]
+        return subprocess.Popen([sys.executable, '-m', 'coverage', 'run',
+                                '--source=patroni', '-p', 'patroni.py', self._config],
                                 stdout=self._log, stderr=subprocess.STDOUT, cwd=self._work_directory)
 
     def stop(self, kill=False, timeout=15, postgres=False):
         if postgres:
             return subprocess.call(['pg_ctl', '-D', self._data_dir, 'stop', '-mi', '-w'])
         super(PatroniController, self).stop(kill, timeout)
+        if isinstance(self._context.dcs_ctl, KubernetesController):
+            self._context.dcs_ctl.delete_pod(self._name[8:])
+        if self.watchdog:
+            self.watchdog.stop()
 
     def _is_accessible(self):
-        return self.query("SELECT 1", fail_ok=True) is not None
+        cursor = self.query("SELECT 1", fail_ok=True)
+        if cursor is not None:
+            cursor.execute("SET synchronous_commit TO 'local'")
+            return True
 
-    def _make_patroni_test_config(self, name, tags):
+    def _make_patroni_test_config(self, name, custom_config):
         patroni_config_name = self.PATRONI_CONFIG.format(name)
         patroni_config_path = os.path.join(self._output_dir, patroni_config_name)
 
         with open(patroni_config_name) as f:
             config = yaml.safe_load(f)
-            config.pop('etcd')
+            config.pop('etcd', None)
 
         host = config['postgresql']['listen'].split(':')[0]
 
         config['postgresql']['listen'] = config['postgresql']['connect_address'] = '{0}:{1}'.format(host, self.__PORT)
 
+        config['name'] = name
+        config['postgresql']['data_dir'] = self._data_dir
+        config['postgresql']['use_unix_socket'] = True
+        config['postgresql']['parameters'].update({
+            'logging_collector': 'on', 'log_destination': 'csvlog', 'log_directory': self._output_dir,
+            'log_filename': name + '.log', 'log_statement': 'all', 'log_min_messages': 'debug1',
+            'unix_socket_directories': self._data_dir})
+
+        if 'bootstrap' in config:
+            config['bootstrap']['post_bootstrap'] = 'psql -w -c "SELECT 1"'
+            if 'initdb' in config['bootstrap']:
+                config['bootstrap']['initdb'].extend([{'auth': 'md5'}, {'auth-host': 'md5'}])
+
+        if custom_config is not None:
+            self.recursive_update(config, custom_config)
+
+        if config['postgresql'].get('callbacks', {}).get('on_role_change'):
+            config['postgresql']['callbacks']['on_role_change'] += ' ' + str(self.__PORT)
+
+        with open(patroni_config_path, 'w') as f:
+            yaml.safe_dump(config, f, default_flow_style=False)
+
         user = config['postgresql'].get('authentication', config['postgresql']).get('superuser', {})
         self._connkwargs = {k: user[n] for n, k in [('username', 'user'), ('password', 'password')] if n in user}
         self._connkwargs.update({'host': host, 'port': self.__PORT, 'database': 'postgres'})
 
-        config['name'] = name
+        self._replication = config['postgresql'].get('authentication', config['postgresql']).get('replication', {})
-        config['postgresql']['data_dir'] = self._data_dir
+        self._replication.update({'host': host, 'port': self.__PORT, 'database': 'postgres'})
-        config['postgresql']['parameters'].update({
-            'logging_collector': 'on', 'log_destination': 'csvlog', 'log_directory': self._output_dir,
-            'log_filename': name + '.log', 'log_statement': 'all', 'log_min_messages': 'debug1'})
-
-        if 'bootstrap' in config and 'initdb' in config['bootstrap']:
-            config['bootstrap']['initdb'].extend([{'auth': 'md5'}, {'auth-host': 'md5'}])
-
-        if tags:
-            config['tags'] = tags
-
-        with open(patroni_config_path, 'w') as f:
-            yaml.safe_dump(config, f, default_flow_style=False)
 
         return patroni_config_path
 
@@ -185,10 +237,101 @@ class PatroniController(AbstractController):
             time.sleep(1)
         return False
 
+    def get_watchdog(self):
+        return self.watchdog
+
+    def _get_pid(self):
+        try:
+            pidfile = os.path.join(self._data_dir, 'postmaster.pid')
+            if not os.path.exists(pidfile):
+                return None
+            return int(open(pidfile).readline().strip())
+        except Exception:
+            return None
+
+    def database_is_running(self):
+        pid = self._get_pid()
+        if not pid:
+            return False
+        try:
+            os.kill(pid, 0)
+        except OSError:
+            return False
+        return True
+
+    def patroni_hang(self, timeout):
+        hang = ProcessHang(self._handle.pid, timeout)
+        self._closables.append(hang)
+        hang.start()
+
+    def checkpoint_hang(self, timeout):
+        pid = self._get_pid()
+        if not pid:
+            return False
+        proc = psutil.Process(pid)
+        for child in proc.children():
+            if 'checkpoint' in child.cmdline()[0]:
+                checkpointer = child
+                break
+        else:
+            return False
+        hang = ProcessHang(checkpointer.pid, timeout)
+        self._closables.append(hang)
+        hang.start()
+        return True
+
+    def cancel_background(self):
+        for obj in self._closables:
+            obj.close()
+        self._closables = []
+
+    def terminate_backends(self):
+        pid = self._get_pid()
+        if not pid:
+            return False
+        proc = psutil.Process(pid)
+        for p in proc.children():
+            if 'process' not in p.cmdline()[0]:
+                p.terminate()
+
+    @property
+    def backup_source(self):
+        return 'postgres://{username}:{password}@{host}:{port}/{database}'.format(**self._replication)
+
+    def backup(self, dest='data/basebackup'):
+        subprocess.call([PatroniPoolController.BACKUP_SCRIPT, '--walmethod=none',
+                         '--datadir=' + os.path.join(self._work_directory, dest),
+                         '--dbname=' + self.backup_source])
+
+
+class ProcessHang(object):
+
+    """A background thread implementing a cancelable process hang via SIGSTOP."""
+
+    def __init__(self, pid, timeout):
+        self._cancelled = threading.Event()
+        self._thread = threading.Thread(target=self.run)
+        self.pid = pid
+        self.timeout = timeout
+
+    def start(self):
+        self._thread.start()
+
+    def run(self):
+        os.kill(self.pid, signal.SIGSTOP)
+        try:
+            self._cancelled.wait(self.timeout)
+        finally:
+            os.kill(self.pid, signal.SIGCONT)
+
+    def close(self):
+        self._cancelled.set()
+        self._thread.join()
+
 
 class AbstractDcsController(AbstractController):
 
-    _CLUSTER_NODE = '/service/batman'
+    _CLUSTER_NODE = '/service/{0}'
 
     def __init__(self, context, mktemp=True):
         work_directory = mktemp and tempfile.mkdtemp() or None
@@ -203,17 +346,13 @@ class AbstractDcsController(AbstractController):
         if self._work_directory:
             shutil.rmtree(self._work_directory)
 
-    def path(self, key=None):
+    def path(self, key=None, scope='batman'):
-        return self._CLUSTER_NODE + (key and '/' + key or '')
+        return self._CLUSTER_NODE.format(scope) + (key and '/' + key or '')
 
     @abc.abstractmethod
-    def query(self, key):
+    def query(self, key, scope='batman'):
         """ query for a value of a given key """
 
-    @abc.abstractmethod
-    def set(self, key, value):
-        """ set a value to a given key """
-
     @abc.abstractmethod
     def cleanup_service_tree(self):
         """ clean all contents stored in the tree used for the tests """
@@ -235,19 +374,21 @@ class ConsulController(AbstractDcsController):
     def __init__(self, context):
         super(ConsulController, self).__init__(context)
         os.environ['PATRONI_CONSUL_HOST'] = 'localhost:8500'
+        os.environ['PATRONI_CONSUL_REGISTER_SERVICE'] = 'on'
         self._client = consul.Consul()
+        self._config_file = None
 
     def _start(self):
-        config_file = self._work_directory + '.json'
+        self._config_file = self._work_directory + '.json'
-        with open(config_file, 'wb') as f:
+        with open(self._config_file, 'wb') as f:
             f.write(b'{"session_ttl_min":"5s","server":true,"bootstrap":true,"advertise_addr":"127.0.0.1"}')
-        return subprocess.Popen(['consul', 'agent', '-config-file', config_file, '-data-dir', self._work_directory],
+        return subprocess.Popen(['consul', 'agent', '-config-file', self._config_file, '-data-dir',
-                                stdout=self._log, stderr=subprocess.STDOUT)
+                                 self._work_directory], stdout=self._log, stderr=subprocess.STDOUT)
 
     def stop(self, kill=False, timeout=15):
         super(ConsulController, self).stop(kill=kill, timeout=timeout)
-        if self._work_directory:
+        if self._config_file:
-            os.unlink(self._work_directory + '.json')
+            os.unlink(self._config_file)
 
     def _is_running(self):
         try:
@@ -255,18 +396,15 @@ class ConsulController(AbstractDcsController):
         except Exception:
             return False
 
-    def path(self, key=None):
+    def path(self, key=None, scope='batman'):
-        return super(ConsulController, self).path(key)[1:]
+        return super(ConsulController, self).path(key, scope)[1:]
 
-    def query(self, key):
+    def query(self, key, scope='batman'):
-        _, value = self._client.kv.get(self.path(key))
+        _, value = self._client.kv.get(self.path(key, scope))
         return value and value['Value'].decode('utf-8')
 
-    def set(self, key, value):
-        self._client.kv.put(self.path(key), value)
-
     def cleanup_service_tree(self):
-        self._client.kv.delete(self.path(), recurse=True)
+        self._client.kv.delete(self.path(scope=''), recurse=True)
 
     def start(self, max_wait_limit=15):
         super(ConsulController, self).start(max_wait_limit)
@@ -285,18 +423,15 @@ class EtcdController(AbstractDcsController):
         return subprocess.Popen(["etcd", "--debug", "--data-dir", self._work_directory],
                                 stdout=self._log, stderr=subprocess.STDOUT)
 
-    def query(self, key):
+    def query(self, key, scope='batman'):
         try:
-            return self._client.get(self.path(key)).value
+            return self._client.get(self.path(key, scope)).value
         except etcd.EtcdKeyNotFound:
             return None
 
-    def set(self, key, value):
-        self._client.set(self.path(key), value)
-
     def cleanup_service_tree(self):
         try:
-            self._client.delete(self.path(), recursive=True)
+            self._client.delete(self.path(scope=''), recursive=True)
         except (etcd.EtcdKeyNotFound, etcd.EtcdConnectionFailed):
             return
         except Exception as e:
@@ -310,6 +445,76 @@ class EtcdController(AbstractDcsController):
             return False
 
 
+class KubernetesController(AbstractDcsController):
+
+    def __init__(self, context):
+        super(KubernetesController, self).__init__(context)
+        self._namespace = 'default'
+        self._labels = {"application": "patroni"}
+        self._label_selector = ','.join('{0}={1}'.format(k, v) for k, v in self._labels.items())
+        os.environ['PATRONI_KUBERNETES_LABELS'] = json.dumps(self._labels)
+        os.environ['PATRONI_KUBERNETES_USE_ENDPOINTS'] = 'true'
+
+        from kubernetes import client as k8s_client, config as k8s_config
+        k8s_config.load_kube_config(context='local')
+        self._client = k8s_client
+        self._api = self._client.CoreV1Api()
+
+    def _start(self):
+        pass
+
+    def create_pod(self, name, scope):
+        labels = self._labels.copy()
+        labels['cluster-name'] = scope
+        metadata = self._client.V1ObjectMeta(namespace=self._namespace, name=name, labels=labels)
+        spec = self._client.V1PodSpec(containers=[self._client.V1Container(name=name, image='empty')])
+        body = self._client.V1Pod(metadata=metadata, spec=spec)
+        self._api.create_namespaced_pod(self._namespace, body)
+
+    def delete_pod(self, name):
+        try:
+            self._api.delete_namespaced_pod(name, self._namespace, self._client.V1DeleteOptions())
+        except:
+            pass
+        while True:
+            try:
+                self._api.read_namespaced_pod(name, self._namespace)
+            except:
+                break
+
+    def query(self, key, scope='batman'):
+        if key.startswith('members/'):
+            pod = self._api.read_namespaced_pod(key[8:], self._namespace)
+            return (pod.metadata.annotations or {}).get('status', '')
+        else:
+            try:
+                e = self._api.read_namespaced_endpoints(scope + ('' if key == 'leader' else '-' + key), self._namespace)
+                if key == 'leader':
+                    return e.metadata.annotations[key]
+                else:
+                    return json.dumps(e.metadata.annotations)
+            except:
+                return None
+
+    def cleanup_service_tree(self):
+        try:
+            self._api.delete_collection_namespaced_pod(self._namespace, label_selector=self._label_selector)
+        except:
+            pass
+        try:
+            self._api.delete_collection_namespaced_endpoints(self._namespace, label_selector=self._label_selector)
+        except:
+            pass
+
+        while True:
+            result = self._api.list_namespaced_pod(self._namespace, label_selector=self._label_selector)
+            if len(result.items) < 1:
+                break
+
+    def _is_running(self):
+        return True
+
+
 class ZooKeeperController(AbstractDcsController):
 
     """ handles all zookeeper related tasks, used for the tests setup and cleanup """
@@ -323,18 +528,15 @@ class ZooKeeperController(AbstractDcsController):
     def _start(self):
         pass  # TODO: implement later
 
-    def query(self, key):
+    def query(self, key, scope='batman'):
         try:
-            return self._client.get(self.path(key))[0].decode('utf-8')
+            return self._client.get(self.path(key, scope))[0].decode('utf-8')
         except kazoo.exceptions.NoNodeError:
             return None
 
-    def set(self, key, value):
-        self._client.set(self.path(key), value.encode('utf-8'))
-
     def cleanup_service_tree(self):
         try:
-            self._client.delete(self.path(), recursive=True)
+            self._client.delete(self.path(scope=''), recursive=True)
         except (kazoo.exceptions.NoNodeError):
             return
         except Exception as e:
@@ -359,6 +561,8 @@ class ExhibitorController(ZooKeeperController):
 
 class PatroniPoolController(object):
 
+    BACKUP_SCRIPT = 'features/backup_create.sh'
+
     def __init__(self, context):
         self._context = context
         self._dcs = None
@@ -383,13 +587,16 @@ class PatroniPoolController(object):
     def output_dir(self):
         return self._output_dir
 
-    def start(self, name, max_wait_limit=20, tags=None):
+    def start(self, name, max_wait_limit=20, custom_config=None):
         if name not in self._processes:
-            self._processes[name] = PatroniController(self._context, name, self.patroni_path, self._output_dir, tags)
+            self._processes[name] = PatroniController(self._context, name, self.patroni_path,
+                                                      self._output_dir, custom_config)
         self._processes[name].start(max_wait_limit)
 
     def __getattr__(self, func):
-        if func not in ['stop', 'query', 'write_label', 'read_label', 'check_role_has_changed_to', 'add_tag_to_config']:
+        if func not in ['stop', 'query', 'write_label', 'read_label', 'check_role_has_changed_to', 'add_tag_to_config',
+                        'get_watchdog', 'database_is_running', 'checkpoint_hang', 'patroni_hang',
+                        'terminate_backends', 'backup']:
             raise AttributeError("PatroniPoolController instance has no attribute '{0}'".format(func))
 
         def wrapper(name, *args, **kwargs):
@@ -398,6 +605,7 @@ class PatroniPoolController(object):
 
     def stop_all(self):
         for ctl in self._processes.values():
+            ctl.cancel_background()
             ctl.stop()
         self._processes.clear()
 
@@ -408,6 +616,61 @@ class PatroniPoolController(object):
         os.makedirs(feature_dir)
         self._output_dir = feature_dir
 
+    def clone(self, from_name, cluster_name, to_name):
+        f = self._processes[from_name]
+        custom_config = {
+            'scope': cluster_name,
+            'bootstrap': {
+                'method': 'pg_basebackup',
+                'pg_basebackup': {
+                    'command': self.BACKUP_SCRIPT + ' --walmethod=stream --dbname=' + f.backup_source
+                },
+                'dcs': {
+                    'postgresql': {
+                        'parameters': {
+                            'max_connections': 101
+                        }
+                    }
+                }
+            },
+            'postgresql': {
+                'parameters': {
+                    'archive_mode': 'on',
+                    'archive_command': 'mkdir -p {0} && test ! -f {0}/%f && cp %p {0}/%f'.format(
+                            os.path.join(self.patroni_path, 'data/wal_archive'))
+                },
+                'authentication': {
+                    'superuser': {'password': 'zalando1'},
+                    'replication': {'password': 'rep-pass1'}
+                }
+            }
+        }
+        self.start(to_name, custom_config=custom_config)
+
+    def bootstrap_from_backup(self, name, cluster_name):
+        custom_config = {
+            'scope': cluster_name,
+            'bootstrap': {
+                'method': 'backup_restore',
+                'backup_restore': {
+                    'command': 'features/backup_restore.sh --sourcedir=' + os.path.join(self.patroni_path,
+                                                                                        'data/basebackup'),
+                    'recovery_conf': {
+                        'recovery_target_action': 'promote',
+                        'recovery_target_timeline': 'latest',
+                        'restore_command': 'cp {0}/data/wal_archive/%f %p'.format(self.patroni_path)
+                    }
+                }
+            },
+            'postgresql': {
+                'authentication': {
+                    'superuser': {'password': 'zalando2'},
+                    'replication': {'password': 'rep-pass2'}
+                }
+            }
+        }
+        self.start(name, custom_config=custom_config)
+
     @property
     def dcs(self):
         if self._dcs is None:
@@ -416,8 +679,125 @@ class PatroniPoolController(object):
         return self._dcs
 
 
+class WatchdogMonitor(object):
+    """Testing harness for emulating a watchdog device as a named pipe. Because we can't easily emulate ioctl's we
+    require a custom driver on Patroni side. The device takes no action, only notes if it was pinged and/or triggered.
+    """
+    def __init__(self, name, work_directory, output_dir):
+        self.fifo_path = os.path.join(work_directory, 'data', 'watchdog.{0}.fifo'.format(name))
+        self.fifo_file = None
+        self._stop_requested = False  # Relying on bool setting being atomic
+        self._thread = None
+        self.last_ping = None
+        self.was_pinged = False
+        self.was_closed = False
+        self._was_triggered = False
+        self.timeout = 60
+        self._log_file = open(os.path.join(output_dir, 'watchdog.{0}.log'.format(name)), 'w')
+        self._log("watchdog {0} initialized".format(name))
+
+    def _log(self, msg):
+        tstamp = datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S,%f")
+        self._log_file.write("{0}: {1}\n".format(tstamp, msg))
+
+    def start(self):
+        assert self._thread is None
+        self._stop_requested = False
+        self._log("starting fifo {0}".format(self.fifo_path))
+        fifo_dir = os.path.dirname(self.fifo_path)
+        if os.path.exists(self.fifo_path):
+            os.unlink(self.fifo_path)
+        elif not os.path.exists(fifo_dir):
+            os.mkdir(fifo_dir)
+        os.mkfifo(self.fifo_path)
+        self.last_ping = time.time()
+
+        self._thread = threading.Thread(target=self.run)
+        self._thread.start()
+
+    def run(self):
+        try:
+            while not self._stop_requested:
+                self._log("opening")
+                self.fifo_file = os.open(self.fifo_path, os.O_RDONLY)
+                try:
+                    self._log("Fifo {0} connected".format(self.fifo_path))
+                    self.was_closed = False
+                    while not self._stop_requested:
+                        c = os.read(self.fifo_file, 1)
+
+                        if c == b'X':
+                            self._log("Stop requested")
+                            return
+                        elif c == b'':
+                            self._log("Pipe closed")
+                            break
+                        elif c == b'C':
+                            command = b''
+                            c = os.read(self.fifo_file, 1)
+                            while c != b'\n' and c != b'':
+                                command += c
+                                c = os.read(self.fifo_file, 1)
+                            command = command.decode('utf8')
+
+                            if command.startswith('timeout='):
+                                self.timeout = int(command.split('=')[1])
+                                self._log("timeout={0}".format(self.timeout))
+                        elif c in [b'V', b'1']:
+                            cur_time = time.time()
+                            if cur_time - self.last_ping > self.timeout:
+                                self._log("Triggered")
+                                self._was_triggered = True
+                            if c == b'V':
+                                self._log("magic close")
+                                self.was_closed = True
+                            elif c == b'1':
+                                self.was_pinged = True
+                                self._log("ping after {0} seconds".format(cur_time - (self.last_ping or cur_time)))
+                                self.last_ping = cur_time
+                        else:
+                            self._log('Unknown command {0} received from fifo'.format(c))
+                finally:
+                    self.was_closed = True
+                    self._log("closing")
+                    os.close(self.fifo_file)
+        except Exception as e:
+            self._log("Error {0}".format(e))
+        finally:
+            self._log("stopping")
+            self._log_file.flush()
+            if os.path.exists(self.fifo_path):
+                os.unlink(self.fifo_path)
+
+    def stop(self):
+        self._log("Monitor stop")
+        self._stop_requested = True
+        try:
+            if os.path.exists(self.fifo_path):
+                fd = os.open(self.fifo_path, os.O_WRONLY)
+                os.write(fd, b'X')
+                os.close(fd)
+        except Exception as e:
+            self._log("err while closing: {0}".format(str(e)))
+        if self._thread:
+            self._thread.join()
+            self._thread = None
+
+    def reset(self):
+        self._log("reset")
+        self.was_pinged = self.was_closed = self._was_triggered = False
+
+    @property
+    def was_triggered(self):
+        delta = time.time() - self.last_ping
+        triggered = self._was_triggered or not self.was_closed and delta > self.timeout
+        self._log("triggered={0}, {1}s left".format(triggered, self.timeout - delta))
+        return triggered
+
+
 # actions to execute on start/stop of the tests and before running invidual features
 def before_all(context):
+    os.environ.update({'PATRONI_RESTAPI_USERNAME': 'username', 'PATRONI_RESTAPI_PASSWORD': 'password'})
     context.ci = 'TRAVIS_BUILD_NUMBER' in os.environ or 'BUILD_NUMBER' in os.environ
     context.timeout_multiplier = 2 if context.ci else 1
     context.pctl = PatroniPoolController(context)
@@ -432,8 +812,8 @@ def before_all(context):
 
 def after_all(context):
     context.dcs_ctl.stop()
-    subprocess.call(['coverage', 'combine'])
+    subprocess.call([sys.executable, '-m', 'coverage', 'combine'])
-    subprocess.call(['coverage', 'report'])
+    subprocess.call([sys.executable, '-m', 'coverage', 'report'])
 
 
 def before_feature(context, feature):
@@ -446,3 +826,5 @@ def after_feature(context, feature):
     context.pctl.stop_all()
     shutil.rmtree(os.path.join(context.pctl.patroni_path, 'data'))
     context.dcs_ctl.cleanup_service_tree()
+    if feature.status == 'failed':
+        shutil.copytree(context.pctl.output_dir, context.pctl.output_dir + '_failed')

features/patroni_api.feature

@@ -8,22 +8,24 @@ Scenario: check API requests on a stand-alone server
 	Then I receive a response code 200
 	And I receive a response state running
 	And I receive a response role master
+	When I issue a GET request to http://127.0.0.1:8008/health
+	Then I receive a response code 200
 	When I issue a GET request to http://127.0.0.1:8008/replica
 	Then I receive a response code 503
 	When I run patronictl.py reinit batman postgres0 --force
 	Then I receive a response returncode 0
 	And I receive a response output "Failed: reinitialize for member postgres0, status code=503, (I am the leader, can not reinitialize)"
-	When I run patronictl.py failover batman --master postgres0 --force
+	When I run patronictl.py switchover batman --master postgres0 --force
 	Then I receive a response returncode 1
-	And I receive a response output "Error: No candidates found to failover to"
+	And I receive a response output "Error: No candidates found to switchover to"
-	When I issue a POST request to http://127.0.0.1:8008/failover with {"leader": "postgres0"}
+	When I issue a POST request to http://127.0.0.1:8008/switchover with {"leader": "postgres0"}
-	Then I receive a response code 500
+	Then I receive a response code 412
-	And I receive a response text failover is not possible: cluster does not have members except leader
+	And I receive a response text switchover is not possible: cluster does not have members except leader
 	When I issue an empty POST request to http://127.0.0.1:8008/failover
 	Then I receive a response code 400
 	When I issue a POST request to http://127.0.0.1:8008/failover with {"foo": "bar"}
 	Then I receive a response code 400
-	And I receive a response text "No values given for required parameters leader and candidate"
+	And I receive a response text "Failover could be performed only to a specific candidate"
 
 Scenario: check local configuration reload
 	Given I issue an empty POST request to http://127.0.0.1:8008/reload
@@ -34,16 +36,16 @@ Scenario: check local configuration reload
 	Then I receive a response code 202
 
 Scenario: check dynamic configuration change via DCS
-	Given I issue a PATCH request to http://127.0.0.1:8008/config with {"ttl": 10, "loop_wait": 2, "postgresql": {"parameters": {"max_connections": 101}}}
+	Given I run patronictl.py edit-config -s 'ttl=10' -s 'loop_wait=2' -p 'max_connections=101' --force batman
-	Then I receive a response code 200
+	Then I receive a response returncode 0
-	And I receive a response loop_wait 2
+	And I receive a response output "+loop_wait: 2"
 	And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 11 seconds
 	When I issue a GET request to http://127.0.0.1:8008/config
 	Then I receive a response code 200
 	And I receive a response loop_wait 2
 	When I issue a GET request to http://127.0.0.1:8008/patroni
 	Then I receive a response code 200
-	And I receive a response tags {'tag': 'new_value'}
+	And I receive a response tags {'new_tag': 'new_value'}
 
 Scenario: check API requests for the primary-replica pair in the pause mode
 	Given I run patronictl.py pause batman
@@ -64,26 +66,42 @@ Scenario: check API requests for the primary-replica pair in the pause mode
 	When I sleep for 10 seconds
 	Then postgres1 role is the secondary after 15 seconds
 
-Scenario: check the failover via the API in the pause mode
+Scenario: check the switchover via the API in the pause mode
-	Given I run patronictl.py failover batman --master postgres0 --candidate postgres1 --force
+	Given I issue a POST request to http://127.0.0.1:8008/switchover with {"leader": "postgres0", "candidate": "postgres1"}
-	Then I receive a response returncode 0
+	Then I receive a response code 200
 	And postgres1 is a leader after 5 seconds
 	And postgres1 role is the primary after 10 seconds
 	And postgres0 role is the secondary after 10 seconds
 	And replication works from postgres1 to postgres0 after 20 seconds
+	When I issue a GET request to http://127.0.0.1:8008/master
+	Then I receive a response code 503
+	When I issue a GET request to http://127.0.0.1:8008/replica
+	Then I receive a response code 200
+	When I issue a GET request to http://127.0.0.1:8009/master
+	Then I receive a response code 200
+	When I issue a GET request to http://127.0.0.1:8009/replica
+	Then I receive a response code 503
 
-Scenario: check the scheduled failover
+Scenario: check the scheduled switchover
-	Given I issue a scheduled failover from postgres1 to postgres0 in 3 seconds
+	Given I issue a scheduled switchover from postgres1 to postgres0 in 3 seconds
 	Then I receive a response returncode 1
-	And I receive a response output "Can't schedule failover in the paused state"
+	And I receive a response output "Can't schedule switchover in the paused state"
 	When I run patronictl.py resume batman
 	Then I receive a response returncode 0
-	Given I issue a scheduled failover from postgres1 to postgres0 in 3 seconds
+	Given I issue a scheduled switchover from postgres1 to postgres0 in 3 seconds
 	Then I receive a response returncode 0
 	And postgres0 is a leader after 20 seconds
 	And postgres0 role is the primary after 10 seconds
 	And postgres1 role is the secondary after 10 seconds
 	And replication works from postgres0 to postgres1 after 25 seconds
+	When I issue a GET request to http://127.0.0.1:8008/master
+	Then I receive a response code 200
+	When I issue a GET request to http://127.0.0.1:8008/replica
+	Then I receive a response code 503
+	When I issue a GET request to http://127.0.0.1:8009/master
+	Then I receive a response code 503
+	When I issue a GET request to http://127.0.0.1:8009/replica
+	Then I receive a response code 200
 
 Scenario: check the scheduled restart
 	Given I issue a PATCH request to http://127.0.0.1:8008/config with {"postgresql": {"parameters": {"superuser_reserved_connections": "6"}}}

features/standby_cluster.feature

@@ -0,0 +1,48 @@
+Feature: standby cluster
+  Scenario: check permanent logical slots are preserved on failover/switchover
+    Given I start postgres1
+    Then postgres1 is a leader after 10 seconds
+    And I sleep for 3 seconds
+    When I issue a PATCH request to http://127.0.0.1:8009/config with {"loop_wait": 2, "slots": {"pm_1": {"type": "physical"}}, "postgresql": {"parameters": {"wal_level": "logical"}}}
+    Then I receive a response code 200
+    And Response on GET http://127.0.0.1:8009/config contains slots after 10 seconds
+    And I sleep for 3 seconds
+    When I issue a PATCH request to http://127.0.0.1:8009/config with {"slots": {"test_logical": {"type": "logical", "database": "postgres", "plugin": "test_decoding"}}}
+    Then I receive a response code 200
+    And I do a backup of postgres1
+    When I start postgres0 with callback configured
+    Then "members/postgres0" key in DCS has state=running after 10 seconds
+    And replication works from postgres1 to postgres0 after 15 seconds
+    When I shut down postgres1
+    Then postgres0 is a leader after 10 seconds
+    And "members/postgres0" key in DCS has role=master after 3 seconds
+    When I issue a GET request to http://127.0.0.1:8008/
+    Then I receive a response code 200
+    And there is a label with "test_logical" in postgres0 data directory
+
+  Scenario: check replication of a single table in a standby cluster
+    Given I start postgres1 in a standby cluster batman1 as a clone of postgres0
+    Then postgres1 is a leader of batman1 after 10 seconds
+    When I add the table foo to postgres0
+    Then table foo is present on postgres1 after 20 seconds
+    When I issue a GET request to http://127.0.0.1:8009/master
+    Then I receive a response code 200
+    When I issue a GET request to http://127.0.0.1:8009/standby_leader
+    Then I receive a response code 200
+    And I receive a response role standby_leader
+    And there is a postgres1_cb.log with "on_start replica batman1\non_role_change standby_leader batman1" in postgres1 data directory
+    When I start postgres2 in a cluster batman1
+    Then postgres2 role is the replica after 24 seconds
+    And table foo is present on postgres2 after 20 seconds
+
+  Scenario: check failover
+    When I kill postgres1
+    And I kill postmaster on postgres1
+    Then postgres2 is replicating from postgres0 after 32 seconds
+    When I issue a GET request to http://127.0.0.1:8010/master
+    Then I receive a response code 200
+    When I issue a GET request to http://127.0.0.1:8010/standby_leader
+    Then I receive a response code 200
+    And I receive a response role standby_leader
+    And replication works from postgres0 to postgres2 after 15 seconds
+    And there is a postgres2_cb.log with "on_start replica batman1\non_role_change standby_leader batman1" in postgres2 data directory

features/steps/basic_replication.py

@@ -11,7 +11,7 @@ def start_patroni(context, name):
 
 @step('I shut down {name:w}')
 def stop_patroni(context, name):
-    return context.pctl.stop(name)
+    return context.pctl.stop(name, timeout=60)
 
 
 @step('I kill {name:w}')

features/steps/cascading_replication.py

@@ -6,12 +6,13 @@ from behave import step, then
 
 @step('I configure and start {name:w} with a tag {tag_name:w} {tag_value:w}')
 def start_patroni_with_a_name_value_tag(context, name, tag_name, tag_value):
-    return context.pctl.start(name, tags={tag_name: tag_value})
+    return context.pctl.start(name, custom_config={'tags': {tag_name: tag_value}})
 
 
-@then('There is a label with "{content:w}" in {name:w} data directory')
+@then('There is a {label} with "{content}" in {name:w} data directory')
-def check_label(context, content, name):
+def check_label(context, label, content, name):
-    label = context.pctl.read_label(name)
+    label = context.pctl.read_label(name, label)
+    label = label.replace('\n', '\\n')
     assert label == content, "{0} is not equal to {1}".format(label, content)
 
 

features/steps/custom_bootstrap.py

@@ -0,0 +1,27 @@
+import time
+
+from behave import step, then
+
+
+@step('I start {name:w} in a cluster {cluster_name:w} as a clone of {name2:w}')
+def start_cluster_clone(context, name, cluster_name, name2):
+    context.pctl.clone(name2, cluster_name, name)
+
+
+@step('I start {name:w} in a cluster {cluster_name:w} from backup')
+def start_cluster_from_backup(context, name, cluster_name):
+    context.pctl.bootstrap_from_backup(name, cluster_name)
+
+
+@then('{name:w} is a leader of {cluster_name:w} after {time_limit:d} seconds')
+def is_a_leader(context, name, cluster_name, time_limit):
+    time_limit *= context.timeout_multiplier
+    max_time = time.time() + int(time_limit)
+    while (context.dcs_ctl.query("leader", scope=cluster_name) != name):
+        time.sleep(1)
+        assert time.time() < max_time, "{0} is not a leader in dcs after {1} seconds".format(name, time_limit)
+
+
+@step('I do a backup of {name:w}')
+def do_backup(context, name):
+    context.pctl.backup(name)

features/steps/patroni_api.py

@@ -1,9 +1,11 @@
+import base64
 import json
 import os
 import parse
 import requests
 import shlex
 import subprocess
+import sys
 import time
 import yaml
 
@@ -78,11 +80,13 @@ def do_post_empty(context, url):
 @step('I issue a {request_method:w} request to {url:url} with {data}')
 def do_request(context, request_method, url, data):
     data = data and json.loads(data) or {}
+    headers = {'Authorization': 'Basic ' + base64.b64encode('username:password'.encode('utf-8')).decode('utf-8'),
+               'Content-Type': 'application/json'}
     try:
         if request_method == 'PATCH':
-            r = requests.patch(url, json=data)
+            r = requests.patch(url, headers=headers, json=data)
         else:
-            r = requests.post(url, json=data)
+            r = requests.post(url, headers=headers, json=data)
     except requests.exceptions.RequestException:
         context.status_code = None
         context.response = None
@@ -92,7 +96,7 @@ def do_request(context, request_method, url, data):
 
 @step('I run {cmd}')
 def do_run(context, cmd):
-    cmd = ['coverage', 'run', '--source=patroni', '-p'] + shlex.split(cmd)
+    cmd = [sys.executable, '-m', 'coverage', 'run', '--source=patroni', '-p'] + shlex.split(cmd)
     try:
         # XXX: Dirty hack! We need to take name/passwd from the config!
         env = os.environ.copy()
@@ -111,7 +115,8 @@ def check_response(context, component, data):
         assert context.status_code == int(data),\
             "status code {0} != {1}, response: {2}".format(context.status_code, data, context.response)
     elif component == 'returncode':
-        assert context.status_code == int(data), "return code {0} != {1}".format(context.status_code, data)
+        assert context.status_code == int(data), "return code {0} != {1}, {2}".format(context.status_code,
+                                                                                      data, context.response)
     elif component == 'text':
         assert context.response == data.strip('"'), "response {0} does not contain {1}".format(context.response, data)
     elif component == 'output':
@@ -121,10 +126,10 @@ def check_response(context, component, data):
         assert str(context.response[component]) == str(data), "{0} does not contain {1}".format(component, data)
 
 
-@step('I issue a scheduled failover from {from_host:w} to {to_host:w} in {in_seconds:d} seconds')
+@step('I issue a scheduled switchover from {from_host:w} to {to_host:w} in {in_seconds:d} seconds')
-def scheduled_failover(context, from_host, to_host, in_seconds):
+def scheduled_switchover(context, from_host, to_host, in_seconds):
     context.execute_steps(u"""
-        Given I run patronictl.py failover batman --master {0} --candidate {1} --scheduled "{2}" --force
+        Given I run patronictl.py switchover batman --master {0} --candidate {1} --scheduled "{2}" --force
     """.format(from_host, to_host, datetime.now(tzutc) + timedelta(seconds=int(in_seconds))))
 
 

features/steps/standby_cluster.py

@@ -0,0 +1,82 @@
+import os
+import time
+
+from behave import step
+
+
+select_replication_query = """
+SELECT * FROM pg_catalog.pg_stat_replication
+WHERE application_name = '{0}'
+"""
+
+callback = "bash -c 'echo \"${*: -3:1} ${*: -2:1} ${*: -1:1}\" >> data/$1/$1_cb.log' -- "
+
+
+@step('I start {name:w} with callback configured')
+def start_patroni_with_callbacks(context, name):
+    return context.pctl.start(name, custom_config={
+        "postgresql": {
+            "callbacks": {
+                "on_role_change": "features/callback.sh"
+            }
+        }
+    })
+
+
+@step('I start {name:w} in a cluster {cluster_name:w}')
+def start_patroni(context, name, cluster_name):
+    return context.pctl.start(name, custom_config={
+        "scope": cluster_name,
+        "postgresql": {
+            "callbacks": {c: callback + name for c in ('on_start', 'on_stop', 'on_restart', 'on_role_change')},
+            "backup_restore": {
+                "command": "features/backup_restore.sh --sourcedir=" + os.path.join(context.pctl.patroni_path,
+                                                                                    "data/basebackup")}
+        }
+    })
+
+
+@step('I start {name:w} in a standby cluster {cluster_name:w} as a clone of {name2:w}')
+def start_patroni_standby_cluster(context, name, cluster_name, name2):
+    # we need to remove patroni.dynamic.json in order to "bootstrap" standby cluster with existing PGDATA
+    os.unlink(os.path.join(context.pctl._processes[name]._data_dir, 'patroni.dynamic.json'))
+    port = context.pctl._processes[name2]._connkwargs.get('port')
+    context.pctl._processes[name].update_config({
+        "scope": cluster_name,
+        "bootstrap": {
+            "dcs": {
+                "ttl": 20,
+                "loop_wait": 2,
+                "retry_timeout": 5,
+                "standby_cluster": {
+                    "host": "localhost",
+                    "port": port,
+                    "primary_slot_name": "pm_1",
+                    "create_replica_methods": ["backup_restore", "basebackup"]
+                }
+            }
+        },
+        "postgresql": {
+            "callbacks": {c: callback + name for c in ('on_start', 'on_stop', 'on_restart', 'on_role_change')}
+        }
+    })
+    return context.pctl.start(name)
+
+
+@step('{pg_name1:w} is replicating from {pg_name2:w} after {timeout:d} seconds')
+def check_replication_status(context, pg_name1, pg_name2, timeout):
+    bound_time = time.time() + timeout
+
+    while time.time() < bound_time:
+        cur = context.pctl.query(
+            pg_name2,
+            select_replication_query.format(pg_name1),
+            fail_ok=True
+        )
+
+        if cur and len(cur.fetchall()) != 0:
+            break
+
+        time.sleep(1)
+    else:
+        assert False, "{0} is not replicating from {1} after {2} seconds".format(pg_name1, pg_name2, timeout)

features/steps/watchdog.py

@@ -0,0 +1,74 @@
+from behave import step, then
+import time
+
+
+def polling_loop(timeout, interval=1):
+    """Returns an iterator that returns values until timeout has passed. Timeout is measured from start of iteration."""
+    start_time = time.time()
+    iteration = 0
+    end_time = start_time + timeout
+    while time.time() < end_time:
+        yield iteration
+        iteration += 1
+        time.sleep(interval)
+
+
+@step('I start {name:w} with watchdog')
+def start_patroni_with_watchdog(context, name):
+    return context.pctl.start(name, custom_config={'watchdog': True})
+
+
+@step('{name:w} watchdog has been pinged after {timeout:d} seconds')
+def watchdog_was_pinged(context, name, timeout):
+    for _ in polling_loop(timeout):
+        if context.pctl.get_watchdog(name).was_pinged:
+            return True
+    return False
+
+
+@then('{name:w} watchdog has been closed')
+def watchdog_was_closed(context, name):
+    assert context.pctl.get_watchdog(name).was_closed
+
+
+@step('I reset {name:w} watchdog state')
+def watchdog_reset_pinged(context, name):
+    context.pctl.get_watchdog(name).reset()
+
+
+@then('{name:w} watchdog is triggered after {timeout:d} seconds')
+def watchdog_was_triggered(context, name, timeout):
+    for _ in polling_loop(timeout):
+        if context.pctl.get_watchdog(name).was_triggered:
+            return True
+    assert False
+
+
+@then('{name:w} watchdog was not triggered')
+def watchdog_was_not_triggered(context, name):
+    assert not context.pctl.get_watchdog(name).was_triggered
+
+
+@step('{name:w} checkpoint takes {timeout:d} seconds')
+def checkpoint_hang(context, name, timeout):
+    assert context.pctl.checkpoint_hang(name, timeout)
+
+
+@step('{name:w} hangs for {timeout:d} seconds')
+def patroni_hang(context, name, timeout):
+    return context.pctl.patroni_hang(name, timeout)
+
+
+@step('I terminate {name:w} user processes')
+def terminate_backends(context, name):
+    return context.pctl.terminate_backends(name)
+
+
+@step('Sleep for {timeout:d} seconds')
+def dcs_connection_lost(context, timeout):
+    time.sleep(timeout)
+
+
+@then('{name:w} database is running')
+def database_is_running(context, name):
+    assert context.pctl.database_is_running(name)

features/watchdog.feature

@@ -0,0 +1,31 @@
+Feature: watchdog
+  Verify that watchdog gets pinged and triggered under appropriate circumstances.
+
+  Scenario: watchdog is opened and pinged
+    Given I start postgres0 with watchdog
+    Then postgres0 is a leader after 10 seconds
+    And postgres0 role is the primary after 10 seconds
+    And postgres0 watchdog has been pinged after 10 seconds
+
+  Scenario: watchdog is disabled during pause
+    Given I run patronictl.py pause batman
+    Then I receive a response returncode 0
+    When I sleep for 2 seconds
+    Then postgres0 watchdog has been closed
+
+  Scenario: watchdog is opened and pinged after resume
+    Given I reset postgres0 watchdog state
+    And I run patronictl.py resume batman
+    Then I receive a response returncode 0
+    And postgres0 watchdog has been pinged after 10 seconds
+
+  Scenario: watchdog is disabled when shutting down
+    Given I shut down postgres0
+    Then postgres0 watchdog has been closed
+
+  Scenario: watchdog is triggered if patroni stops responding
+    Given I reset postgres0 watchdog state
+    And I start postgres0 with watchdog
+    Then postgres0 role is the primary after 10 seconds
+    When postgres0 hangs for 30 seconds
+    Then postgres0 watchdog is triggered after 30 seconds

haproxy.cfg

@@ -1,21 +1,25 @@
 global
-	maxconn 100
+    maxconn 100
 
 defaults
-	log	global
+    log global
-	mode	tcp
+    mode tcp
-	retries 2
+    retries 2
-	timeout client 30m
+    timeout client 30m
-	timeout connect 4s
+    timeout connect 4s
-	timeout server 30m
+    timeout server 30m
-	timeout check 5s
+    timeout check 5s
 
-frontend ft_postgresql
+listen stats
-	bind *:5000
+    mode http
-	default_backend bk_db
+    bind *:7000
-
+    stats enable
-backend bk_db
+    stats uri /
-	option httpchk
 
+listen batman
+    bind *:5000
+    option httpchk
+    http-check expect status 200
+    default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
     server postgresql_127.0.0.1_5432 127.0.0.1:5432 maxconn 100 check port 8008
     server postgresql_127.0.0.1_5433 127.0.0.1:5433 maxconn 100 check port 8009

kubernetes/Dockerfile

@@ -0,0 +1,34 @@
+FROM postgres:11
+MAINTAINER Alexander Kukushkin <[email protected]>
+
+RUN export DEBIAN_FRONTEND=noninteractive \
+    && echo 'APT::Install-Recommends "0";\nAPT::Install-Suggests "0";' > /etc/apt/apt.conf.d/01norecommend \
+    && apt-get update -y \
+    && apt-get upgrade -y \
+    && apt-cache depends patroni | sed -n -e 's/.* Depends: \(python3-.\+\)$/\1/p' \
+            | grep -Ev '^python3-(sphinx|etcd|consul|kazoo|kubernetes)' \
+            | xargs apt-get install -y vim-tiny curl jq locales git python3-pip python3-wheel \
+    ## Make sure we have a en_US.UTF-8 locale available
+    && localedef -i en_US -c -f UTF-8 -A /usr/share/locale/locale.alias en_US.UTF-8 \
+    && pip3 install setuptools \
+    && pip3 install 'git+https://github.com/zalando/patroni.git#egg=patroni[kubernetes]' \
+    && PGHOME=/home/postgres \
+    && mkdir -p $PGHOME \
+    && chown postgres $PGHOME \
+    && sed -i "s|/var/lib/postgresql.*|$PGHOME:/bin/bash|" /etc/passwd \
+    # Set permissions for OpenShift
+    && chmod 775 $PGHOME \
+    && chmod 664 /etc/passwd \
+    # Clean up
+    && apt-get remove -y git python3-pip python3-wheel \
+    && apt-get autoremove -y \
+    && apt-get clean -y \
+    && rm -rf /var/lib/apt/lists/* /root/.cache
+
+ADD entrypoint.sh /
+
+EXPOSE 5432 8008
+ENV LC_ALL=en_US.UTF-8 LANG=en_US.UTF-8 EDITOR=/usr/bin/editor
+USER postgres
+WORKDIR /home/postgres
+CMD ["/bin/bash", "/entrypoint.sh"]

kubernetes/entrypoint.sh

@@ -0,0 +1,39 @@
+#!/bin/bash
+
+if [[ $UID -ge 10000 ]]; then
+    GID=$(id -g)
+    sed -e "s/^postgres:x:[^:]*:[^:]*:/postgres:x:$UID:$GID:/" /etc/passwd > /tmp/passwd
+    cat /tmp/passwd > /etc/passwd
+    rm /tmp/passwd
+fi
+
+cat > /home/postgres/patroni.yml <<__EOF__
+bootstrap:
+  dcs:
+    postgresql:
+      use_pg_rewind: true
+  initdb:
+  - auth-host: md5
+  - auth-local: trust
+  - encoding: UTF8
+  - locale: en_US.UTF-8
+  - data-checksums
+  pg_hba:
+  - host all all 0.0.0.0/0 md5
+  - host replication ${PATRONI_REPLICATION_USERNAME} ${PATRONI_KUBERNETES_POD_IP}/16 md5
+restapi:
+  connect_address: '${PATRONI_KUBERNETES_POD_IP}:8008'
+postgresql:
+  connect_address: '${PATRONI_KUBERNETES_POD_IP}:5432'
+  authentication:
+    superuser:
+      password: '${PATRONI_SUPERUSER_PASSWORD}'
+    replication:
+      password: '${PATRONI_REPLICATION_PASSWORD}'
+__EOF__
+
+unset PATRONI_SUPERUSER_PASSWORD PATRONI_REPLICATION_PASSWORD
+export KUBERNETES_NAMESPACE=$PATRONI_KUBERNETES_NAMESPACE
+export POD_NAME=$PATRONI_NAME
+
+exec /usr/bin/python3 /usr/local/bin/patroni /home/postgres/patroni.yml

kubernetes/openshift-example/README.md

@@ -0,0 +1,49 @@
+# Patroni OpenShift Configuration
+Patroni can be run in OpenShift. Based on the kubernetes configuration, the Dockerfile and Entrypoint has been modified to support the dynamic UID/GID configuration that is applied in OpenShift. This can be run under the standard `restricted` SCC. 
+
+# Examples
+
+## Create test project
+
+```
+oc new-project patroni-test
+```
+
+## Build the image
+
+Note: Update the references when merged upstream. 
+Note: If deploying as a template for multiple users, the following commands should be performed in a shared namespace like `openshift`. 
+
+```
+oc import-image postgres:10 --confirm -n openshift
+oc new-build https://github.com/zalando/patroni --context-dir=kubernetes -n openshift
+```
+
+## Deploy the Image 
+Two configuration templates exist in [templates](templates) directory: 
+- Patroni Ephemeral
+- Patroni Persistent
+
+The only difference is whether or not the statefulset requests persistent storage. 
+
+## Create the Template
+Install the template into the `openshift` namespace if this should be shared across projects: 
+
+```
+oc create -f templates/template_patroni_ephemeral.yml -n openshift
+```
+
+Then, from your own project: 
+
+```
+oc new-app patroni-pgsql-ephemeral
+```
+
+Once the pods are running, two configmaps should be available: 
+
+```
+$ oc get configmap
+NAME                DATA      AGE
+patroniocp-config   0         1m
+patroniocp-leader   0         1m
+```

kubernetes/openshift-example/templates/template_patroni_ephemeral.yml

@@ -0,0 +1,287 @@
+apiVersion: v1
+kind: Template
+metadata:
+  name: patroni-pgsql-ephemeral
+  annotations:
+    description: |-
+      Patroni Postgresql database cluster, without persistent storage.
+
+      WARNING: Any data stored will be lost upon pod destruction. Only use this template for testing.
+    iconClass: icon-postgresql
+    openshift.io/display-name: Patroni Postgresql (Ephemeral)
+    openshift.io/long-description: This template deploys a a patroni postgresql HA cluster without persistent storage.
+    tags: postgresql
+objects:
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_CLUSTER_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_MASTER_SERVICE_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    selector:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+      role: master
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: v1
+  kind: Secret
+  metadata:
+    name: ${PATRONI_CLUSTER_NAME}
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+  stringData:
+    superuser-password: ${PATRONI_SUPERUSER_PASSWORD}
+    replication-password: ${PATRONI_REPLICATION_PASSWORD}
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_REPLICA_SERVICE_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    selector:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+      role: replica
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: apps/v1
+  kind: StatefulSet
+  metadata:
+    creationTimestamp: null
+    generation: 3
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${APPLICATION_NAME}
+  spec:
+    podManagementPolicy: OrderedReady
+    replicas: 3
+    revisionHistoryLimit: 10
+    selector:
+      matchLabels:
+        application: ${APPLICATION_NAME}
+        cluster-name: ${PATRONI_CLUSTER_NAME}
+    serviceName: ${APPLICATION_NAME}
+    template:
+      metadata:
+        creationTimestamp: null
+        labels:
+          application: ${APPLICATION_NAME}
+          cluster-name: ${PATRONI_CLUSTER_NAME}
+      spec:
+        containers:
+        - env:
+          - name: PATRONI_KUBERNETES_POD_IP
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: status.podIP
+          - name: PATRONI_KUBERNETES_NAMESPACE
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: metadata.namespace
+          - name: PATRONI_KUBERNETES_LABELS
+            value: '{application: ${APPLICATION_NAME}, cluster-name: ${PATRONI_CLUSTER_NAME}}'
+          - name: PATRONI_SUPERUSER_USERNAME
+            value: ${PATRONI_SUPERUSER_USERNAME}
+          - name: PATRONI_SUPERUSER_PASSWORD
+            valueFrom:
+              secretKeyRef:
+                key: superuser-password
+                name: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_REPLICATION_USERNAME
+            value: ${PATRONI_REPLICATION_USERNAME}
+          - name: PATRONI_REPLICATION_PASSWORD
+            valueFrom:
+              secretKeyRef:
+                key: replication-password
+                name: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_SCOPE
+            value: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_NAME
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: metadata.name
+          - name: PATRONI_POSTGRESQL_DATA_DIR
+            value: /home/postgres/pgdata/pgroot/data
+          - name: PATRONI_POSTGRESQL_PGPASS
+            value: /tmp/pgpass
+          - name: PATRONI_POSTGRESQL_LISTEN
+            value: 0.0.0.0:5432
+          - name: PATRONI_RESTAPI_LISTEN
+            value: 0.0.0.0:8008
+          image: docker-registry.default.svc:5000/${NAMESPACE}/patroni:latest
+          imagePullPolicy: IfNotPresent
+          name: ${APPLICATION_NAME}
+          ports:
+          - containerPort: 8008
+            protocol: TCP
+          - containerPort: 5432
+            protocol: TCP
+          resources: {}
+          terminationMessagePath: /dev/termination-log
+          terminationMessagePolicy: File
+          volumeMounts:
+          - mountPath: /home/postgres/pgdata
+            name: pgdata
+        dnsPolicy: ClusterFirst
+        restartPolicy: Always
+        schedulerName: default-scheduler
+        securityContext: {}
+        serviceAccount: ${SERVICE_ACCOUNT}
+        serviceAccountName: ${SERVICE_ACCOUNT}
+        terminationGracePeriodSeconds: 0
+        volumes:
+        - name: pgdata
+          emptyDir: {}
+    updateStrategy:
+      type: OnDelete
+- apiVersion: v1
+  kind: Endpoints
+  metadata:
+    name: ${APPLICATION_NAME}
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+  subsets: []
+- apiVersion: v1
+  kind: ServiceAccount
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: Role
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+  rules:
+  - apiGroups:
+    - ""
+    resources:
+    - configmaps
+    verbs:
+    - create
+    - get
+    - list
+    - patch
+    - update
+    - watch
+    # delete is required only for 'patronictl remove'
+    - delete
+  - apiGroups:
+    - ""
+    resources:
+    - endpoints
+    verbs:
+    - get
+    - patch
+    - update
+    # the following three privileges are necessary only when using endpoints
+    - create
+    - list
+    - watch
+    # delete is required only for for 'patronictl remove'
+    - delete
+  - apiGroups:
+    - ""
+    resources:
+    - pods
+    verbs:
+    - get
+    - list
+    - patch
+    - update
+    - watch
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: RoleBinding
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+  roleRef:
+    apiGroup: rbac.authorization.k8s.io
+    kind: Role
+    name: ${SERVICE_ACCOUNT}
+  subjects:
+  - kind: ServiceAccount
+    name: ${SERVICE_ACCOUNT}
+parameters:
+- description: The name of the application for labelling all artifacts.
+  displayName: Application Name
+  name: APPLICATION_NAME
+  value: patroni-ephemeral
+- description: The name of the patroni-pgsql cluster.
+  displayName: Cluster Name
+  name: PATRONI_CLUSTER_NAME
+  value: patroni-ephemeral
+- description: The name of the OpenShift Service exposed for the patroni-ephemeral-master container.
+  displayName: Master service name.  
+  name: PATRONI_MASTER_SERVICE_NAME
+  value: patroni-ephemeral-master
+- description: The name of the OpenShift Service exposed for the patroni-ephemeral-replica containers.
+  displayName: Replica service name. 
+  name: PATRONI_REPLICA_SERVICE_NAME
+  value: patroni-ephemeral-replica
+- description: Maximum amount of memory the container can use.
+  displayName: Memory Limit
+  name: MEMORY_LIMIT
+  value: 512Mi
+- description: The OpenShift Namespace where the patroni and postgresql ImageStream resides.
+  displayName: ImageStream Namespace
+  name: NAMESPACE
+  value: openshift
+- description: Username of the superuser account for initialization.
+  displayName: Superuser Username
+  name: PATRONI_SUPERUSER_USERNAME
+  value: postgres
+- description: Password of the superuser account for initialization.
+  displayName: Superuser Passsword
+  name: PATRONI_SUPERUSER_PASSWORD
+  value: postgres
+- description: Username of the replication account for initialization.
+  displayName: Replication Username
+  name: PATRONI_REPLICATION_USERNAME
+  value: postgres
+- description: Password of the replication account for initialization.
+  displayName: Repication Passsword
+  name: PATRONI_REPLICATION_PASSWORD
+  value: postgres
+- description: Service account name used for pods and rolebindings to form a cluster in the project. 
+  displayName: Service Account
+  name: SERVICE_ACCOUNT
+  value: patroniocp

kubernetes/openshift-example/templates/template_patroni_persistent.yaml

@@ -0,0 +1,303 @@
+apiVersion: v1
+kind: Template
+metadata:
+  name: patroni-pgsql-persistent
+  annotations:
+    description: |-
+      Patroni Postgresql database cluster, with persistent storage.
+
+      WARNING: Any data stored will be lost upon pod destruction. Only use this template for testing.
+    iconClass: icon-postgresql
+    openshift.io/display-name: Patroni Postgresql (Persistent)
+    openshift.io/long-description: This template deploys a a patroni postgresql HA cluster without persistent storage.
+    tags: postgresql
+objects:
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_CLUSTER_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_MASTER_SERVICE_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    selector:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+      role: master
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: v1
+  kind: Secret
+  metadata:
+    name: ${PATRONI_CLUSTER_NAME}
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+  stringData:
+    superuser-password: ${PATRONI_SUPERUSER_PASSWORD}
+    replication-password: ${PATRONI_REPLICATION_PASSWORD}
+- apiVersion: v1
+  kind: Service
+  metadata:
+    creationTimestamp: null
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${PATRONI_REPLICA_SERVICE_NAME}
+  spec:
+    ports:
+    - port: 5432
+      protocol: TCP
+      targetPort: 5432
+    selector:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+      role: replica
+    sessionAffinity: None
+    type: ClusterIP
+  status:
+    loadBalancer: {}
+- apiVersion: apps/v1
+  kind: StatefulSet
+  metadata:
+    creationTimestamp: null
+    generation: 3
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+    name: ${APPLICATION_NAME}
+  spec:
+    podManagementPolicy: OrderedReady
+    replicas: 3
+    revisionHistoryLimit: 10
+    selector:
+      matchLabels:
+        application: ${APPLICATION_NAME}
+        cluster-name: ${PATRONI_CLUSTER_NAME}
+    serviceName: ${APPLICATION_NAME}
+    template:
+      metadata:
+        creationTimestamp: null
+        labels:
+          application: ${APPLICATION_NAME}
+          cluster-name: ${PATRONI_CLUSTER_NAME}
+      spec:
+        containers:
+        - env:
+          - name: PATRONI_KUBERNETES_POD_IP
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: status.podIP
+          - name: PATRONI_KUBERNETES_NAMESPACE
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: metadata.namespace
+          - name: PATRONI_KUBERNETES_LABELS
+            value: '{application: ${APPLICATION_NAME}, cluster-name: ${PATRONI_CLUSTER_NAME}}'
+          - name: PATRONI_SUPERUSER_USERNAME
+            value: ${PATRONI_SUPERUSER_USERNAME}
+          - name: PATRONI_SUPERUSER_PASSWORD
+            valueFrom:
+              secretKeyRef:
+                key: superuser-password
+                name: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_REPLICATION_USERNAME
+            value: ${PATRONI_REPLICATION_USERNAME}
+          - name: PATRONI_REPLICATION_PASSWORD
+            valueFrom:
+              secretKeyRef:
+                key: replication-password
+                name: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_SCOPE
+            value: ${PATRONI_CLUSTER_NAME}
+          - name: PATRONI_NAME
+            valueFrom:
+              fieldRef:
+                apiVersion: v1
+                fieldPath: metadata.name
+          - name: PATRONI_POSTGRESQL_DATA_DIR
+            value: /home/postgres/pgdata/pgroot/data
+          - name: PATRONI_POSTGRESQL_PGPASS
+            value: /tmp/pgpass
+          - name: PATRONI_POSTGRESQL_LISTEN
+            value: 0.0.0.0:5432
+          - name: PATRONI_RESTAPI_LISTEN
+            value: 0.0.0.0:8008
+          image: docker-registry.default.svc:5000/${NAMESPACE}/patroni:latest
+          imagePullPolicy: IfNotPresent
+          name: ${APPLICATION_NAME}
+          ports:
+          - containerPort: 8008
+            protocol: TCP
+          - containerPort: 5432
+            protocol: TCP
+          resources: {}
+          terminationMessagePath: /dev/termination-log
+          terminationMessagePolicy: File
+          volumeMounts:
+          - mountPath: /home/postgres/pgdata
+            name: ${APPLICATION_NAME}
+        dnsPolicy: ClusterFirst
+        restartPolicy: Always
+        schedulerName: default-scheduler
+        securityContext: {}
+        serviceAccount: ${SERVICE_ACCOUNT}
+        serviceAccountName: ${SERVICE_ACCOUNT}
+        terminationGracePeriodSeconds: 0
+        volumes:
+        - name: ${APPLICATION_NAME}
+          persistentVolumeClaim: 
+            claimName: ${APPLICATION_NAME}
+    volumeClaimTemplates:
+    - metadata:
+        labels:
+          application: ${APPLICATION_NAME}
+        name: ${APPLICATION_NAME}
+      spec:
+        accessModes:
+        - ReadWriteOnce
+        resources:
+          requests:
+            storage: ${PVC_SIZE}
+    updateStrategy:
+      type: OnDelete
+- apiVersion: v1
+  kind: Endpoints
+  metadata:
+    name: ${APPLICATION_NAME}
+    labels:
+      application: ${APPLICATION_NAME}
+      cluster-name: ${PATRONI_CLUSTER_NAME}
+  subsets: []
+- apiVersion: v1
+  kind: ServiceAccount
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: Role
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+  rules:
+  - apiGroups:
+    - ""
+    resources:
+    - configmaps
+    verbs:
+    - create
+    - get
+    - list
+    - patch
+    - update
+    - watch
+    # delete is required only for 'patronictl remove'
+    - delete
+  - apiGroups:
+    - ""
+    resources:
+    - endpoints
+    verbs:
+    - get
+    - patch
+    - update
+    # the following three privileges are necessary only when using endpoints
+    - create
+    - list
+    - watch
+    # delete is required only for for 'patronictl remove'
+    - delete
+  - apiGroups:
+    - ""
+    resources:
+    - pods
+    verbs:
+    - get
+    - list
+    - patch
+    - update
+    - watch
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: RoleBinding
+  metadata:
+    name: ${SERVICE_ACCOUNT}
+  roleRef:
+    apiGroup: rbac.authorization.k8s.io
+    kind: Role
+    name: ${SERVICE_ACCOUNT}
+  subjects:
+  - kind: ServiceAccount
+    name: ${SERVICE_ACCOUNT}
+parameters:
+- description: The name of the application for labelling all artifacts.
+  displayName: Application Name
+  name: APPLICATION_NAME
+  value: patroni-persistent
+- description: The name of the patroni-pgsql cluster.
+  displayName: Cluster Name
+  name: PATRONI_CLUSTER_NAME
+  value: patroni-persistent
+- description: The name of the OpenShift Service exposed for the patroni-persistent-master container.
+  displayName: Master service name.  
+  name: PATRONI_MASTER_SERVICE_NAME
+  value: patroni-persistent-master
+- description: The name of the OpenShift Service exposed for the patroni-persistent-replica containers.
+  displayName: Replica service name. 
+  name: PATRONI_REPLICA_SERVICE_NAME
+  value: patroni-persistent-replica
+- description: Maximum amount of memory the container can use.
+  displayName: Memory Limit
+  name: MEMORY_LIMIT
+  value: 512Mi
+- description: The OpenShift Namespace where the patroni and postgresql ImageStream resides.
+  displayName: ImageStream Namespace
+  name: NAMESPACE
+  value: openshift
+- description: Username of the superuser account for initialization.
+  displayName: Superuser Username
+  name: PATRONI_SUPERUSER_USERNAME
+  value: postgres
+- description: Password of the superuser account for initialization.
+  displayName: Superuser Passsword
+  name: PATRONI_SUPERUSER_PASSWORD
+  value: postgres
+- description: Username of the replication account for initialization.
+  displayName: Replication Username
+  name: PATRONI_REPLICATION_USERNAME
+  value: postgres
+- description: Password of the replication account for initialization.
+  displayName: Repication Passsword
+  name: PATRONI_REPLICATION_PASSWORD
+  value: postgres
+- description: Service account name used for pods and rolebindings to form a cluster in the project. 
+  displayName: Service Account
+  name: SERVICE_ACCOUNT
+  value: patroni-persistent
+- description: The size of the persistent volume to create. 
+  displayName: Persistent Volume Size
+  name: PVC_SIZE
+  value: 5Gi

kubernetes/openshift-example/test/Jenkinsfile

@@ -0,0 +1,43 @@
+pipeline {
+  agent any
+  stages {
+    stage ('Deploy test pod'){
+      when {
+          expression {
+            openshift.withCluster() {
+              openshift.withProject() {
+                return !openshift.selector( "dc", "pgbench" ).exists()
+              }
+            }
+          }
+        }
+      steps {
+        script {
+            openshift.withCluster() {
+                openshift.withProject() {
+                  def pgbench = openshift.newApp( "https://github.com/stewartshea/docker-pgbench/", "--name=pgbench", "-e PGPASSWORD=postgres", "-e PGUSER=postgres", "-e PGHOST=patroni-persistent-master", "-e PGDATABASE=postgres", "-e TEST_CLIENT_COUNT=20", "-e TEST_DURATION=120" )
+                  def pgbenchdc = openshift.selector( "dc", "pgbench" )
+                  timeout(5) { 
+                    pgbenchdc.rollout().status()
+                    }
+                  }
+                }
+            }
+        }
+    }
+    stage ('Run benchmark Test'){
+      steps {
+          sh '''
+          oc exec $(oc get pods -l app=pgbench | grep Running | awk '{print $1}')  ./test.sh
+          '''
+      }
+    }
+    stage ('Clean up pgtest pod'){
+      steps {
+          sh '''
+          oc delete all -l app=pgbench
+          '''
+      }
+    }
+  }
+}

kubernetes/openshift-example/test/README.md

@@ -0,0 +1,2 @@
+# Jenkins Test
+This pipeline test will create a separate deployment config for a pgbench pod and execute a test against the patroni cluster. This is a sample and should be customized. 

kubernetes/patroni_k8s.yaml

@@ -0,0 +1,212 @@
+# headless service to avoid deletion of patronidemo-config endpoint
+apiVersion: v1
+kind: Service
+metadata:
+  name: patronidemo-config
+  labels:
+    application: patroni
+    cluster-name: patronidemo
+spec:
+  clusterIP: None
+
+---
+apiVersion: apps/v1beta1
+kind: StatefulSet
+metadata:
+  name: &cluster_name patronidemo
+  labels:
+    application: patroni
+    cluster-name: *cluster_name
+spec:
+  replicas: 3
+  serviceName: *cluster_name
+  template:
+    metadata:
+      labels:
+        application: patroni
+        cluster-name: *cluster_name
+    spec:
+      serviceAccountName: patronidemo
+      containers:
+      - name: *cluster_name
+        image: patroni  # docker build -t patroni .
+        imagePullPolicy: IfNotPresent
+        ports:
+        - containerPort: 8008
+          protocol: TCP
+        - containerPort: 5432
+          protocol: TCP
+        volumeMounts:
+        - mountPath: /home/postgres/pgdata
+          name: pgdata
+        env:
+        - name: PATRONI_KUBERNETES_POD_IP
+          valueFrom:
+            fieldRef:
+              fieldPath: status.podIP
+        - name: PATRONI_KUBERNETES_NAMESPACE
+          valueFrom:
+            fieldRef:
+              fieldPath: metadata.namespace
+        - name: PATRONI_KUBERNETES_USE_ENDPOINTS
+          value: 'true'
+        - name: PATRONI_KUBERNETES_LABELS
+          value: '{application: patroni, cluster-name: patronidemo}'
+        - name: PATRONI_SUPERUSER_USERNAME
+          value: postgres
+        - name: PATRONI_SUPERUSER_PASSWORD
+          valueFrom:
+            secretKeyRef:
+              name: *cluster_name
+              key: superuser-password
+        - name: PATRONI_REPLICATION_USERNAME
+          value: standby
+        - name: PATRONI_REPLICATION_PASSWORD
+          valueFrom:
+            secretKeyRef:
+              name: *cluster_name
+              key: replication-password
+        - name: PATRONI_SCOPE
+          value: *cluster_name
+        - name: PATRONI_NAME
+          valueFrom:
+            fieldRef:
+              fieldPath: metadata.name
+        - name: PATRONI_POSTGRESQL_DATA_DIR
+          value: /home/postgres/pgdata/pgroot/data
+        - name: PATRONI_POSTGRESQL_PGPASS
+          value: /tmp/pgpass
+        - name: PATRONI_POSTGRESQL_LISTEN
+          value: '0.0.0.0:5432'
+        - name: PATRONI_RESTAPI_LISTEN
+          value: '0.0.0.0:8008'
+      terminationGracePeriodSeconds: 0
+      volumes:
+      - name: pgdata
+        emptyDir: {}
+#  volumeClaimTemplates:
+#  - metadata:
+#      labels:
+#        application: spilo
+#        spilo-cluster: *cluster_name
+#      annotations:
+#        volume.alpha.kubernetes.io/storage-class: anything
+#      name: pgdata
+#    spec:
+#      accessModes:
+#      - ReadWriteOnce
+#      resources:
+#        requests:
+#          storage: 5Gi
+
+---
+apiVersion: v1
+kind: Endpoints
+metadata:
+  name: &cluster_name patronidemo
+  labels:
+    application: patroni
+    cluster-name: *cluster_name
+subsets: []
+
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: &cluster_name patronidemo
+  labels:
+    application: patroni
+    cluster-name: *cluster_name
+spec:
+  type: ClusterIP
+  ports:
+  - port: 5432
+    targetPort: 5432
+
+---
+apiVersion: v1
+kind: Secret
+metadata:
+  name: &cluster_name patronidemo
+  labels:
+    application: patroni
+    cluster-name: *cluster_name
+type: Opaque
+data:
+  superuser-password: emFsYW5kbw==
+  replication-password: cmVwLXBhc3M=
+
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: patronidemo
+
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: patronidemo
+rules:
+- apiGroups:
+  - ""
+  resources:
+  - configmaps
+  verbs:
+  - create
+  - get
+  - list
+  - patch
+  - update
+  - watch
+  # delete and deletecollection are required only for 'patronictl remove'
+  - delete
+  - deletecollection
+- apiGroups:
+  - ""
+  resources:
+  - endpoints
+  verbs:
+  - get
+  - patch
+  - update
+  # the following three privileges are necessary only when using endpoints
+  - create
+  - list
+  - watch
+  # delete and deletecollection are required only for for 'patronictl remove'
+  - delete
+  - deletecollection
+- apiGroups:
+  - ""
+  resources:
+  - pods
+  verbs:
+  - get
+  - list
+  - patch
+  - update
+  - watch
+# The following privilege is only necessary for creation of headless service
+# for patronidemo-config endpoint, in order to prevent cleaning it up by the
+# k8s master. You can avoid giving this privilege by explicitly creating the
+# service like it is done in this manifest (lines 2..10)
+- apiGroups:
+  - ""
+  resources:
+  - services
+  verbs:
+  - create
+
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: patronidemo
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: patronidemo
+subjects:
+- kind: ServiceAccount
+  name: patronidemo

patroni.py

@@ -1,6 +1,14 @@
 #!/usr/bin/env python
 from patroni import main
 
+import os
 
 if __name__ == '__main__':
+
+    if os.getenv("PATRONI_DEBUG_MODE"):
+        # XXX Visual Code specific https://github.com/microsoft/ptvsd/issues/1443
+        # create processes by spawning new Python interpreters instead of forking the current one 
+        import multiprocessing
+        multiprocessing.set_start_method('spawn', True)
+
     main()

patroni/__init__.py

@@ -14,19 +14,25 @@ class Patroni(object):
         from patroni.config import Config
         from patroni.dcs import get_dcs
         from patroni.ha import Ha
+        from patroni.log import PatroniLogger
         from patroni.postgresql import Postgresql
         from patroni.version import __version__
+        from patroni.watchdog import Watchdog
 
         self.setup_signal_handlers()
 
         self.version = __version__
+        self.logger = PatroniLogger()
         self.config = Config()
+        self.logger.reload_config(self.config.get('log', {}))
         self.dcs = get_dcs(self.config)
+        self.watchdog = Watchdog(self.config)
         self.load_dynamic_configuration()
 
         self.postgresql = Postgresql(self.config['postgresql'])
         self.api = RestApiServer(self, self.config['restapi'])
         self.ha = Ha(self)
+        self.is_in_debug_mode = bool(os.getenv("PATRONI_DEBUG_MODE"))
 
         self.tags = self.get_tags()
         self.next_run = time.time()
@@ -37,15 +43,17 @@ class Patroni(object):
         while True:
             try:
                 cluster = self.dcs.get_cluster()
-                if cluster and cluster.config:
+                if cluster and cluster.config and cluster.config.data:
                     if self.config.set_dynamic_configuration(cluster.config):
                         self.dcs.reload_config(self.config)
+                        self.watchdog.reload_config(self.config)
                 elif not self.config.dynamic_configuration and 'bootstrap' in self.config:
                     if self.config.set_dynamic_configuration(self.config['bootstrap']['dcs']):
                         self.dcs.reload_config(self.config)
                 break
             except DCSError:
                 logger.warning('Can not get cluster from dcs')
+                time.sleep(5)
 
     def get_tags(self):
         return {tag: value for tag, value in self.config.get('tags', {}).items()
@@ -62,7 +70,9 @@ class Patroni(object):
     def reload_config(self):
         try:
             self.tags = self.get_tags()
+            self.logger.reload_config(self.config.get('log', {}))
             self.dcs.reload_config(self.config)
+            self.watchdog.reload_config(self.config)
             self.api.reload_config(self.config['restapi'])
             self.postgresql.reload_config(self.config['postgresql'])
         except Exception:
@@ -76,9 +86,10 @@ class Patroni(object):
         self._received_sighup = True
 
     def sigterm_handler(self, *args):
-        if not self._received_sigterm:
+        with self._sigterm_lock:
-            self._received_sigterm = True
+            if not self._received_sigterm:
-            sys.exit()
+                self._received_sigterm = True
+                sys.exit()
 
     @property
     def noloadbalance(self):
@@ -92,16 +103,22 @@ class Patroni(object):
             self.next_run = current_time
             # Release the GIL so we don't starve anyone waiting on async_executor lock
             time.sleep(0.001)
-            # Warn user that Patroni is not keeping up
+            # Warn user that Patroni is not keeping up or runs in debug
-            logger.warning("Loop time exceeded, rescheduling immediately.")
+            msg = "Patroni runs in the debug mode: keys' TTL is infinite, loop wait disabled" if self.is_in_debug_mode else "Loop time exceeded, rescheduling immediately."
+            logger.warning(msg)
         elif self.ha.watch(nap_time):
             self.next_run = time.time()
 
+    @property
+    def received_sigterm(self):
+        with self._sigterm_lock:
+            return self._received_sigterm
+
     def run(self):
         self.api.start()
         self.next_run = time.time()
 
-        while not self._received_sigterm:
+        while not self.received_sigterm:
             if self._received_sighup:
                 self._received_sighup = False
                 if self.config.reload_local_configuration():
@@ -109,66 +126,78 @@ class Patroni(object):
 
             logger.info(self.ha.run_cycle())
 
-            cluster = self.dcs.cluster
+            if self.dcs.cluster and self.dcs.cluster.config and self.dcs.cluster.config.data \
-            if cluster and cluster.config and self.config.set_dynamic_configuration(cluster.config):
+                    and self.config.set_dynamic_configuration(self.dcs.cluster.config):
                 self.reload_config()
 
-            if not self.postgresql.data_directory_empty():
+            if self.postgresql.role != 'uninitialized':
                 self.config.save_cache()
 
             self.schedule_next_run()
 
     def setup_signal_handlers(self):
+        from threading import Lock
+
         self._received_sighup = False
+        self._sigterm_lock = Lock()
         self._received_sigterm = False
-        signal.signal(signal.SIGHUP, self.sighup_handler)
+        if os.name != 'nt':
+            signal.signal(signal.SIGHUP, self.sighup_handler)
         signal.signal(signal.SIGTERM, self.sigterm_handler)
 
+    def shutdown(self):
+        with self._sigterm_lock:
+            self._received_sigterm = True
+        try:
+            self.api.shutdown()
+        except Exception:
+            logger.exception('Exception during RestApi.shutdown')
+        self.ha.shutdown()
+        self.logger.shutdown()
+
 
 def patroni_main():
-    logging.basicConfig(format='%(asctime)s %(levelname)s: %(message)s', level=logging.INFO)
-    logging.getLogger('requests').setLevel(logging.WARNING)
-
     patroni = Patroni()
     try:
         patroni.run()
     except KeyboardInterrupt:
         pass
     finally:
-        patroni.api.shutdown()
+        patroni.shutdown()
-        if patroni.ha.is_paused():
-            logger.info('Leader key is not deleted and Postgresql is not stopped due paused state')
-        else:
-            patroni.ha.while_not_sync_standby(lambda: patroni.postgresql.stop(checkpoint=False))
-            patroni.dcs.delete_leader()
 
 
-def pg_ctl_start(args):
+def fatal(string, *args):
-    import subprocess
+    sys.stderr.write('FATAL: ' + string.format(*args) + '\n')
-    postmaster = subprocess.Popen(args)
+    sys.exit(1)
-    print(postmaster.pid)
 
 
-def call_self(args, **kwargs):
+def check_psycopg2():
-    """This function executes Patroni once again with provided arguments.
+    min_psycopg2 = (2, 5, 4)
+    min_psycopg2_str = '.'.join(map(str, min_psycopg2))
 
-    :args: list of arguments to call Patroni with.
+    def parse_version(version):
-    :returns: `Popen` object"""
+        for e in version.split('.'):
+            try:
+                yield int(e)
+            except ValueError:
+                break
 
-    exe = [sys.executable]
+    try:
-    if not getattr(sys, 'frozen', False):  # Binary distribution?
+        import psycopg2
-        exe.append(sys.argv[0])
+        version_str = psycopg2.__version__.split(' ')[0]
-
+        version = tuple(parse_version(version_str))
-    import subprocess
+        if version < min_psycopg2:
-    return subprocess.Popen(exe + args, **kwargs)
+            fatal('Patroni requires psycopg2>={0}, but only {1} is available', min_psycopg2_str, version_str)
+    except ImportError:
+        fatal('Patroni requires psycopg2>={0} or psycopg2-binary', min_psycopg2_str)
 
 
 def main():
+    check_psycopg2()
     if os.getpid() != 1:
-        if len(sys.argv) > 5 and sys.argv[1] == 'pg_ctl_start':
-            return pg_ctl_start(sys.argv[2:])
         return patroni_main()
 
+    # Patroni started with PID=1, it looks like we are in the container
     pid = 0
 
     # Looks like we are in a docker, so we will act like init
@@ -179,7 +208,7 @@ def main():
                 if ret == (0, 0):
                     break
                 elif ret[0] != pid:
-                    logging.info('Reaped pid=%s, exit status=%s', *ret)
+                    logger.info('Reaped pid=%s, exit status=%s', *ret)
         except OSError:
             pass
 
@@ -187,14 +216,18 @@ def main():
         if pid:
             os.kill(pid, signo)
 
-    signal.signal(signal.SIGCHLD, sigchld_handler)
+    if os.name != 'nt':
-    signal.signal(signal.SIGHUP, passtochild)
+        signal.signal(signal.SIGCHLD, sigchld_handler)
+        signal.signal(signal.SIGHUP, passtochild)
+        signal.signal(signal.SIGQUIT, passtochild)
+        signal.signal(signal.SIGUSR1, passtochild)
+        signal.signal(signal.SIGUSR2, passtochild)
     signal.signal(signal.SIGINT, passtochild)
-    signal.signal(signal.SIGUSR1, passtochild)
+    signal.signal(signal.SIGABRT, passtochild)
-    signal.signal(signal.SIGUSR2, passtochild)
-    signal.signal(signal.SIGQUIT, passtochild)
     signal.signal(signal.SIGTERM, passtochild)
 
-    patroni = call_self(sys.argv[1:])
+    import multiprocessing
+    patroni = multiprocessing.Process(target=patroni_main)
+    patroni.start()
     pid = patroni.pid
-    patroni.wait()
+    patroni.join()

patroni/api.py

@@ -1,14 +1,18 @@
 import base64
-import fcntl
 import json
 import logging
 import psycopg2
 import time
+import traceback
 import dateutil.parser
 import datetime
+import os
+import socket
 
-from patroni.exceptions import PostgresConnectionException
+from patroni.postgresql import PostgresConnectionException
-from patroni.utils import deep_compare, patch_config, Retry, RetryFailedError, is_valid_pg_version, parse_int, tzutc
+from patroni.postgresql.misc import postgres_version_to_int, PostgresException
+from patroni.utils import deep_compare, parse_bool, patch_config, Retry, \
+    RetryFailedError, parse_int, split_host_port, tzutc, uri
 from six.moves.BaseHTTPServer import BaseHTTPRequestHandler, HTTPServer
 from six.moves.socketserver import ThreadingMixIn
 from threading import Thread
@@ -24,9 +28,9 @@ def check_auth(func):
     def do_PUT_foo():
         pass
     """
-    def wrapper(handler):
+    def wrapper(handler, *args, **kwargs):
         if handler.check_auth_header():
-            return func(handler)
+            return func(handler, *args, **kwargs)
     return wrapper
 
 
@@ -68,34 +72,56 @@ class RestApiHandler(BaseHTTPRequestHandler):
             response['scheduled_restart'] = patroni.scheduled_restart.copy()
             del response['scheduled_restart']['postmaster_start_time']
             response['scheduled_restart']['schedule'] = (response['scheduled_restart']['schedule']).isoformat()
+        if not patroni.ha.watchdog.is_healthy:
+            response['watchdog_failed'] = True
+        if patroni.ha.is_paused():
+            response['pause'] = True
+        qsize = patroni.logger.queue_size
+        if qsize > patroni.logger.NORMAL_LOG_QUEUE_SIZE:
+            response['logger_queue_size'] = qsize
+            lost = patroni.logger.records_lost
+            if lost:
+                response['logger_records_lost'] = lost
         self._write_json_response(status_code, response)
 
     def do_GET(self, write_status_code_only=False):
         """Default method for processing all GET requests which can not be routed to other methods"""
 
+        time_start = time.time()
+        request_type = 'OPTIONS' if write_status_code_only else 'GET'
+
         path = '/master' if self.path == '/' else self.path
         response = self.get_postgresql_status()
 
         patroni = self.server.patroni
         cluster = patroni.dcs.cluster
-        if cluster:  # dcs available
+
-            if cluster.leader and cluster.leader.name == patroni.postgresql.name:  # is_leader
+        if not cluster and patroni.ha.is_paused():
-                status_code = 200 if 'master' in path else 503
+            primary_status_code = 200 if response['role'] == 'master' else 503
-            elif 'role' not in response:
-                status_code = 503
-            elif response['role'] == 'master':  # running as master but without leader lock!!!!
-                status_code = 503
-            elif response['role'] in path:  # response['role'] != 'master'
-                status_code = 503 if patroni.noloadbalance else 200
-            else:
-                status_code = 503
-        elif 'role' in response and response['role'] in path:
-            status_code = 503 if response['role'] != 'master' and patroni.noloadbalance else 200
-        elif patroni.ha.restart_scheduled() and patroni.postgresql.role == 'master' and 'master' in path:
-            # exceptional case for master node when the postgres is being restarted via API
-            status_code = 200
         else:
-            status_code = 503
+            primary_status_code = 200 if patroni.ha.is_leader() else 503
+
+        replica_status_code = 200 if not patroni.noloadbalance and \
+            response.get('role') == 'replica' and response.get('state') == 'running' else 503
+        status_code = 503
+
+        if patroni.ha.is_standby_cluster() and ('standby_leader' in path or 'standby-leader' in path):
+            status_code = 200 if patroni.ha.is_leader() else 503
+        elif 'master' in path or 'leader' in path or 'primary' in path or 'read-write' in path:
+            status_code = primary_status_code
+        elif 'replica' in path:
+            status_code = replica_status_code
+        elif 'read-only' in path:
+            status_code = 200 if primary_status_code == 200 else replica_status_code
+        elif 'health' in path:
+            status_code = 200 if response.get('state') == 'running' else 503
+        elif cluster:  # dcs is available
+            is_synchronous = cluster.is_synchronous_mode() and cluster.sync \
+                    and cluster.sync.sync_standby == patroni.postgresql.name
+            if path in ('/sync', '/synchronous') and is_synchronous:
+                status_code = replica_status_code
+            elif path in ('/async', '/asynchronous') and not is_synchronous:
+                status_code = replica_status_code
 
         if write_status_code_only:  # when haproxy sends OPTIONS request it reads only status code and nothing more
             message = self.responses[status_code][0]
@@ -103,6 +129,10 @@ class RestApiHandler(BaseHTTPRequestHandler):
         else:
             self._write_status_response(status_code, response)
 
+        time_end = time.time()
+        self.log_message('%s %s %s latency: %s ms', request_type, path,
+                         status_code, (time_end - time_start) * 1000)
+
     def do_OPTIONS(self):
         self.do_GET(write_status_code_only=True)
 
@@ -219,9 +249,11 @@ class RestApiHandler(BaseHTTPRequestHandler):
                     data = "PostgreSQL role should be either master or replica"
                     break
             elif k == 'postgres_version':
-                if not is_valid_pg_version(request[k]):
+                try:
+                    postgres_version_to_int(request[k])
+                except PostgresException as e:
                     status_code = 400
-                    data = "PostgreSQL version should be in the first.major.minor format"
+                    data = e.value
                     break
             elif k == 'timeout':
                 request[k] = parse_int(request[k], 's')
@@ -262,7 +294,14 @@ class RestApiHandler(BaseHTTPRequestHandler):
 
     @check_auth
     def do_POST_reinitialize(self):
-        data = self.server.patroni.ha.reinitialize()
+        request = self._read_json_content(body_is_optional=True)
+
+        if request:
+            logger.debug('received reinitialize request: %s', request)
+
+        force = isinstance(request, dict) and parse_bool(request.get('force')) or False
+
+        data = self.server.patroni.ha.reinitialize(force)
         if data is None:
             status_code = 200
             data = 'reinitialize started'
@@ -270,43 +309,50 @@ class RestApiHandler(BaseHTTPRequestHandler):
             status_code = 503
         self._write_response(status_code, data)
 
-    def poll_failover_result(self, leader, candidate):
+    def poll_failover_result(self, leader, candidate, action):
         timeout = max(10, self.server.patroni.dcs.loop_wait)
         for _ in range(0, timeout*2):
             time.sleep(1)
             try:
                 cluster = self.server.patroni.dcs.get_cluster()
-                if cluster.leader and cluster.leader.name != leader:
+                if not cluster.is_unlocked() and cluster.leader.name != leader:
                     if not candidate or candidate == cluster.leader.name:
-                        return 200, 'Successfully failed over to "{0}"'.format(cluster.leader.name)
+                        return 200, 'Successfully {0}ed over to "{1}"'.format(action[:-4], cluster.leader.name)
                     else:
-                        return 200, 'Failed over to "{0}" instead of "{1}"'.format(cluster.leader.name, candidate)
+                        return 200, '{0}ed over to "{1}" instead of "{2}"'.format(action[:-4].title(),
+                                                                                  cluster.leader.name, candidate)
                 if not cluster.failover:
-                    return 503, 'Failover failed'
+                    return 503, action.title() + ' failed'
             except Exception as e:
-                logger.debug('Exception occured during polling failover result: %s', e)
+                logger.debug('Exception occured during polling %s result: %s', action, e)
-        return 503, 'Failover status unknown'
+        return 503, action.title() + ' status unknown'
 
-    def is_failover_possible(self, cluster, leader, candidate):
+    def is_failover_possible(self, cluster, leader, candidate, action):
         if leader and (not cluster.leader or cluster.leader.name != leader):
             return 'leader name does not match'
         if candidate:
+            if action == 'switchover' and cluster.is_synchronous_mode() and cluster.sync.sync_standby != candidate:
+                return 'candidate name does not match with sync_standby'
             members = [m for m in cluster.members if m.name == candidate]
             if not members:
                 return 'candidate does not exists'
+        elif cluster.is_synchronous_mode():
+            members = [m for m in cluster.members if m.name == cluster.sync.sync_standby]
+            if not members:
+                return action + ' is not possible: can not find sync_standby'
         else:
             members = [m for m in cluster.members if m.name != cluster.leader.name and m.api_url]
             if not members:
-                return 'failover is not possible: cluster does not have members except leader'
+                return action + ' is not possible: cluster does not have members except leader'
-        for _, reachable, _, _, tags in self.server.patroni.ha.fetch_nodes_statuses(members):
+        for st in self.server.patroni.ha.fetch_nodes_statuses(members):
-            if reachable and not tags.get('nofailover', False):
+            if st.failover_limitation() is None:
                 return None
-        return 'failover is not possible: no good candidates have been found'
+        return action + ' is not possible: no good candidates have been found'
 
     @check_auth
-    def do_POST_failover(self):
+    def do_POST_failover(self, action='failover'):
         request = self._read_json_content()
-        status_code = 500
+        (status_code, data) = (400, '')
         if not request:
             return
 
@@ -315,39 +361,47 @@ class RestApiHandler(BaseHTTPRequestHandler):
         scheduled_at = request.get('scheduled_at')
         cluster = self.server.patroni.dcs.get_cluster()
 
-        if scheduled_at and cluster.is_paused():
+        logger.info("received %s request with leader=%s candidate=%s scheduled_at=%s",
-            self._write_response(status_code, "Can't schedule failover in the paused state")
+                    action, leader, candidate, scheduled_at)
 
-        logger.info("received failover request with leader=%s candidate=%s scheduled_at=%s",
+        if action == 'failover' and not candidate:
-                    leader, candidate, scheduled_at)
+            data = 'Failover could be performed only to a specific candidate'
+        elif action == 'switchover' and not leader:
+            data = 'Switchover could be performed only from a specific leader'
 
-        data = ''
+        if not data and scheduled_at:
-        if leader or candidate:
+            if not leader:
-            if scheduled_at:
+                data = 'Scheduled {0} is possible only from a specific leader'.format(action)
-                (_, data, scheduled_at) = self.parse_schedule(scheduled_at, "failover")
+            if not data and cluster.is_paused():
-                if _:
+                data = "Can't schedule {0} in the paused state".format(action)
-                    status_code = _
+            if not data:
-                elif self.server.patroni.dcs.manual_failover(leader, candidate, scheduled_at=scheduled_at):
+                (status_code, data, scheduled_at) = self.parse_schedule(scheduled_at, action)
-                    self.server.patroni.ha.wakeup()
+
-                    data = 'Failover scheduled'
+        if not data and cluster.is_paused() and not candidate:
+            data = action.title() + ' is possible only to a specific candidate in a paused state'
+
+        if not data and not scheduled_at:
+            data = self.is_failover_possible(cluster, leader, candidate, action)
+            if data:
+                status_code = 412
+
+        if not data:
+            if self.server.patroni.dcs.manual_failover(leader, candidate, scheduled_at=scheduled_at):
+                self.server.patroni.ha.wakeup()
+                if scheduled_at:
+                    data = action.title() + ' scheduled'
                     status_code = 202
                 else:
-                    data = 'failed to write failover key into DCS'
+                    status_code, data = self.poll_failover_result(cluster.leader and cluster.leader.name,
-                    status_code = 503
+                                                                  candidate, action)
             else:
-                data = self.is_failover_possible(cluster, leader, candidate)
+                data = 'failed to write {0} key into DCS'.format(action)
-                if not data:
+                status_code = 503
-                    if self.server.patroni.dcs.manual_failover(leader, candidate):
-                        self.server.patroni.ha.wakeup()
-                        status_code, data = self.poll_failover_result(cluster.leader and cluster.leader.name, candidate)
-                    else:
-                        data = 'failed to write failover key into DCS'
-                        status_code = 503
-        else:
-            status_code = 400
-            data = 'No values given for required parameters leader and candidate'
         self._write_response(status_code, data)
 
+    def do_POST_switchover(self):
+        self.do_POST_failover(action='switchover')
+
     def parse_request(self):
         """Override parse_request method to enrich basic functionality of `BaseHTTPRequestHandler` class
 
@@ -374,38 +428,54 @@ class RestApiHandler(BaseHTTPRequestHandler):
 
     def get_postgresql_status(self, retry=False):
         try:
-            row = self.query("""WITH replication_info AS (
+            cluster = self.server.patroni.dcs.cluster
-                                    SELECT usename, application_name, client_addr, state, sync_state, sync_priority
+
-                                      FROM pg_stat_replication
+            if self.server.patroni.postgresql.state not in ('running', 'restarting', 'starting'):
-                                )
+                raise RetryFailedError('')
-                                SELECT to_char(pg_postmaster_start_time(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),
+            stmt = ("WITH replication_info AS ("
-                                       pg_is_in_recovery(),
+                    "SELECT usename, application_name, client_addr, state, sync_state, sync_priority"
-                                       CASE WHEN pg_is_in_recovery()
+                    " FROM pg_catalog.pg_stat_replication) SELECT"
-                                            THEN 0
+                    " pg_catalog.to_char(pg_catalog.pg_postmaster_start_time(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),"
-                                            ELSE pg_xlog_location_diff(pg_current_xlog_location(), '0/0')::bigint
+                    " CASE WHEN pg_catalog.pg_is_in_recovery() THEN 0"
-                                       END,
+                    " ELSE ('x' || pg_catalog.substr(pg_catalog.pg_{0}file_name("
-                                       pg_xlog_location_diff(COALESCE(pg_last_xlog_receive_location(),
+                    "pg_catalog.pg_current_{0}_{1}()), 1, 8))::bit(32)::int END,"
-                                                                      pg_last_xlog_replay_location()), '0/0')::bigint,
+                    " CASE WHEN pg_catalog.pg_is_in_recovery() THEN 0"
-                                       pg_xlog_location_diff(pg_last_xlog_replay_location(), '0/0')::bigint,
+                    " ELSE pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_current_{0}_{1}(), '0/0')::bigint END,"
-                                       to_char(pg_last_xact_replay_timestamp(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),
+                    " pg_catalog.pg_{0}_{1}_diff(COALESCE(pg_catalog.pg_last_{0}_receive_{1}(),"
-                                       pg_is_in_recovery() AND pg_is_xlog_replay_paused(),
+                    " pg_catalog.pg_last_{0}_replay_{1}()), '0/0')::bigint,"
-                                       (SELECT array_to_json(array_agg(row_to_json(ri))) FROM replication_info ri)""",
+                    " pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_last_{0}_replay_{1}(), '0/0')::bigint,"
-                             retry=retry)[0]
+                    " pg_catalog.to_char(pg_catalog.pg_last_xact_replay_timestamp(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),"
+                    " pg_catalog.pg_is_in_recovery() AND pg_catalog.pg_is_{0}_replay_paused(), "
+                    "(SELECT pg_catalog.array_to_json(pg_catalog.array_agg("
+                    "pg_catalog.row_to_json(ri))) FROM replication_info ri)")
+
+            row = self.query(stmt.format(self.server.patroni.postgresql.wal_name,
+                                         self.server.patroni.postgresql.lsn_name), retry=retry)[0]
 
             result = {
                 'state': self.server.patroni.postgresql.state,
                 'postmaster_start_time': row[0],
-                'role': 'replica' if row[1] else 'master',
+                'role': 'replica' if row[1] == 0 else 'master',
                 'server_version': self.server.patroni.postgresql.server_version,
+                'cluster_unlocked': bool(not cluster or cluster.is_unlocked()),
                 'xlog': ({
                     'received_location': row[3],
                     'replayed_location': row[4],
                     'replayed_timestamp': row[5],
-                    'paused': row[6]} if row[1] else {
+                    'paused': row[6]} if row[1] == 0 else {
                     'location': row[2]
                 })
             }
 
+            if result['role'] == 'replica' and self.server.patroni.ha.is_standby_cluster():
+                result['role'] = self.server.patroni.postgresql.role
+
+            if row[1] > 0:
+                result['timeline'] = row[1]
+            else:
+                leader_timeline = None if not cluster or cluster.is_unlocked() else cluster.leader.timeline
+                result['timeline'] = self.server.patroni.postgresql.replica_cached_timeline(leader_timeline)
+
             if row[7]:
                 result['replication'] = row[7]
 
@@ -425,6 +495,7 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
 
     def __init__(self, patroni, config):
         self.patroni = patroni
+        self.__listen = None
         self.__initialize(config)
         self.__set_config_parameters(config)
         self.daemon = True
@@ -442,8 +513,10 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
 
     @staticmethod
     def _set_fd_cloexec(fd):
-        flags = fcntl.fcntl(fd, fcntl.F_GETFD)
+        if os.name != 'nt':
-        fcntl.fcntl(fd, fcntl.F_SETFD, flags | fcntl.FD_CLOEXEC)
+            import fcntl
+            flags = fcntl.fcntl(fd, fcntl.F_GETFD)
+            fcntl.fcntl(fd, fcntl.F_SETFD, flags | fcntl.FD_CLOEXEC)
 
     def check_basic_auth_key(self, key):
         return self.__auth_key == key
@@ -459,18 +532,51 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
     def __get_ssl_options(config):
         return {option: config[option] for option in ['certfile', 'keyfile'] if option in config}
 
-    def __set_connection_string(self, connect_address):
-        self.connection_string = '{0}://{1}/patroni'.format(self.__protocol, connect_address or self.__listen)
-
     def __set_config_parameters(self, config):
         self.__auth_key = base64.b64encode(config['auth'].encode('utf-8')).decode('utf-8') if 'auth' in config else None
-        self.__set_connection_string(config.get('connect_address'))
+        self.connection_string = uri(self.__protocol, config.get('connect_address') or self.__listen, 'patroni')
+
+    @staticmethod
+    def __has_dual_stack():
+        if hasattr(socket, 'AF_INET6') and hasattr(socket, 'IPPROTO_IPV6') and hasattr(socket, 'IPV6_V6ONLY'):
+            sock = None
+            try:
+                sock = socket.socket(socket.AF_INET6, socket.SOCK_STREAM)
+                sock.setsockopt(socket.IPPROTO_IPV6, socket.IPV6_V6ONLY, False)
+                return True
+            except socket.error as e:
+                logger.debug('Error when working with ipv6 socket: %s', e)
+            finally:
+                if sock:
+                    sock.close()
+        return False
+
+    def __httpserver_init(self, host, port):
+        dual_stack = self.__has_dual_stack()
+        if host == '':
+            host = None
+
+        info = socket.getaddrinfo(host, port, socket.AF_UNSPEC, socket.SOCK_STREAM, 0, socket.AI_PASSIVE)
+        # in case dual stack is not supported we want IPv4 to be preferred over IPv6
+        info.sort(key=lambda x: x[0] == socket.AF_INET, reverse=not dual_stack)
+
+        self.address_family = info[0][0]
+        HTTPServer.__init__(self, info[0][-1][:2], RestApiHandler)
 
     def __initialize(self, config):
-        self.__ssl_options = self.__get_ssl_options(config)
+        try:
+            host, port = split_host_port(config['listen'], None)
+        except Exception:
+            raise ValueError('Invalid "restapi" config: expected <HOST>:<PORT> for "listen", but got "{0}"'
+                             .format(config['listen']))
+
+        if self.__listen is not None:  # changing config in runtime
+            self.shutdown()
+
         self.__listen = config['listen']
-        host, port = config['listen'].split(':')
+        self.__ssl_options = self.__get_ssl_options(config)
-        HTTPServer.__init__(self, (host, int(port)), RestApiHandler)
+
+        self.__httpserver_init(host, port)
         Thread.__init__(self, target=self.serve_forever)
         self._set_fd_cloexec(self.socket)
 
@@ -480,13 +586,23 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
         # Sometime it's also needed to pass reference to a 'keyfile'.
         if self.__ssl_options.get('certfile'):
             import ssl
-            self.socket = ssl.wrap_socket(self.socket, server_side=True, **self.__ssl_options)
+            ctx = ssl.create_default_context(ssl.Purpose.CLIENT_AUTH)
+            ctx.load_cert_chain(**self.__ssl_options)
+            self.socket = ctx.wrap_socket(self.socket, server_side=True)
             self.__protocol = 'https'
-        self.__set_connection_string(config.get('connect_address'))
+        return True
 
     def reload_config(self, config):
-        self.__set_config_parameters(config)
+        if 'listen' not in config:  # changing config in runtime
-        if self.__listen != config['listen'] or self.__ssl_options != self.__get_ssl_options(config):
+            raise ValueError('Can not find "restapi.listen" config')
-            self.shutdown()
+
-            self.__initialize(config)
+        elif (self.__listen != config['listen'] or self.__ssl_options != self.__get_ssl_options(config)) \
+                and self.__initialize(config):
             self.start()
+        self.__set_config_parameters(config)
+
+    @staticmethod
+    def handle_error(request, client_address):
+        address, port = client_address
+        logger.warning('Exception happened during processing of request from {}:{}'.format(address, port))
+        logger.warning(traceback.format_exc())

patroni/async_executor.py

@@ -1,26 +1,78 @@
 import logging
-from threading import RLock, Thread
+from threading import Event, Lock, RLock, Thread
 
 logger = logging.getLogger(__name__)
 
 
+class CriticalTask(object):
+    """Represents a critical task in a background process that we either need to cancel or get the result of.
+
+    Fields of this object may be accessed only when holding a lock on it. To perform the critical task the background
+    thread must, while holding lock on this object, check `is_cancelled` flag, run the task and mark the task as
+    complete using `complete()`.
+
+    The main thread must hold async lock to prevent the task from completing, hold lock on critical task object,
+    call cancel. If the task has completed `cancel()` will return False and `result` field will contain the result of
+    the task. When cancel returns True it is guaranteed that the background task will notice the `is_cancelled` flag.
+    """
+    def __init__(self):
+        self._lock = Lock()
+        self.is_cancelled = False
+        self.result = None
+
+    def reset(self):
+        """Must be called every time the background task is finished.
+
+        Must be called from async thread. Caller must hold lock on async executor when calling."""
+        self.is_cancelled = False
+        self.result = None
+
+    def cancel(self):
+        """Tries to cancel the task, returns True if the task has already run.
+
+        Caller must hold lock on async executor and the task when calling."""
+        if self.result is not None:
+            return False
+        self.is_cancelled = True
+        return True
+
+    def complete(self, result):
+        """Mark task as completed along with a result.
+
+        Must be called from async thread. Caller must hold lock on task when calling."""
+        self.result = result
+
+    def __enter__(self):
+        self._lock.acquire()
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        self._lock.release()
+
+
 class AsyncExecutor(object):
 
-    def __init__(self, ha_wakeup):
+    def __init__(self, cancellable, ha_wakeup):
+        self._cancellable = cancellable
         self._ha_wakeup = ha_wakeup
         self._thread_lock = RLock()
         self._scheduled_action = None
         self._scheduled_action_lock = RLock()
+        self._is_cancelled = False
+        self._finish_event = Event()
+        self.critical_task = CriticalTask()
 
     @property
     def busy(self):
         return self.scheduled_action is not None
 
-    def schedule(self, action, immediately=False):
+    def schedule(self, action):
         with self._scheduled_action_lock:
             if self._scheduled_action is not None:
                 return self._scheduled_action
             self._scheduled_action = action
+            self._is_cancelled = False
+            self._finish_event.set()
         return None
 
     @property
@@ -35,20 +87,43 @@ class AsyncExecutor(object):
     def run(self, func, args=()):
         wakeup = False
         try:
+            with self:
+                if self._is_cancelled:
+                    return
+                self._finish_event.clear()
+
+            self._cancellable.reset_is_cancelled()
             # if the func returned something (not None) - wake up main HA loop
             wakeup = func(*args) if args else func()
             return wakeup
-        except:
+        except Exception:
             logger.exception('Exception during execution of long running task %s', self.scheduled_action)
         finally:
             with self:
                 self.reset_scheduled_action()
+                self._finish_event.set()
+                with self.critical_task:
+                    self.critical_task.reset()
             if wakeup is not None:
                 self._ha_wakeup()
 
     def run_async(self, func, args=()):
         Thread(target=self.run, args=(func, args)).start()
 
+    def cancel(self):
+        with self:
+            with self._scheduled_action_lock:
+                if self._scheduled_action is None:
+                    return
+                logger.warning('Cancelling long running task %s', self._scheduled_action)
+            self._is_cancelled = True
+
+        self._cancellable.cancel()
+        self._finish_event.wait()
+
+        with self:
+            self.reset_scheduled_action()
+
     def __enter__(self):
         self._thread_lock.acquire()
 

patroni/config.py

@@ -1,6 +1,7 @@
 import json
 import logging
 import os
+import shutil
 import sys
 import tempfile
 import yaml
@@ -8,8 +9,9 @@ import yaml
 from collections import defaultdict
 from copy import deepcopy
 from patroni.dcs import ClusterConfig
-from patroni.postgresql import Postgresql
+from patroni.postgresql.config import ConfigHandler
-from patroni.utils import deep_compare, parse_int, patch_config
+from patroni.utils import deep_compare, parse_bool, parse_int, patch_config
+from requests.structures import CaseInsensitiveDict
 
 logger = logging.getLogger(__name__)
 
@@ -41,12 +43,26 @@ class Config(object):
     __DEFAULT_CONFIG = {
         'ttl': 30, 'loop_wait': 10, 'retry_timeout': 10,
         'maximum_lag_on_failover': 1048576,
+        'check_timeline': False,
         'master_start_timeout': 300,
         'synchronous_mode': False,
+        'synchronous_mode_strict': False,
+        'standby_cluster': {
+            'create_replica_methods': '',
+            'host': '',
+            'port': '',
+            'primary_slot_name': '',
+            'restore_command': '',
+            'archive_cleanup_command': '',
+            'recovery_min_apply_delay': ''
+        },
         'postgresql': {
             'bin_dir': '',
             'use_slots': True,
-            'parameters': {p: v[0] for p, v in Postgresql.CMDLINE_OPTIONS.items()}
+            'parameters': CaseInsensitiveDict({p: v[0] for p, v in ConfigHandler.CMDLINE_OPTIONS.items()})
+        },
+        'watchdog': {
+            'mode': 'automatic',
         }
     }
 
@@ -83,6 +99,9 @@ class Config(object):
     def dynamic_configuration(self):
         return deepcopy(self._dynamic_configuration)
 
+    def check_mode(self, mode):
+        return bool(parse_bool(self._dynamic_configuration.get(mode)))
+
     def _load_config_file(self):
         """Loads config.yaml from filesystem and applies some values which were set via ENV"""
         with open(self._config_file) as f:
@@ -106,7 +125,7 @@ class Config(object):
                 with os.fdopen(fd, 'w') as f:
                     fd = None
                     json.dump(self.dynamic_configuration, f)
-                tmpfile = os.rename(tmpfile, self._cache_file)
+                tmpfile = shutil.move(tmpfile, self._cache_file)
                 self._cache_needs_saving = False
             except Exception:
                 logger.exception('Exception when saving file: %s', self._cache_file)
@@ -150,6 +169,8 @@ class Config(object):
                     self._local_configuration = configuration
                     self.__effective_configuration = new_configuration
                     return True
+                else:
+                    logger.info('No configuration items changed, nothing to reload.')
             except Exception:
                 logger.exception('Exception when reloading local configuration from %s', self.config_file)
                 if dry_run:
@@ -157,11 +178,9 @@ class Config(object):
 
     @staticmethod
     def _process_postgresql_parameters(parameters, is_local=False):
-        ret = {}
+        return {name: value for name, value in (parameters or {}).items()
-        for name, value in (parameters or {}).items():
+                if name not in ConfigHandler.CMDLINE_OPTIONS or
-            if name not in Postgresql.CMDLINE_OPTIONS or not is_local and Postgresql.CMDLINE_OPTIONS[name][1](value):
+                not is_local and ConfigHandler.CMDLINE_OPTIONS[name][1](value)}
-                ret[name] = value
-        return ret
 
     def _safe_copy_dynamic_configuration(self, dynamic_configuration):
         config = deepcopy(self.__DEFAULT_CONFIG)
@@ -173,8 +192,12 @@ class Config(object):
                         config['postgresql'][name].update(self._process_postgresql_parameters(value))
                     elif name not in ('connect_address', 'listen', 'data_dir', 'pgpass', 'authentication'):
                         config['postgresql'][name] = deepcopy(value)
+            elif name == 'standby_cluster':
+                for name, value in (value or {}).items():
+                    if name in self.__DEFAULT_CONFIG['standby_cluster']:
+                        config['standby_cluster'][name] = deepcopy(value)
             elif name in config:  # only variables present in __DEFAULT_CONFIG allowed to be overriden from DCS
-                if name == 'synchronous_mode':
+                if name in ('synchronous_mode', 'synchronous_mode_strict'):
                     config[name] = value
                 else:
                     config[name] = int(value)
@@ -192,6 +215,15 @@ class Config(object):
             if value:
                 ret[param] = value
 
+        def _fix_log_env(name, oldname):
+            value = _popenv(oldname)
+            name = Config.PATRONI_ENV_PREFIX + 'LOG_' + name.upper()
+            if value and name not in os.environ:
+                os.environ[name] = value
+
+        for name, oldname in (('level', 'loglevel'), ('format', 'logformat'), ('dateformat', 'log_datefmt')):
+            _fix_log_env(name, oldname)
+
         def _set_section_values(section, params):
             for param in params:
                 value = _popenv(section + '_' + param)
@@ -199,7 +231,24 @@ class Config(object):
                     ret[section][param] = value
 
         _set_section_values('restapi', ['listen', 'connect_address', 'certfile', 'keyfile'])
-        _set_section_values('postgresql', ['listen', 'connect_address', 'data_dir', 'pgpass', 'bin_dir'])
+        _set_section_values('postgresql', ['listen', 'connect_address', 'config_dir', 'data_dir', 'pgpass', 'bin_dir'])
+        _set_section_values('log', ['level', 'format', 'dateformat', 'max_queue_size',
+                                    'dir', 'file_size', 'file_num', 'loggers'])
+
+        def _parse_dict(value):
+            if not value.strip().startswith('{'):
+                value = '{{{0}}}'.format(value)
+            try:
+                return yaml.safe_load(value)
+            except Exception:
+                logger.exception('Exception when parsing dict %s', value)
+                return None
+
+        value = ret.get('log', {}).pop('loggers', None)
+        if value:
+            value = _parse_dict(value)
+            if value:
+                ret['log']['loggers'] = value
 
         def _get_auth(name):
             ret = {}
@@ -214,7 +263,7 @@ class Config(object):
             ret['restapi']['authentication'] = restapi_auth
 
         authentication = {}
-        for user_type in ('replication', 'superuser'):
+        for user_type in ('replication', 'superuser', 'rewind'):
             entry = _get_auth(user_type)
             if entry:
                 authentication[user_type] = entry
@@ -222,8 +271,6 @@ class Config(object):
         if authentication:
             ret['postgresql']['authentication'] = authentication
 
-        users = {}
-
         def _parse_list(value):
             if not (value.strip().startswith('-') or '[' in value):
                 value = '[{0}]'.format(value)
@@ -233,30 +280,42 @@ class Config(object):
                 logger.exception('Exception when parsing list %s', value)
                 return None
 
+        for param in list(os.environ.keys()):
+            if param.startswith(Config.PATRONI_ENV_PREFIX):
+                # PATRONI_(ETCD|CONSUL|ZOOKEEPER|EXHIBITOR|...)_(HOSTS?|PORT|..)
+                name, suffix = (param[8:].split('_', 1) + [''])[:2]
+                if suffix in ('HOST', 'HOSTS', 'PORT', 'USE_PROXIES', 'PROTOCOL', 'SRV', 'URL', 'PROXY',
+                              'CACERT', 'CERT', 'KEY', 'VERIFY', 'TOKEN', 'CHECKS', 'DC', 'CONSISTENCY',
+                              'REGISTER_SERVICE', 'SERVICE_CHECK_INTERVAL', 'NAMESPACE', 'CONTEXT',
+                              'USE_ENDPOINTS', 'SCOPE_LABEL', 'ROLE_LABEL', 'POD_IP', 'PORTS', 'LABELS') and name:
+                    value = os.environ.pop(param)
+                    if suffix == 'PORT':
+                        value = value and parse_int(value)
+                    elif suffix in ('HOSTS', 'PORTS', 'CHECKS'):
+                        value = value and _parse_list(value)
+                    elif suffix == 'LABELS':
+                        value = _parse_dict(value)
+                    elif suffix in ('USE_PROXIES', 'REGISTER_SERVICE'):
+                        value = parse_bool(value)
+                    if value:
+                        ret[name.lower()][suffix.lower()] = value
+        if 'etcd' in ret:
+            ret['etcd'].update(_get_auth('etcd'))
+
+        users = {}
         for param in list(os.environ.keys()):
             if param.startswith(Config.PATRONI_ENV_PREFIX):
                 name, suffix = (param[8:].rsplit('_', 1) + [''])[:2]
-                if name and suffix:
+                # PATRONI_<username>_PASSWORD=<password>, PATRONI_<username>_OPTIONS=<option1,option2,...>
-                    # PATRONI_(ETCD|CONSUL|ZOOKEEPER|EXHIBITOR|...)_(HOSTS?|PORT|..)
+                # CREATE USER "<username>" WITH <OPTIONS> PASSWORD '<password>'
-                    if suffix in ('HOST', 'HOSTS', 'PORT', 'SRV', 'URL', 'PROXY', 'CACERT', 'CERT', 'KEY') \
+                if name and suffix == 'PASSWORD':
-                            and '_' not in name:
+                    password = os.environ.pop(param)
-                        value = os.environ.pop(param)
+                    if password:
-                        if suffix == 'PORT':
+                        users[name] = {'password': password}
-                            value = value and parse_int(value)
+                        options = os.environ.pop(param[:-9] + '_OPTIONS', None)
-                        elif suffix == 'HOSTS':
+                        options = options and _parse_list(options)
-                            value = value and _parse_list(value)
+                        if options:
-                        if value:
+                            users[name]['options'] = options
-                            ret[name.lower()][suffix.lower()] = value
-                    # PATRONI_<username>_PASSWORD=<password>, PATRONI_<username>_OPTIONS=<option1,option2,...>
-                    # CREATE USER "<username>" WITH <OPTIONS> PASSWORD '<password>'
-                    elif suffix == 'PASSWORD':
-                        password = os.environ.pop(param)
-                        if password:
-                            users[name] = {'password': password}
-                            options = os.environ.pop(param[:-9] + '_OPTIONS', None)
-                            options = options and _parse_list(options)
-                            if options:
-                                users[name]['options'] = options
         if users:
             ret['bootstrap']['users'] = users
 
@@ -271,8 +330,13 @@ class Config(object):
                         config['postgresql'][name].update(self._process_postgresql_parameters(value, True))
                     elif name != 'use_slots':  # replication slots must be enabled/disabled globally
                         config['postgresql'][name] = deepcopy(value)
-            elif name not in config:
+            elif name not in config or name in ['watchdog']:
                 config[name] = deepcopy(value) if value else {}
+        
+        if os.getenv("PATRONI_DEBUG_MODE"):
+            config['ttl'] = 24 * 60 * 60 # practical infinity for a debugging session
+            config['loop_wait'] = -1
+
 
         # restapi server expects to get restapi.auth = 'username:password'
         if 'authentication' in config['restapi']:
@@ -298,8 +362,15 @@ class Config(object):
         if 'name' not in config and 'name' in pg_config:
             config['name'] = pg_config['name']
 
-        pg_config.update({p: config[p] for p in ('name', 'scope', 'retry_timeout',
+        updated_fields = (
-                          'synchronous_mode', 'maximum_lag_on_failover') if p in config})
+            'name',
+            'scope',
+            'retry_timeout',
+            'synchronous_mode',
+            'synchronous_mode_strict',
+        )
+
+        pg_config.update({p: config[p] for p in updated_fields if p in config})
 
         return config
 

patroni/ctl.py

@@ -4,27 +4,37 @@ Patroni Control
 
 import base64
 import click
+import codecs
 import datetime
 import dateutil.parser
+import cdiff
+import copy
+import difflib
+import io
 import json
 import logging
 import os
-import psycopg2
 import random
 import requests
+import subprocess
 import sys
+import tempfile
 import time
 import tzlocal
 import yaml
 
 from click import ClickException
+from contextlib import contextmanager
 from patroni.config import Config
 from patroni.dcs import get_dcs as _get_dcs
 from patroni.exceptions import PatroniException
 from patroni.postgresql import Postgresql
-from patroni.utils import is_valid_pg_version
+from patroni.postgresql.misc import postgres_version_to_int
+from patroni.utils import patch_config, polling_loop
+from patroni.version import __version__
 from prettytable import PrettyTable
 from six.moves.urllib_parse import urlparse
+from six import text_type
 
 CONFIG_DIR_PATH = click.get_app_dir('patroni')
 CONFIG_FILE_PATH = os.path.join(CONFIG_DIR_PATH, 'patronictl.yaml')
@@ -53,12 +63,15 @@ def parse_dcs(dcs):
     elif scheme not in DCS_DEFAULTS:
         raise PatroniCtlException('Unknown dcs scheme: {}'.format(scheme))
 
-    dcs_info = DCS_DEFAULTS[scheme]
+    default = DCS_DEFAULTS[scheme]
-    return yaml.load(dcs_info['template'].format(host=parsed.hostname or 'localhost', port=port or dcs_info['port']))
+    return yaml.safe_load(default['template'].format(host=parsed.hostname or 'localhost', port=port or default['port']))
 
 
 def load_config(path, dcs):
-    logging.debug('Loading configuration from file %s', path)
+    if not (os.path.exists(path) and os.access(path, os.R_OK)):
+        logging.debug('Ignoring configuration file "%s". It does not exists or is not readable.', path)
+    else:
+        logging.debug('Loading configuration from file %s', path)
     config = {}
     old_argv = list(sys.argv)
     try:
@@ -87,19 +100,30 @@ def store_config(config, path):
         yaml.dump(config, fd)
 
 
-option_format = click.option('--format', '-f', 'fmt', help='Output format (pretty, json)', default='pretty')
+option_format = click.option('--format', '-f', 'fmt', help='Output format (pretty, json, yaml)', default='pretty')
 option_watchrefresh = click.option('-w', '--watch', type=float, help='Auto update the screen every X seconds')
 option_watch = click.option('-W', is_flag=True, help='Auto update the screen every 2 seconds')
 option_force = click.option('--force', is_flag=True, help='Do not ask for confirmation at any point')
+arg_cluster_name = click.argument('cluster_name', required=False,
+                                  default=lambda: click.get_current_context().obj.get('scope'))
+option_insecure = click.option('-k', '--insecure', is_flag=True, help='Allow connections to SSL sites without certs')
 
 
 @click.group()
-@click.option('--config-file', '-c', help='Configuration file', default=CONFIG_FILE_PATH)
+@click.option('--config-file', '-c', help='Configuration file',
+              envvar='PATRONICTL_CONFIG_FILE', default=CONFIG_FILE_PATH)
 @click.option('--dcs', '-d', help='Use this DCS', envvar='DCS')
+@option_insecure
 @click.pass_context
-def ctl(ctx, config_file, dcs):
+def ctl(ctx, config_file, dcs, insecure):
-    logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s', level=os.environ.get('LOGLEVEL', 'WARNING'))
+    level = 'WARNING'
+    for name in ('LOGLEVEL', 'PATRONI_LOGLEVEL', 'PATRONI_LOG_LEVEL'):
+        level = os.environ.get(name, level)
+    logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s', level=level)
+    logging.captureWarnings(True)  # Capture eventual SSL warning
     ctx.obj = load_config(config_file, dcs)
+    # backward compatibility for configuration file where ctl section is not define
+    ctx.obj.setdefault('ctl', {})['insecure'] = ctx.obj.get('ctl', {}).get('insecure') or insecure
 
 
 def get_dcs(config, scope):
@@ -117,6 +141,7 @@ def auth_header(config):
 
 
 def request_patroni(member, request_type, endpoint, content=None, headers=None):
+    ctx = click.get_current_context()  # the current click context
     headers = headers or {}
     url_parts = urlparse(member.api_url)
     logging.debug(url_parts)
@@ -125,8 +150,21 @@ def request_patroni(member, request_type, endpoint, content=None, headers=None):
 
     url = '{0}://{1}/{2}'.format(url_parts.scheme, url_parts.netloc, endpoint)
 
+    insecure = ctx.obj.get('ctl', {}).get('insecure', False)
+    # Get certfile if any from several configuration namespace
+    cert = ctx.obj.get('ctl', {}).get('cacert') or \
+        ctx.obj.get('restapi', {}).get('cacert') or \
+        ctx.obj.get('restapi', {}).get('certfile')
+    # In the case we specificaly disable SSL cert verification we don't want to have the warning
+    if insecure:
+        verify = False
+    elif cert:
+        verify = cert
+    else:
+        verify = True
     return getattr(requests, request_type)(url, headers=headers,
-                                           data=json.dumps(content) if content else None, timeout=60)
+                                           data=json.dumps(content) if content else None, timeout=60,
+                                           verify=verify)
 
 
 def print_output(columns, rows=None, alignment=None, fmt='pretty', header=True, delimiter='\t'):
@@ -140,13 +178,16 @@ def print_output(columns, rows=None, alignment=None, fmt='pretty', header=True,
         click.echo(t)
         return
 
-    if fmt == 'json':
+    if fmt in ['json', 'yaml', 'yml']:
         elements = [dict(zip(columns, r)) for r in rows]
-        click.echo(json.dumps(elements))
+        if fmt == 'json':
+            click.echo(json.dumps(elements))
+        elif fmt in ('yaml', 'yml'):
+            click.echo(yaml.safe_dump(elements, encoding=None, default_flow_style=False, allow_unicode=True, width=200))
 
     if fmt == 'tsv':
         if columns is not None and header:
-            click.echo(delimiter.join(columns) + '\n')
+            click.echo(delimiter.join(columns))
 
         for r in rows:
             c = [str(c) for c in r]
@@ -212,13 +253,14 @@ def get_cursor(cluster, connect_parameters, role='master', member=None):
     else:
         params.pop('database')
 
+    import psycopg2
     conn = psycopg2.connect(**params)
     conn.autocommit = True
     cursor = conn.cursor()
     if role == 'any':
         return cursor
 
-    cursor.execute('SELECT pg_is_in_recovery()')
+    cursor.execute('SELECT pg_catalog.pg_is_in_recovery()')
     in_recovery = cursor.fetchone()[0]
 
     if in_recovery and role == 'replica' or not in_recovery and role == 'master':
@@ -229,7 +271,7 @@ def get_cursor(cluster, connect_parameters, role='master', member=None):
     return None
 
 
-def get_members(cluster, cluster_name, member_names, role, force, action):
+def get_members(cluster, cluster_name, member_names, role, force, action, scheduled_at=None):
     candidates = {m.name: m for m in cluster.members}
 
     if not force or role:
@@ -248,14 +290,22 @@ def get_members(cluster, cluster_name, member_names, role, force, action):
         member_names = [click.prompt('Which member do you want to {0} [{1}]?'.format(action,
                         ', '.join(candidates.keys())), type=str, default='')]
 
-    for mn in member_names:
+    for member_name in member_names:
-        if mn not in candidates:
+        if member_name not in candidates:
-            raise PatroniCtlException('{0} is not a member of cluster'.format(mn))
+            raise PatroniCtlException('{0} is not a member of cluster'.format(member_name))
 
-    if not force:
+    if scheduled_at:
-        confirm = click.confirm('Are you sure you want to {0} members {1}?'.format(action, ', '.join(member_names)))
+        if not force:
-        if not confirm:
+            confirm = click.confirm('Are you sure you want to schedule {0} of members {1} at {2}?'
-            raise PatroniCtlException('Aborted {0}'.format(action))
+                                    .format(action, ', '.join(member_names), scheduled_at))
+            if not confirm:
+                raise PatroniCtlException('Aborted scheduled {0}'.format(action))
+    else:
+        if not force:
+            confirm = click.confirm('Are you sure you want to {0} members {1}?'
+                                    .format(action, ', '.join(member_names)))
+            if not confirm:
+                raise PatroniCtlException('Aborted {0}'.format(action))
 
     return [candidates[n] for n in member_names]
 
@@ -264,7 +314,7 @@ def get_members(cluster, cluster_name, member_names, role, force, action):
 @click.option('--role', '-r', help='Give a dsn of any member with this role', type=click.Choice(['master', 'replica',
               'any']), default=None)
 @click.option('--member', '-m', help='Generate a dsn for this member', type=str)
-@click.argument('cluster_name')
+@arg_cluster_name
 @click.pass_obj
 def dsn(obj, cluster_name, role, member):
     if role is not None and member is not None:
@@ -282,7 +332,7 @@ def dsn(obj, cluster_name, role, member):
 
 
 @ctl.command('query', help='Query a Patroni PostgreSQL member')
-@click.argument('cluster_name')
+@arg_cluster_name
 @option_format
 @click.option('--format', 'fmt', help='Output format (pretty, json)', default='tsv')
 @click.option('--file', '-f', 'p_file', help='Execute the SQL commands from this file', type=click.File('rb'))
@@ -346,6 +396,7 @@ def query(
 
 
 def query_member(cluster, cursor, member, role, command, connect_parameters):
+    import psycopg2
     try:
         if cursor is None:
             cursor = get_cursor(cluster, connect_parameters, role=role, member=member)
@@ -358,7 +409,7 @@ def query_member(cluster, cursor, member, role, command, connect_parameters):
             logging.debug(message)
             return [[timestamp(0), message]], None
 
-        cursor.execute('SELECT pg_is_in_recovery()')
+        cursor.execute('SELECT pg_catalog.pg_is_in_recovery()')
         in_recovery = cursor.fetchone()[0]
 
         if in_recovery and role == 'master' or not in_recovery and role == 'replica':
@@ -397,7 +448,7 @@ def remove(obj, cluster_name, fmt):
     if message != confirm:
         raise PatroniCtlException('You did not exactly type "{0}"'.format(message))
 
-    if cluster.leader:
+    if cluster.leader and cluster.leader.name:
         confirm = click.prompt('This cluster currently is healthy. Please specify the master name to continue')
         if confirm != cluster.leader.name:
             raise PatroniCtlException('You did not specify the current master of the cluster')
@@ -410,8 +461,10 @@ def check_response(response, member_name, action_name, silent_success=False):
         click.echo('Failed: {0} for member {1}, status code={2}, ({3})'.format(
             action_name, member_name, response.status_code, response.text
         ))
+        return False
     elif not silent_success:
         click.echo('Success: {0} for member {1}'.format(action_name, member_name))
+    return True
 
 
 def parse_scheduled(scheduled):
@@ -428,6 +481,33 @@ def parse_scheduled(scheduled):
     return None
 
 
+@ctl.command('reload', help='Reload cluster member configuration')
+@click.argument('cluster_name')
+@click.argument('member_names', nargs=-1)
+@click.option('--role', '-r', help='Reload only members with this role', default='any',
+              type=click.Choice(['master', 'replica', 'any']))
+@option_force
+@click.pass_obj
+def reload(obj, cluster_name, member_names, force, role):
+    cluster = get_dcs(obj, cluster_name).get_cluster()
+
+    members = get_members(cluster, cluster_name, member_names, role, force, 'reload')
+
+    content = {}
+    for member in members:
+        r = request_patroni(member, 'post', 'reload', content, auth_header(obj))
+        if r.status_code == 200:
+            click.echo('No changes to apply on member {0}'.format(member.name))
+        elif r.status_code == 202:
+            click.echo('Reload request received for member {0} and will be processed within {1} seconds'.format(
+                member.name, cluster.config.data.get('loop_wait'))
+            )
+        else:
+            click.echo('Failed: reload for member {0}, status code={1}, ({2})'.format(
+                member.name, r.status_code, r.text)
+            )
+
+
 @ctl.command('restart', help='Restart cluster member')
 @click.argument('cluster_name')
 @click.argument('member_names', nargs=-1)
@@ -446,7 +526,14 @@ def parse_scheduled(scheduled):
 def restart(obj, cluster_name, member_names, force, role, p_any, scheduled, version, pending, timeout):
     cluster = get_dcs(obj, cluster_name).get_cluster()
 
-    members = get_members(cluster, cluster_name, member_names, role, force, 'restart')
+    if scheduled is None and not force:
+        next_hour = (datetime.datetime.now() + datetime.timedelta(hours=1)).strftime('%Y-%m-%dT%H:%M')
+        scheduled = click.prompt('When should the restart take place (e.g. ' + next_hour + ') ',
+                                 type=str, default='now')
+
+    scheduled_at = parse_scheduled(scheduled)
+
+    members = get_members(cluster, cluster_name, member_names, role, force, 'restart', scheduled_at)
     if p_any:
         random.shuffle(members)
         members = members[:1]
@@ -460,16 +547,13 @@ def restart(obj, cluster_name, member_names, force, role, p_any, scheduled, vers
         content['restart_pending'] = True
 
     if version:
-        if not is_valid_pg_version(version):
+        try:
-            message = 'PostgreSQL version should be in the first.major.minor format'
+            postgres_version_to_int(version)
-            raise PatroniCtlException(message)
+        except PatroniException as e:
-        else:
+            raise PatroniCtlException(e.value)
-            content['postgres_version'] = version
 
-    if scheduled is None and not force:
+        content['postgres_version'] = version
-        scheduled = click.prompt('When should the restart take place (e.g. 2015-10-01T14:30) ', type=str, default='now')
 
-    scheduled_at = parse_scheduled(scheduled)
     if scheduled_at:
         if cluster.is_paused():
             raise PatroniCtlException("Can't schedule restart in the paused state")
@@ -507,68 +591,73 @@ def reinit(obj, cluster_name, member_names, force):
     members = get_members(cluster, cluster_name, member_names, None, force, 'reinitialize')
 
     for member in members:
-        r = request_patroni(member, 'post', 'reinitialize', headers=auth_header(obj))
+        body = {'force': force}
-        check_response(r, member.name, 'reinitialize')
+        while True:
+            r = request_patroni(member, 'post', 'reinitialize', body, auth_header(obj))
+            if not check_response(r, member.name, 'reinitialize') and r.text.endswith(' already in progress') \
+                    and not force and click.confirm('Do you want to cancel it and reinitialize anyway?'):
+                body['force'] = True
+                continue
+            break
 
 
-@ctl.command('failover', help='Failover to a replica')
+def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, force, scheduled=None):
-@click.argument('cluster_name')
-@click.option('--master', help='The name of the current master', default=None)
-@click.option('--candidate', help='The name of the candidate', default=None)
-@click.option('--scheduled', help='Timestamp of a scheduled failover in unambiguous format (e.g. ISO 8601)',
-              default=None)
-@option_force
-@click.pass_obj
-def failover(obj, cluster_name, master, candidate, force, scheduled):
     """
-        We want to trigger a failover for the specified cluster name.
+        We want to trigger a failover or switchover for the specified cluster name.
 
         We verify that the cluster name, master name and candidate name are correct.
-        If so, we trigger a failover and keep the client up to date.
+        If so, we trigger an action and keep the client up to date.
     """
 
     dcs = get_dcs(obj, cluster_name)
     cluster = dcs.get_cluster()
 
-    if cluster.leader is None and not cluster.is_paused():
+    if action == 'switchover' and cluster.leader is None:
         raise PatroniCtlException('This cluster has no master')
 
-    if master is None and (not cluster.is_paused() or cluster.leader):
+    if master is None:
-        if force:
+        if force or action == 'failover':
-            master = cluster.leader.member.name
+            master = cluster.leader and cluster.leader.name
         else:
             master = click.prompt('Master', type=str, default=cluster.leader.member.name)
 
     if master is not None and cluster.leader and cluster.leader.member.name != master:
         raise PatroniCtlException('Member {0} is not the leader of cluster {1}'.format(master, cluster_name))
 
-    candidate_names = [str(m.name) for m in cluster.members if m.name != master]
+    # excluding members with nofailover tag
+    candidate_names = [str(m.name) for m in cluster.members if m.name != master and not m.nofailover]
     # We sort the names for consistent output to the client
     candidate_names.sort()
 
     if not candidate_names:
-        raise PatroniCtlException('No candidates found to failover to')
+        raise PatroniCtlException('No candidates found to {0} to'.format(action))
 
     if candidate is None and not force:
         candidate = click.prompt('Candidate ' + str(candidate_names), type=str, default='')
 
+    if action == 'failover' and not candidate:
+        raise PatroniCtlException('Failover could be performed only to a specific candidate')
+
     if candidate == master:
-        raise PatroniCtlException('Failover target and source are the same.')
+        raise PatroniCtlException(action.title() + ' target and source are the same.')
 
     if candidate and candidate not in candidate_names:
         raise PatroniCtlException('Member {0} does not exist in cluster {1}'.format(candidate, cluster_name))
 
-    if scheduled is None and not force:
-        scheduled = click.prompt('When should the failover take place (e.g. 2015-10-01T14:30) ', type=str,
-                                 default='now')
-
-    scheduled_at = parse_scheduled(scheduled)
-
     scheduled_at_str = None
-    if scheduled_at:
+    scheduled_at = None
-        if cluster.is_paused():
+
-            raise PatroniCtlException("Can't schedule failover in the paused state")
+    if action == 'switchover':
-        scheduled_at_str = scheduled_at.isoformat()
+        if scheduled is None and not force:
+            next_hour = (datetime.datetime.now() + datetime.timedelta(hours=1)).strftime('%Y-%m-%dT%H:%M')
+            scheduled = click.prompt('When should the switchover take place (e.g. ' + next_hour + ' ) ',
+                                     type=str, default='now')
+
+        scheduled_at = parse_scheduled(scheduled)
+        if scheduled_at:
+            if cluster.is_paused():
+                raise PatroniCtlException("Can't schedule switchover in the paused state")
+            scheduled_at_str = scheduled_at.isoformat()
 
     failover_value = {'leader': master, 'candidate': candidate, 'scheduled_at': scheduled_at_str}
 
@@ -579,46 +668,85 @@ def failover(obj, cluster_name, master, candidate, force, scheduled):
     output_members(dcs.get_cluster(), cluster_name)
 
     if not force:
-        a = \
+        demote_msg = ', demoting current master ' + master if master else ''
-            click.confirm('Are you sure you want to failover cluster {0}, demoting current master {1}?'.format(
+        if scheduled_at_str:
-                cluster_name, master))
+            if not click.confirm('Are you sure you want to schedule {0} of cluster {1} at {2}{3}?'
-        if not a:
+                                 .format(action, cluster_name, scheduled_at_str, demote_msg)):
-            raise PatroniCtlException('Aborting failover')
+                raise PatroniCtlException('Aborting scheduled ' + action)
+        else:
+            if not click.confirm('Are you sure you want to {0} cluster {1}{2}?'
+                                 .format(action, cluster_name, demote_msg)):
+                raise PatroniCtlException('Aborting ' + action)
 
     r = None
     try:
-        member = cluster.leader.member if cluster.leader else [m for m in cluster.members if m.name == candidate][0]
+        member = cluster.leader.member if cluster.leader else cluster.get_member(candidate, False)
+
+        r = request_patroni(member, 'post', action, failover_value, auth_header(obj))
+
+        # probably old patroni, which doesn't support switchover yet
+        if r.status_code == 501 and action == 'switchover' and 'Server does not support this operation' in r.text:
+            r = request_patroni(member, 'post', 'failover', failover_value, auth_header(obj))
 
-        r = request_patroni(member, 'post', 'failover', failover_value, auth_header(obj))
         if r.status_code in (200, 202):
             logging.debug(r)
             cluster = dcs.get_cluster()
             logging.debug(cluster)
             click.echo('{0} {1}'.format(timestamp(), r.text))
         else:
-            click.echo('Failover failed, details: {0}, {1}'.format(r.status_code, r.text))
+            click.echo('{0} failed, details: {1}, {2}'.format(action.title(), r.status_code, r.text))
             return
     except Exception:
         logging.exception(r)
         logging.warning('Failing over to DCS')
-        click.echo(timestamp() + ' Could not failover using Patroni api, falling back to DCS')
+        click.echo('{0} Could not {1} using Patroni api, falling back to DCS'.format(timestamp(), action))
-        click.echo(timestamp() + ' Initializing failover from master {0}'.format(master))
         dcs.manual_failover(master, candidate, scheduled_at=scheduled_at)
 
     output_members(cluster, cluster_name)
 
 
+@ctl.command('failover', help='Failover to a replica')
+@arg_cluster_name
+@click.option('--master', help='The name of the current master', default=None)
+@click.option('--candidate', help='The name of the candidate', default=None)
+@option_force
+@click.pass_obj
+def failover(obj, cluster_name, master, candidate, force):
+    action = 'switchover' if master else 'failover'
+    _do_failover_or_switchover(obj, action, cluster_name, master, candidate, force)
+
+
+@ctl.command('switchover', help='Switchover to a replica')
+@arg_cluster_name
+@click.option('--master', help='The name of the current master', default=None)
+@click.option('--candidate', help='The name of the candidate', default=None)
+@click.option('--scheduled', help='Timestamp of a scheduled switchover in unambiguous format (e.g. ISO 8601)',
+              default=None)
+@option_force
+@click.pass_obj
+def switchover(obj, cluster_name, master, candidate, force, scheduled):
+    _do_failover_or_switchover(obj, 'switchover', cluster_name, master, candidate, force, scheduled)
+
+
 def output_members(cluster, name, extended=False, fmt='pretty'):
     rows = []
     logging.debug(cluster)
     leader_name = None
     if cluster.leader:
-        leader_name = cluster.leader.member.name
+        leader_name = cluster.leader.name
 
     xlog_location_cluster = cluster.last_leader_operation or 0
 
     # Mainly for consistent pretty printing and watching we sort the output
     cluster.members.sort(key=lambda x: x.name)
+
+    has_scheduled_restarts = any(m.data.get('scheduled_restart') for m in cluster.members)
+    has_pending_restarts = any(m.data.get('pending_restart') for m in cluster.members)
+
+    # Show Host as 'host:port' if somebody is running on non-standard port or two nodes are running on the same host
+    append_port = any(str(m.conn_kwargs()['port']) != '5432' for m in cluster.members) or\
+        len(set(m.conn_kwargs()['host'] for m in cluster.members)) < len(cluster.members)
+
     for m in cluster.members:
         logging.debug(m)
 
@@ -628,22 +756,23 @@ def output_members(cluster, name, extended=False, fmt='pretty'):
         elif m.name == cluster.sync.sync_standby:
             role = 'Sync standby'
 
-        host = m.conn_kwargs()['host']
+        xlog_location = m.data.get('xlog_location')
-
-        xlog_location = m.data.get('xlog_location') or 0
         lag = ''
-        if xlog_location_cluster >= xlog_location:
+        if xlog_location is None:
+            lag = 'unknown'
+        elif xlog_location_cluster >= xlog_location:
             lag = round((xlog_location_cluster - xlog_location)/1024/1024)
 
-        row = [
+        host = m.conn_kwargs()['host']
-            name,
+        if append_port:
-            m.name,
+            host += ':{0}'.format(m.conn_kwargs()['port'])
-            host,
+
-            role,
+        row = [name, m.name, host, role, m.data.get('state', ''), m.data.get('timeline', ''), lag]
-            m.data.get('state', ''),
+
-            lag,
+        if extended or has_pending_restarts:
-        ]
+            row.append('*' if m.data.get('pending_restart') else '')
-        if extended:
+
+        if extended or has_scheduled_restarts:
             value = ''
             scheduled_restart = m.data.get('scheduled_restart')
             if scheduled_restart:
@@ -655,39 +784,55 @@ def output_members(cluster, name, extended=False, fmt='pretty'):
 
         rows.append(row)
 
-    columns = [
+    columns = ['Cluster', 'Member', 'Host', 'Role', 'State', 'TL', 'Lag in MB']
-        'Cluster',
+    alignment = {'Lag in MB': 'r', 'TL': 'r'}
-        'Member',
-        'Host',
-        'Role',
-        'State',
-        'Lag in MB',
-    ]
-    alignment = {'Cluster': 'l', 'Member': 'l', 'Host': 'l', 'Lag in MB': 'r'}
 
-    if extended:
+    if extended or has_pending_restarts:
+        columns.append('Pending restart')
+
+    if extended or has_scheduled_restarts:
         columns.append('Scheduled restart')
-        alignment['Scheduled restart'] = 'l'
 
     print_output(columns, rows, alignment, fmt)
 
+    service_info = []
+    if cluster.is_paused():
+        service_info.append('Maintenance mode: on')
+
+    if cluster.failover and cluster.failover.scheduled_at:
+        info = 'Switchover scheduled at: ' + cluster.failover.scheduled_at.isoformat()
+        if cluster.failover.leader:
+            info += '\n                    from: ' + cluster.failover.leader
+        if cluster.failover.candidate:
+            info += '\n                      to: ' + cluster.failover.candidate
+        service_info.append(info)
+
+    if service_info:
+        click.echo(' ' + '\n '.join(service_info))
+
 
 @ctl.command('list', help='List the Patroni members for a given Patroni')
 @click.argument('cluster_names', nargs=-1)
 @click.option('--extended', '-e', help='Show some extra information', is_flag=True)
+@click.option('--timestamp', '-t', 'ts', help='Print timestamp', is_flag=True)
 @option_format
 @option_watch
 @option_watchrefresh
 @click.pass_obj
-def members(obj, cluster_names, fmt, watch, w, extended):
+def members(obj, cluster_names, fmt, watch, w, extended, ts):
     if not cluster_names:
-        logging.warning('Listing members: No cluster names were provided')
+        if 'scope' in obj:
-        return
+            cluster_names = [obj['scope']]
+        if not cluster_names:
+            return logging.warning('Listing members: No cluster names were provided')
 
     for cluster_name in cluster_names:
         dcs = get_dcs(obj, cluster_name)
 
         for _ in watching(w, watch):
+            if ts:
+                click.echo(timestamp(0))
+
             cluster = dcs.get_cluster()
             output_members(cluster, cluster_name, extended, fmt)
 
@@ -723,7 +868,7 @@ def touch_member(config, dcs):
         'role': p.role
     }
 
-    return dcs.touch_member(json.dumps(data, separators=(',', ':')), permanent=True)
+    return dcs.touch_member(data, permanent=True)
 
 
 def set_defaults(config, cluster_name):
@@ -780,8 +925,27 @@ def flush(obj, cluster_name, member_names, force, role, target):
                 click.echo('No scheduled restart for member {0}'.format(member.name))
 
 
-def toggle_pause(config, cluster_name, paused):
+def wait_until_pause_is_applied(dcs, paused, old_cluster):
-    cluster = get_dcs(config, cluster_name).get_cluster()
+    click.echo("'{0}' request sent, waiting until it is recognized by all nodes".format(paused and 'pause' or 'resume'))
+    old = {m.name: m.index for m in old_cluster.members if m.api_url}
+    loop_wait = old_cluster.config.data.get('loop_wait', dcs.loop_wait)
+
+    for _ in polling_loop(loop_wait + 1):
+        cluster = dcs.get_cluster()
+        if all(m.data.get('pause', False) == paused for m in cluster.members if m.name in old):
+            break
+    else:
+        remaining = [m.name for m in cluster.members if m.data.get('pause', False) != paused
+                     and m.name in old and old[m.name] != m.index]
+        if remaining:
+            return click.echo("{0} members didn't recognized pause state after {1} seconds"
+                              .format(', '.join(remaining), loop_wait))
+    return click.echo('Success: cluster management is {0}'.format(paused and 'paused' or 'resumed'))
+
+
+def toggle_pause(config, cluster_name, paused, wait):
+    dcs = get_dcs(config, cluster_name)
+    cluster = dcs.get_cluster()
     if cluster.is_paused() == paused:
         raise PatroniCtlException('Cluster is {0} paused'.format(paused and 'already' or 'not'))
 
@@ -793,12 +957,16 @@ def toggle_pause(config, cluster_name, paused):
     for member in members:
         try:
             r = request_patroni(member, 'patch', 'config', {'pause': paused or None}, auth_header(config))
-        except Exception:
+        except Exception as err:
+            logging.warning(str(err))
             logging.warning('Member %s is not accessible', member.name)
             continue
 
         if r.status_code == 200:
-            click.echo('Success: cluster management is {0}'.format(paused and 'paused' or 'resumed'))
+            if wait:
+                wait_until_pause_is_applied(dcs, paused, cluster)
+            else:
+                click.echo('Success: cluster management is {0}'.format(paused and 'paused' or 'resumed'))
         else:
             click.echo('Failed: {0} cluster management status code={1}, ({2})'.format(
                        paused and 'pause' or 'resume', r.status_code, r.text))
@@ -808,14 +976,275 @@ def toggle_pause(config, cluster_name, paused):
 
 
 @ctl.command('pause', help='Disable auto failover')
-@click.argument('cluster_name')
+@arg_cluster_name
 @click.pass_obj
-def pause(obj, cluster_name):
+@click.option('--wait', help='Wait until pause is applied on all nodes', is_flag=True)
-    return toggle_pause(obj, cluster_name, True)
+def pause(obj, cluster_name, wait):
+    return toggle_pause(obj, cluster_name, True, wait)
 
 
 @ctl.command('resume', help='Resume auto failover')
-@click.argument('cluster_name')
+@arg_cluster_name
+@click.option('--wait', help='Wait until pause is cleared on all nodes', is_flag=True)
 @click.pass_obj
-def resume(obj, cluster_name):
+def resume(obj, cluster_name, wait):
-    return toggle_pause(obj, cluster_name, False)
+    return toggle_pause(obj, cluster_name, False, wait)
+
+
+@contextmanager
+def temporary_file(contents, suffix='', prefix='tmp'):
+    """Creates a temporary file with specified contents that persists for the context.
+
+    :param contents: binary string that will be written to the file.
+    :param prefix: will be prefixed to the filename.
+    :param suffix: will be appended to the filename.
+    :returns path of the created file.
+    """
+    tmp = tempfile.NamedTemporaryFile(suffix=suffix, prefix=prefix, delete=False)
+    with tmp:
+        tmp.write(contents)
+
+    try:
+        yield tmp.name
+    finally:
+        os.unlink(tmp.name)
+
+
+def show_diff(before_editing, after_editing):
+    """Shows a diff between two strings.
+
+    If the output is to a tty the diff will be colored. Inputs are expected to be unicode strings.
+    """
+    def listify(string):
+        return [l+'\n' for l in string.rstrip('\n').split('\n')]
+
+    unified_diff = difflib.unified_diff(listify(before_editing), listify(after_editing))
+
+    if sys.stdout.isatty():
+        buf = io.StringIO()
+        for line in unified_diff:
+            # Force cast to unicode as difflib on Python 2.7 returns a mix of unicode and str.
+            buf.write(text_type(line))
+        buf.seek(0)
+
+        class opts:
+            side_by_side = False
+            width = 80
+            tab_width = 8
+        cdiff.markup_to_pager(cdiff.PatchStream(buf), opts)
+    else:
+        for line in unified_diff:
+            click.echo(line.rstrip('\n'))
+
+
+def format_config_for_editing(data):
+    """Formats configuration as YAML for human consumption.
+
+    :param data: configuration as nested dictionaries
+    :returns unicode YAML of the configuration"""
+    return yaml.safe_dump(data, default_flow_style=False, encoding=None, allow_unicode=True)
+
+
+def apply_config_changes(before_editing, data, kvpairs):
+    """Applies config changes specified as a list of key-value pairs.
+
+    Keys are interpreted as dotted paths into the configuration data structure. Except for paths beginning with
+    `postgresql.parameters` where rest of the path is used directly to allow for PostgreSQL GUCs containing dots.
+    Values are interpreted as YAML values.
+
+    :param before_editing: human representation before editing
+    :param data: configuration datastructure
+    :param kvpairs: list of strings containing key value pairs separated by =
+    :returns tuple of human readable and parsed datastructure after changes
+    """
+    changed_data = copy.deepcopy(data)
+
+    def set_path_value(config, path, value, prefix=()):
+        # Postgresql GUCs can't be nested, but can contain dots so we re-flatten the structure for this case
+        if prefix == ('postgresql', 'parameters'):
+            path = ['.'.join(path)]
+
+        key = path[0]
+        if len(path) == 1:
+            if value is None:
+                config.pop(key, None)
+            else:
+                config[key] = value
+        else:
+            if not isinstance(config.get(key), dict):
+                config[key] = {}
+            set_path_value(config[key], path[1:], value, prefix + (key,))
+            if config[key] == {}:
+                del config[key]
+
+    for pair in kvpairs:
+        if not pair or "=" not in pair:
+            raise PatroniCtlException("Invalid parameter setting {0}".format(pair))
+        key_path, value = pair.split("=", 1)
+        set_path_value(changed_data, key_path.strip().split("."), yaml.safe_load(value))
+
+    return format_config_for_editing(changed_data), changed_data
+
+
+def apply_yaml_file(data, filename):
+    """Applies changes from a YAML file to configuration
+
+    :param data: configuration datastructure
+    :param filename: name of the YAML file, - is taken to mean standard input
+    :returns tuple of human readable and parsed datastructure after changes
+    """
+    changed_data = copy.deepcopy(data)
+
+    if filename == '-':
+        new_options = yaml.safe_load(sys.stdin)
+    else:
+        with open(filename) as fd:
+            new_options = yaml.safe_load(fd)
+
+    patch_config(changed_data, new_options)
+
+    return format_config_for_editing(changed_data), changed_data
+
+
+def find_executable(executable, path=None):
+    _, ext = os.path.splitext(executable)
+
+    if (sys.platform == 'win32') and (ext != '.exe'):
+        executable = executable + '.exe'
+
+    if os.path.isfile(executable):
+        return executable
+
+    if path is None:
+        path = os.environ.get('PATH', os.defpath)
+
+    for p in path.split(os.pathsep):
+        f = os.path.join(p, executable)
+        if os.path.isfile(f):
+            return f
+
+
+def invoke_editor(before_editing, cluster_name):
+    """Starts editor command to edit configuration in human readable format
+
+    :param before_editing: human representation before editing
+    :returns tuple of human readable and parsed datastructure after changes
+    """
+
+    editor_cmd = os.environ.get('EDITOR')
+    if not editor_cmd:
+        for editor in ('editor', 'vi'):
+            editor_cmd = find_executable(editor)
+            if editor_cmd:
+                logging.debug('Setting fallback editor_cmd=%s', editor)
+                break
+    if not editor_cmd:
+        raise PatroniCtlException('EDITOR environment variable is not set. editor or vi are not available')
+
+    with temporary_file(contents=before_editing.encode('utf-8'),
+                        suffix='.yaml',
+                        prefix='{0}-config-'.format(cluster_name)) as tmpfile:
+        ret = subprocess.call([editor_cmd, tmpfile])
+        if ret:
+            raise PatroniCtlException("Editor exited with return code {0}".format(ret))
+
+        with codecs.open(tmpfile, encoding='utf-8') as fd:
+            after_editing = fd.read()
+
+        return after_editing, yaml.safe_load(after_editing)
+
+
+@ctl.command('edit-config', help="Edit cluster configuration")
+@arg_cluster_name
+@click.option('--quiet', '-q', is_flag=True, help='Do not show changes')
+@click.option('--set', '-s', 'kvpairs', multiple=True,
+              help='Set specific configuration value. Can be specified multiple times')
+@click.option('--pg', '-p', 'pgkvpairs', multiple=True,
+              help='Set specific PostgreSQL parameter value. Shorthand for -s postgresql.parameters. '
+                   'Can be specified multiple times')
+@click.option('--apply', 'apply_filename', help='Apply configuration from file. Use - for stdin.')
+@click.option('--replace', 'replace_filename', help='Apply configuration from file, replacing existing configuration.'
+              ' Use - for stdin.')
+@option_force
+@click.pass_obj
+def edit_config(obj, cluster_name, force, quiet, kvpairs, pgkvpairs, apply_filename, replace_filename):
+    dcs = get_dcs(obj, cluster_name)
+    cluster = dcs.get_cluster()
+
+    before_editing = format_config_for_editing(cluster.config.data)
+
+    after_editing = None  # Serves as a flag if any changes were requested
+    changed_data = cluster.config.data
+
+    if replace_filename:
+        after_editing, changed_data = apply_yaml_file({}, replace_filename)
+
+    if apply_filename:
+        after_editing, changed_data = apply_yaml_file(changed_data, apply_filename)
+
+    if kvpairs or pgkvpairs:
+        all_pairs = list(kvpairs) + ['postgresql.parameters.'+v.lstrip() for v in pgkvpairs]
+        after_editing, changed_data = apply_config_changes(before_editing, changed_data, all_pairs)
+
+    # If no changes were specified on the command line invoke editor
+    if after_editing is None:
+        after_editing, changed_data = invoke_editor(before_editing, cluster_name)
+
+    if cluster.config.data == changed_data:
+        if not quiet:
+            click.echo("Not changed")
+        return
+
+    if not quiet:
+        show_diff(before_editing, after_editing)
+
+    if (apply_filename == '-' or replace_filename == '-') and not force:
+        click.echo("Use --force option to apply changes")
+        return
+
+    if force or click.confirm('Apply these changes?'):
+        if not dcs.set_config_value(json.dumps(changed_data), cluster.config.index):
+            raise PatroniCtlException("Config modification aborted due to concurrent changes")
+        click.echo("Configuration changed")
+
+
+@ctl.command('show-config', help="Show cluster configuration")
+@arg_cluster_name
+@click.pass_obj
+def show_config(obj, cluster_name):
+    cluster = get_dcs(obj, cluster_name).get_cluster()
+
+    click.echo(format_config_for_editing(cluster.config.data))
+
+
+@ctl.command('version', help='Output version of patronictl command or a running Patroni instance')
+@click.argument('cluster_name', required=False)
+@click.argument('member_names', nargs=-1)
+@click.pass_obj
+def version(obj, cluster_name, member_names):
+    click.echo("patronictl version {0}".format(__version__))
+
+    if not cluster_name:
+        return
+
+    click.echo("")
+    cluster = get_dcs(obj, cluster_name).get_cluster()
+    for m in cluster.members:
+        if m.api_url:
+            if not member_names or m.name in member_names:
+                try:
+                    response = request_patroni(m, 'get', 'patroni')
+                    data = response.json()
+                    version = data.get('patroni', {}).get('version')
+                    pg_version = data.get('server_version')
+                    pg_version_str = " PostgreSQL {0}".format(format_pg_version(pg_version)) if pg_version else ""
+                    click.echo("{0}: Patroni {1}{2}".format(m.name, version, pg_version_str))
+                except Exception as e:
+                    click.echo("{0}: failed to get version: {1}".format(m.name, e))
+
+
+def format_pg_version(version):
+    if version < 100000:
+        return "{0}.{1}.{2}".format(version // 10000, version // 100 % 100, version % 100)
+    else:
+        return "{0}.{1}".format(version // 10000, version % 100)

patroni/dcs/__init__.py

@@ -3,17 +3,41 @@ import dateutil
 import importlib
 import inspect
 import json
+import logging
 import os
 import pkgutil
+import re
 import six
 import sys
+import time
 
-from collections import namedtuple
+from collections import defaultdict, namedtuple
+from copy import deepcopy
 from patroni.exceptions import PatroniException
+from patroni.utils import parse_bool, uri
 from random import randint
 from six.moves.urllib_parse import urlparse, urlunparse, parse_qsl
 from threading import Event, Lock
 
+slot_name_re = re.compile('^[a-z0-9_]{1,63}$')
+logger = logging.getLogger(__name__)
+
+
+def slot_name_from_member_name(member_name):
+    """Translate member name to valid PostgreSQL slot name.
+
+    PostgreSQL replication slot names must be valid PostgreSQL names. This function maps the wider space of
+    member names to valid PostgreSQL names. Names are lowercased, dashes and periods common in hostnames
+    are replaced with underscores, other characters are encoded as their unicode codepoint. Name is truncated
+    to 64 characters. Multiple different member names may map to a single slot name."""
+
+    def replace_char(match):
+        c = match.group(0)
+        return '_' if c in '-.' else "u{:04d}".format(ord(c))
+
+    slot_name = re.sub('[^a-z0-9_]', replace_char, member_name.lower())
+    return slot_name[0:63]
+
 
 def parse_connection_string(value):
     """Original Governor stores connection strings for each cluster members if a following format:
@@ -51,18 +75,22 @@ def dcs_modules():
 def get_dcs(config):
     available_implementations = set()
     for module_name in dcs_modules():
-        module = importlib.import_module(module_name)
+        try:
-        for name in filter(lambda name: not name.startswith('__'), dir(module)):  # iterate through module content
+            module = importlib.import_module(module_name)
-            value = getattr(module, name)
+            for name in filter(lambda name: not name.startswith('__'), dir(module)):  # iterate through module content
-            name = name.lower()
+                item = getattr(module, name)
-            # try to find implementation of AbstractDCS interface, class name must match with module_name
+                name = name.lower()
-            if inspect.isclass(value) and issubclass(value, AbstractDCS) and __package__ + '.' + name == module_name:
+                # try to find implementation of AbstractDCS interface, class name must match with module_name
-                available_implementations.add(name)
+                if inspect.isclass(item) and issubclass(item, AbstractDCS) and __package__ + '.' + name == module_name:
-                if name in config:  # which has configuration section in the config file
+                    available_implementations.add(name)
-                    # propagate some parameters
+                    if name in config:  # which has configuration section in the config file
-                    config[name].update({p: config[p] for p in ('namespace', 'name', 'scope', 'loop_wait',
+                        # propagate some parameters
-                                         'patronictl', 'ttl', 'retry_timeout') if p in config})
+                        config[name].update({p: config[p] for p in ('namespace', 'name', 'scope', 'loop_wait',
-                    return value(config[name])
+                                             'patronictl', 'ttl', 'retry_timeout') if p in config})
+                        return item(config[name])
+        except ImportError:
+            if not config.get('patronictl'):
+                logger.info('Failed to import %s', module_name)
     raise PatroniException("""Can not find suitable configuration of distributed configuration store
 Available implementations: """ + ', '.join(available_implementations))
 
@@ -100,12 +128,26 @@ class Member(namedtuple('Member', 'index,name,session,data')):
 
     @property
     def conn_url(self):
-        return self.data.get('conn_url')
+        conn_url = self.data.get('conn_url')
+        conn_kwargs = self.data.get('conn_kwargs')
+        if conn_url:
+            return conn_url
+
+        if conn_kwargs:
+            conn_url = uri('postgresql', (conn_kwargs.get('host'), conn_kwargs.get('port', 5432)))
+            self.data['conn_url'] = conn_url
+            return conn_url
 
     def conn_kwargs(self, auth=None):
+        defaults = {
+            "host": "",
+            "port": "",
+            "database": ""
+        }
         ret = self.data.get('conn_kwargs')
         if ret:
-            ret = ret.copy()
+            defaults.update(ret)
+            ret = defaults
         else:
             r = urlparse(self.conn_url)
             ret = {
@@ -151,6 +193,26 @@ class Member(namedtuple('Member', 'index,name,session,data')):
         return self.state == 'running'
 
 
+class RemoteMember(Member):
+    """ Represents a remote master for a standby cluster
+    """
+    def __new__(cls, name, data):
+        return super(RemoteMember, cls).__new__(cls, None, name, None, data)
+
+    @staticmethod
+    def allowed_keys():
+        return ('primary_slot_name',
+                'create_replica_methods',
+                'restore_command',
+                'archive_cleanup_command',
+                'recovery_min_apply_delay',
+                'no_replication_slot')
+
+    def __getattr__(self, name):
+        if name in RemoteMember.allowed_keys():
+            return self.data.get(name)
+
+
 class Leader(namedtuple('Leader', 'index,session,member')):
 
     """Immutable object (namedtuple) which represents leader key.
@@ -170,40 +232,62 @@ class Leader(namedtuple('Leader', 'index,session,member')):
     def conn_url(self):
         return self.member.conn_url
 
+    @property
+    def timeline(self):
+        return self.member.data.get('timeline')
+
+    @property
+    def checkpoint_after_promote(self):
+        """
+        >>> Leader(1, '', Member.from_node(1, '', '', '{"version":"z"}')).checkpoint_after_promote
+        """
+        version = self.member.data.get('version')
+        if version:
+            try:
+                # 1.5.6 is the last version which doesn't expose checkpoint_after_promote: false
+                if tuple(map(int, version.split('.'))) > (1, 5, 6):
+                    return self.member.data['role'] == 'master' and 'checkpoint_after_promote' not in self.member.data
+            except Exception:
+                logger.debug('Failed to parse Patroni version %s', version)
+
 
 class Failover(namedtuple('Failover', 'index,leader,candidate,scheduled_at')):
 
     """
     >>> 'Failover' in str(Failover.from_node(1, '{"leader": "cluster_leader"}'))
     True
+    >>> 'Failover' in str(Failover.from_node(1, {"leader": "cluster_leader"}))
+    True
     >>> 'Failover' in str(Failover.from_node(1, '{"leader": "cluster_leader", "member": "cluster_candidate"}'))
     True
     >>> Failover.from_node(1, 'null') is None
-    True
+    False
     >>> n = '{"leader": "cluster_leader", "member": "cluster_candidate", "scheduled_at": "2016-01-14T10:09:57.1394Z"}'
     >>> 'tzinfo=' in str(Failover.from_node(1, n))
     True
     >>> Failover.from_node(1, None) is None
-    True
+    False
     >>> Failover.from_node(1, '{}') is None
-    True
+    False
     >>> 'abc' in Failover.from_node(1, 'abc:def')
     True
     """
     @staticmethod
     def from_node(index, value):
-        if not value:
+        if isinstance(value, dict):
-            return None
+            data = value
-
+        elif value:
-        try:
+            try:
-            data = json.loads(value)
+                data = json.loads(value)
-            if not data:
+                if not isinstance(data, dict):
-                return None
+                    data = {}
-        except ValueError:
+            except ValueError:
-            t = [a.strip() for a in value.split(':')]
+                t = [a.strip() for a in value.split(':')]
-            leader = t[0]
+                leader = t[0]
-            candidate = t[1] if len(t) > 1 else None
+                candidate = t[1] if len(t) > 1 else None
-            return Failover(index, leader, candidate, None) if leader or candidate else None
+                return Failover(index, leader, candidate, None) if leader or candidate else None
+        else:
+            data = {}
 
         if data.get('scheduled_at'):
             data['scheduled_at'] = dateutil.parser.parse(data['scheduled_at'])
@@ -220,14 +304,24 @@ class ClusterConfig(namedtuple('ClusterConfig', 'index,data,modify_index')):
     def from_node(index, data, modify_index=None):
         """
         >>> ClusterConfig.from_node(1, '{') is None
-        True
+        False
         """
 
         try:
             data = json.loads(data)
         except (TypeError, ValueError):
-            return None
+            data = None
-        return ClusterConfig(index, data, modify_index or index)
+            modify_index = 0
+        if not isinstance(data, dict):
+            data = {}
+        return ClusterConfig(index, data, index if modify_index is None else modify_index)
+
+    @property
+    def permanent_slots(self):
+        return isinstance(self.data, dict) and (
+                self.data.get('permanent_replication_slots') or
+                self.data.get('permanent_slots') or self.data.get('slots')
+        ) or {}
 
 
 class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
@@ -251,8 +345,12 @@ class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
         True
         >>> SyncState.from_node(1, '{"leader": "leader"}').leader == "leader"
         True
+        >>> SyncState.from_node(1, {"leader": "leader"}).leader == "leader"
+        True
         """
-        if value:
+        if isinstance(value, dict):
+            data = value
+        elif value:
             try:
                 data = json.loads(value)
                 if not isinstance(data, dict):
@@ -282,7 +380,26 @@ class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
         return name is not None and name in (self.leader, self.sync_standby)
 
 
-class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operation,members,failover,sync')):
+class TimelineHistory(namedtuple('TimelineHistory', 'index,value,lines')):
+    """Object representing timeline history file"""
+
+    @staticmethod
+    def from_node(index, value):
+        """
+        >>> h = TimelineHistory.from_node(1, 2)
+        >>> h.lines
+        []
+        """
+        try:
+            lines = json.loads(value)
+        except (TypeError, ValueError):
+            lines = None
+        if not isinstance(lines, list):
+            lines = []
+        return TimelineHistory(index, value, lines)
+
+
+class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operation,members,failover,sync,history')):
 
     """Immutable object (namedtuple) which represents PostgreSQL cluster.
     Consists of the following fields:
@@ -294,6 +411,7 @@ class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operat
     :param members: list of Member object, all PostgreSQL cluster members including leader
     :param failover: reference to `Failover` object
     :param sync: reference to `SyncState` object, last observed synchronous replication state.
+    :param history: reference to `TimelineHistory` object
     """
 
     def is_unlocked(self):
@@ -310,8 +428,93 @@ class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operat
         candidates = [m for m in self.members if m.clonefrom and m.is_running and m.name not in exclude]
         return candidates[randint(0, len(candidates) - 1)] if candidates else self.leader
 
+    def check_mode(self, mode):
+        return bool(self.config and parse_bool(self.config.data.get(mode)))
+
     def is_paused(self):
-        return self.config and self.config.data.get('pause', False) or False
+        return self.check_mode('pause')
+
+    def is_synchronous_mode(self):
+        return self.check_mode('synchronous_mode')
+
+    def get_replication_slots(self, name, role):
+        # if the replicatefrom tag is set on the member - we should not create the replication slot for it on
+        # the current master, because that member would replicate from elsewhere. We still create the slot if
+        # the replicatefrom destination member is currently not a member of the cluster (fallback to the
+        # master), or if replicatefrom destination member happens to be the current master
+        if role in ('master', 'standby_leader'):
+            slot_members = [m.name for m in self.members if m.name != name and
+                            (m.replicatefrom is None or m.replicatefrom == name or
+                             not self.has_member(m.replicatefrom))]
+            permanent_slots = (self.config and self.config.permanent_slots or {}).copy()
+        else:
+            # only manage slots for replicas that replicate from this one, except for the leader among them
+            slot_members = [m.name for m in self.members if m.replicatefrom == name and m.name != self.leader.name]
+            permanent_slots = {}
+
+        slots = {slot_name_from_member_name(name): {'type': 'physical'} for name in slot_members}
+
+        if len(slots) < len(slot_members):
+            # Find which names are conflicting for a nicer error message
+            slot_conflicts = defaultdict(list)
+            for name in slot_members:
+                slot_conflicts[slot_name_from_member_name(name)].append(name)
+            logger.error("Following cluster members share a replication slot name: %s",
+                         "; ".join("{} map to {}".format(", ".join(v), k)
+                                   for k, v in slot_conflicts.items() if len(v) > 1))
+
+        # "merge" replication slots for members with permanent_replication_slots
+        for name, value in permanent_slots.items():
+            if not slot_name_re.match(name):
+                logger.error("Invalid permanent replication slot name '%s'", name)
+                logger.error("Slot name may only contain lower case letters, numbers, and the underscore chars")
+                continue
+
+            if name in slots:
+                logger.error("Permanent replication slot {'%s': %s} is conflicting with" +
+                             " physical replication slot for cluster member", name, value)
+                continue
+
+            value = deepcopy(value)
+            if not value:
+                value = {'type': 'physical'}
+
+            if isinstance(value, dict):
+                if 'type' not in value:
+                    value['type'] = 'logical' if value.get('database') and value.get('plugin') else 'physical'
+
+                if value['type'] == 'physical' or value['type'] == 'logical' \
+                        and value.get('database') and value.get('plugin'):
+                    slots[name] = value
+                    continue
+
+            logger.error("Bad value for slot '%s' in permanent_slots: %s", name, permanent_slots[name])
+
+        return slots
+
+    def has_permanent_logical_slots(self, name):
+        slots = self.get_replication_slots(name, 'master').values()
+        return any(v for v in slots if v.get("type") == "logical")
+
+    @property
+    def timeline(self):
+        """
+        >>> Cluster(0, 0, 0, 0, 0, 0, 0, 0).timeline
+        0
+        >>> Cluster(0, 0, 0, 0, 0, 0, 0, TimelineHistory.from_node(1, '[]')).timeline
+        1
+        >>> Cluster(0, 0, 0, 0, 0, 0, 0, TimelineHistory.from_node(1, '[["a"]]')).timeline
+        0
+        """
+        if self.history:
+            if self.history.lines:
+                try:
+                    return int(self.history.lines[-1][0]) + 1
+                except Exception:
+                    logger.error('Failed to parse cluster history from DCS: %s', self.history.lines)
+            elif self.history.value == '[]':
+                return 1
+        return 0
 
 
 @six.add_metaclass(abc.ABCMeta)
@@ -321,6 +524,7 @@ class AbstractDCS(object):
     _CONFIG = 'config'
     _LEADER = 'leader'
     _FAILOVER = 'failover'
+    _HISTORY = 'history'
     _MEMBERS = 'members/'
     _OPTIME = 'optime'
     _LEADER_OPTIME = _OPTIME + '/' + _LEADER
@@ -332,12 +536,12 @@ class AbstractDCS(object):
             i.e.: `zookeeper` for zookeeper, `etcd` for etcd, etc...
         """
         self._name = config['name']
-        self._namespace = '/{0}'.format(config.get('namespace', '/service/').strip('/'))
+        self._base_path = re.sub('/+', '/', '/'.join(['', config.get('namespace', 'service'), config['scope']]))
-        self._base_path = '/'.join([self._namespace, config['scope']])
         self._set_loop_wait(config.get('loop_wait', 10))
 
         self._ctl = bool(config.get('patronictl', False))
         self._cluster = None
+        self._cluster_valid_till = 0
         self._cluster_thread_lock = Lock()
         self._last_leader_operation = ''
         self.event = Event()
@@ -369,6 +573,10 @@ class AbstractDCS(object):
     def failover_path(self):
         return self.client_path(self._FAILOVER)
 
+    @property
+    def history_path(self):
+        return self.client_path(self._HISTORY)
+
     @property
     def leader_optime_path(self):
         return self.client_path(self._LEADER_OPTIME)
@@ -381,6 +589,10 @@ class AbstractDCS(object):
     def set_ttl(self, ttl):
         """Set the new ttl value for leader key"""
 
+    @abc.abstractmethod
+    def ttl(self):
+        """Get new ttl value"""
+
     @abc.abstractmethod
     def set_retry_timeout(self, retry_timeout):
         """Set the new value for retry_timeout"""
@@ -408,22 +620,26 @@ class AbstractDCS(object):
            instance would be demoted."""
 
     def get_cluster(self):
+        try:
+            cluster = self._load_cluster()
+        except Exception:
+            self.reset_cluster()
+            raise
+
         with self._cluster_thread_lock:
-            try:
+            self._cluster = cluster
-                self._load_cluster()
+            self._cluster_valid_till = time.time() + self.ttl
-            except:
+            return cluster
-                self._cluster = None
-                raise
-            return self._cluster
 
     @property
     def cluster(self):
         with self._cluster_thread_lock:
-            return self._cluster
+            return self._cluster if self._cluster_valid_till > time.time() else None
 
     def reset_cluster(self):
         with self._cluster_thread_lock:
             self._cluster = None
+            self._cluster_valid_till = 0
 
     @abc.abstractmethod
     def _write_leader_optime(self, last_operation):
@@ -436,7 +652,7 @@ class AbstractDCS(object):
             self._last_leader_operation = last_operation
 
     @abc.abstractmethod
-    def update_leader(self):
+    def _update_leader(self):
         """Update leader key (or session) ttl
 
         :returns: `!True` if leader key (or session) has been updated successfully.
@@ -445,6 +661,18 @@ class AbstractDCS(object):
         You have to use CAS (Compare And Swap) operation in order to update leader key,
         for example for etcd `prevValue` parameter must be used."""
 
+    def update_leader(self, last_operation, access_is_restricted=False):
+        """Update leader key (or session) ttl and optime/leader
+
+        :param last_operation: absolute xlog location in bytes
+        :returns: `!True` if leader key (or session) has been updated successfully.
+            If not, `!False` must be returned and current instance would be demoted."""
+
+        ret = self._update_leader()
+        if ret and last_operation:
+            self.write_leader_optime(last_operation)
+        return ret
+
     @abc.abstractmethod
     def attempt_to_acquire_leader(self, permanent=False):
         """Attempt to acquire leader lock
@@ -470,7 +698,6 @@ class AbstractDCS(object):
 
         if scheduled_at:
             failover_value['scheduled_at'] = scheduled_at.isoformat()
-
         return self.set_failover_value(json.dumps(failover_value, separators=(',', ':')), index)
 
     @abc.abstractmethod
@@ -478,12 +705,12 @@ class AbstractDCS(object):
         """Create or update `/config` key"""
 
     @abc.abstractmethod
-    def touch_member(self, data, ttl=None, permanent=False):
+    def touch_member(self, data, permanent=False):
         """Update member key in DCS.
         This method should create or update key with the name = '/members/' + `~self._name`
         and value = data in a given DCS.
 
-        :param data: json serialized information about instance (including connection strings)
+        :param data: information about instance (including connection strings)
         :param ttl: ttl for member key, optional parameter. If it is None `~self.member_ttl will be used`
         :param permanent: if set to `!True`, the member key will never expire.
          Used in patronictl for the external master.
@@ -520,8 +747,18 @@ class AbstractDCS(object):
     def delete_cluster(self):
         """Delete cluster from DCS"""
 
+    @staticmethod
+    def sync_state(leader, sync_standby):
+        """Build sync_state dict"""
+        return {'leader': leader, 'sync_standby': sync_standby}
+
     def write_sync_state(self, leader, sync_standby, index=None):
-        return self.set_sync_state_value(json.dumps({'leader': leader, 'sync_standby': sync_standby}), index=index)
+        sync_value = self.sync_state(leader, sync_standby)
+        return self.set_sync_state_value(json.dumps(sync_value, separators=(',', ':')), index)
+
+    @abc.abstractmethod
+    def set_history_value(self, value):
+        """"""
 
     @abc.abstractmethod
     def set_sync_state_value(self, value, index=None):

patroni/dcs/consul.py

@@ -1,16 +1,19 @@
 from __future__ import absolute_import
+import json
 import logging
 import os
+import re
 import socket
+import ssl
 import time
 import urllib3
 
 from consul import ConsulException, NotFound, base
-from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState
+from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
 from patroni.exceptions import DCSError
-from patroni.utils import Retry, RetryFailedError
+from patroni.utils import deep_compare, parse_bool, Retry, RetryFailedError, split_host_port, uri
 from urllib3.exceptions import HTTPError
-from six.moves.urllib.parse import urlencode
+from six.moves.urllib.parse import urlencode, urlparse, quote
 from six.moves.http_client import HTTPException
 
 logger = logging.getLogger(__name__)
@@ -24,21 +27,43 @@ class ConsulInternalError(ConsulException):
     """An internal Consul server error occurred"""
 
 
+class InvalidSessionTTL(ConsulException):
+    """Session TTL is too small or too big"""
+
+
+class InvalidSession(ConsulException):
+    """invalid session"""
+
+
 class HTTPClient(object):
 
-    def __init__(self, host='127.0.0.1', port=8500, scheme='http', verify=True, timeout=10):
+    def __init__(self, host='127.0.0.1', port=8500, token=None, scheme='http', verify=True, cert=None, ca_cert=None):
-        self.host = host
+        self.token = token
-        self.port = port
+        self._read_timeout = 10
-        self.scheme = scheme
+        self.base_uri = uri(scheme, (host, port))
-        self.verify = verify
+        kwargs = {}
-        self.set_read_timeout(timeout)
+        if cert:
-        self.base_uri = '{0}://{1}:{2}'.format(self.scheme, self.host, self.port)
+            if isinstance(cert, tuple):
-        self.http = urllib3.PoolManager(num_pools=10)
+                # Key and cert are separate
+                kwargs['cert_file'] = cert[0]
+                kwargs['key_file'] = cert[1]
+            else:
+                # combined certificate
+                kwargs['cert_file'] = cert
+        if ca_cert:
+            kwargs['ca_certs'] = ca_cert
+        if verify or ca_cert:
+            kwargs['cert_reqs'] = ssl.CERT_REQUIRED
+        self.http = urllib3.PoolManager(num_pools=10, **kwargs)
         self._ttl = None
 
     def set_read_timeout(self, timeout):
         self._read_timeout = timeout/3.0
 
+    @property
+    def ttl(self):
+        return self._ttl
+
     def set_ttl(self, ttl):
         ret = self._ttl != ttl
         self._ttl = ttl
@@ -48,7 +73,13 @@ class HTTPClient(object):
     def response(response):
         data = response.data.decode('utf-8')
         if response.status == 500:
-            raise ConsulInternalError('{0} {1}'.format(response.status, data))
+            msg = '{0} {1}'.format(response.status, data)
+            if data.startswith('Invalid Session TTL'):
+                raise InvalidSessionTTL(msg)
+            elif data.startswith('invalid session'):
+                raise InvalidSession(msg)
+            else:
+                raise ConsulInternalError(msg)
         return base.Response(response.status, response.headers, data)
 
     def uri(self, path, params=None):
@@ -67,20 +98,47 @@ class HTTPClient(object):
                     data = '{' + ttl + '}'
                 else:
                     data = data[:-1] + ', ' + ttl + '}'
+            if isinstance(params, list):  # starting from v1.1.0 python-consul switched from `dict` to `list` for params
+                params = {k: v for k, v in params}
             kwargs = {'retries': 0, 'preload_content': False, 'body': data}
             if method == 'get' and isinstance(params, dict) and 'index' in params:
-                kwargs['timeout'] = (float(params['wait'][:-1]) if 'wait' in params else 300) + 1
+                timeout = float(params['wait'][:-1]) if 'wait' in params else 300
+                # According to the documentation a small random amount of additional wait time is added to the
+                # supplied maximum wait time to spread out the wake up time of any concurrent requests. This adds
+                # up to wait / 16 additional time to the maximum duration. Since our goal is actually getting a
+                # response rather read timeout we will add to the timeout a sligtly bigger value.
+                kwargs['timeout'] = timeout + max(timeout/15.0, 1)
             else:
                 kwargs['timeout'] = self._read_timeout
+            token = params.pop('token', self.token) if isinstance(params, dict) else self.token
+            if token:
+                kwargs['headers'] = {'X-Consul-Token': token}
             return callback(self.response(self.http.request(method.upper(), self.uri(path, params), **kwargs)))
         return wrapper
 
 
 class ConsulClient(base.Consul):
 
-    @staticmethod
+    def __init__(self, *args, **kwargs):
-    def connect(host, port, scheme, verify=True):
+        self._cert = kwargs.pop('cert', None)
-        return HTTPClient(host, port, scheme, verify)
+        self._ca_cert = kwargs.pop('ca_cert', None)
+        self.token = kwargs.get('token')
+        super(ConsulClient, self).__init__(*args, **kwargs)
+
+    def connect(self, *args, **kwargs):
+        kwargs.update(dict(zip(['host', 'port', 'scheme', 'verify'], args)))
+        if self._cert:
+            kwargs['cert'] = self._cert
+        if self._ca_cert:
+            kwargs['ca_cert'] = self._ca_cert
+        if self.token:
+            kwargs['token'] = self.token
+        return HTTPClient(**kwargs)
+
+    def reload_config(self, config):
+        self.http.token = self.token = config.get('token')
+        self.consistency = config.get('consistency', 'default')
+        self.dc = config.get('dc')
 
 
 def catch_consul_errors(func):
@@ -92,6 +150,31 @@ def catch_consul_errors(func):
     return wrapper
 
 
+def force_if_last_failed(func):
+    def wrapper(*args, **kwargs):
+        if wrapper.last_result is False:
+            kwargs['force'] = True
+        wrapper.last_result = func(*args, **kwargs)
+        return wrapper.last_result
+
+    wrapper.last_result = None
+    return wrapper
+
+
+def service_name_from_scope_name(scope_name):
+    """Translate scope name to service name which can be used in dns.
+
+    230 = 253 - len('replica.') - len('.service.consul')
+    """
+
+    def replace_char(match):
+        c = match.group(0)
+        return '-' if c in '. _' else "u{:04d}".format(ord(c))
+
+    service_name = re.sub(r'[^a-z0-9\-]', replace_char, scope_name.lower())
+    return service_name[0:230]
+
+
 class Consul(AbstractDCS):
 
     def __init__(self, config):
@@ -103,12 +186,43 @@ class Consul(AbstractDCS):
                             retry_exceptions=(ConsulInternalError, HTTPException,
                                               HTTPError, socket.error, socket.timeout))
 
-        self._my_member_data = None
+        kwargs = {}
-        host, port = config.get('host', '127.0.0.1:8500').split(':')
+        if 'url' in config:
-        self._client = ConsulClient(host=host, port=port)
+            r = urlparse(config['url'])
+            config.update({'scheme': r.scheme, 'host': r.hostname, 'port': r.port or 8500})
+        elif 'host' in config:
+            host, port = split_host_port(config.get('host', '127.0.0.1:8500'), 8500)
+            config['host'] = host
+            if 'port' not in config:
+                config['port'] = int(port)
+
+        if config.get('cacert'):
+            config['ca_cert'] = config.pop('cacert')
+
+        if config.get('key') and config.get('cert'):
+            config['cert'] = (config['cert'], config['key'])
+
+        config_keys = ('host', 'port', 'token', 'scheme', 'cert', 'ca_cert', 'dc', 'consistency')
+        kwargs = {p: config.get(p) for p in config_keys if config.get(p)}
+
+        verify = config.get('verify')
+        if not isinstance(verify, bool):
+            verify = parse_bool(verify)
+        if isinstance(verify, bool):
+            kwargs['verify'] = verify
+
+        self._client = ConsulClient(**kwargs)
         self.set_retry_timeout(config['retry_timeout'])
         self.set_ttl(config.get('ttl') or 30)
         self._last_session_refresh = 0
+        self.__session_checks = config.get('checks')
+        self._register_service = config.get('register_service', False)
+        if self._register_service:
+            self._service_name = service_name_from_scope_name(self._scope)
+            if self._scope != self._service_name:
+                logger.warning('Using %s as consul service name instead of scope name %s', self._service_name,
+                               self._scope)
+        self._service_check_interval = config.get('service_check_interval', '5s')
         if not self._ctl:
             self.create_session()
 
@@ -123,15 +237,32 @@ class Consul(AbstractDCS):
                 logger.info('waiting on consul')
                 time.sleep(5)
 
+    def reload_config(self, config):
+        super(Consul, self).reload_config(config)
+        self._client.reload_config(config.get('consul', {}))
+
     def set_ttl(self, ttl):
         if self._client.http.set_ttl(ttl/2.0):  # Consul multiplies the TTL by 2x
             self._session = None
             self.__do_not_watch = True
 
+    @property
+    def ttl(self):
+        return self._client.http.ttl
+
     def set_retry_timeout(self, retry_timeout):
         self._retry.deadline = retry_timeout
         self._client.http.set_read_timeout(retry_timeout)
 
+    def adjust_ttl(self):
+        try:
+            settings = self._client.agent.self()
+            min_ttl = (settings['Config']['SessionTTLMin'] or 10000000000)/1000000000.0
+            logger.warning('Changing Session TTL from %s to %s', self._client.http.ttl, min_ttl)
+            self._client.http.set_ttl(min_ttl)
+        except Exception:
+            logger.exception('adjust_ttl')
+
     def _do_refresh_session(self):
         """:returns: `!True` if it had to create new session"""
         if self._session and self._last_session_refresh + self._loop_wait > time.time():
@@ -144,8 +275,15 @@ class Consul(AbstractDCS):
                 self._session = None
         ret = not self._session
         if ret:
-            self._session = self._client.session.create(name=self._scope + '-' + self._name,
+            try:
-                                                        lock_delay=0.001, behavior='delete')
+                self._session = self._client.session.create(name=self._scope + '-' + self._name,
+                                                            checks=self.__session_checks,
+                                                            lock_delay=0.001, behavior='delete')
+            except InvalidSessionTTL:
+                logger.exception('session.create')
+                self.adjust_ttl()
+                raise
+
         self._last_session_refresh = time.time()
         return ret
 
@@ -174,7 +312,7 @@ class Consul(AbstractDCS):
             nodes = {}
             for node in results:
                 node['Value'] = (node['Value'] or b'').decode('utf-8')
-                nodes[os.path.relpath(node['Key'], path)] = node
+                nodes[node['Key'][len(path):].lstrip('/')] = node
 
             # get initialize flag
             initialize = nodes.get(self._INITIALIZE)
@@ -184,6 +322,10 @@ class Consul(AbstractDCS):
             config = nodes.get(self._CONFIG)
             config = config and ClusterConfig.from_node(config['ModifyIndex'], config['Value'])
 
+            # get timeline history
+            history = nodes.get(self._HISTORY)
+            history = history and TimelineHistory.from_node(history['ModifyIndex'], history['Value'])
+
             # get last leader operation
             last_leader_operation = nodes.get(self._LEADER_OPTIME)
             last_leader_operation = 0 if last_leader_operation is None else int(last_leader_operation['Value'])
@@ -213,46 +355,111 @@ class Consul(AbstractDCS):
             sync = nodes.get(self._SYNC)
             sync = SyncState.from_node(sync and sync['ModifyIndex'], sync and sync['Value'])
 
-            self._cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync)
+            return Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
         except NotFound:
-            self._cluster = Cluster(None, None, None, None, [], None, None)
+            return Cluster(None, None, None, None, [], None, None, None)
-        except:
+        except Exception:
             logger.exception('get_cluster')
             raise ConsulError('Consul is not responding properly')
 
-    def touch_member(self, data, **kwargs):
+    @catch_consul_errors
+    def touch_member(self, data, permanent=False):
         cluster = self.cluster
         member = cluster and cluster.get_member(self._name, fallback_to_leader=False)
-        create_member = self.refresh_session()
+        create_member = not permanent and self.refresh_session()
 
         if member and (create_member or member.session != self._session):
-            try:
+            self._client.kv.delete(self.member_path)
-                self._client.kv.delete(self.member_path)
+            create_member = True
-                create_member = True
-            except Exception:
-                return False
 
-        if not create_member and member and data == self._my_member_data:
+        if not create_member and member and deep_compare(data, member.data):
             return True
 
         try:
-            args = {} if kwargs.get('permanent', False) else {'acquire': self._session}
+            args = {} if permanent else {'acquire': self._session}
-            self._client.kv.put(self.member_path, data, **args)
+            self._client.kv.put(self.member_path, json.dumps(data, separators=(',', ':')), **args)
-            self._my_member_data = data
+            if self._register_service:
+                self.update_service(not create_member and member and member.data or {}, data)
             return True
+        except InvalidSession:
+            self._session = None
+            logger.error('Our session disappeared from Consul, can not "touch_member"')
         except Exception:
             logger.exception('touch_member')
         return False
 
     @catch_consul_errors
+    def register_service(self, service_name, **kwargs):
+        logger.info('Register service %s, params %s', service_name, kwargs)
+        return self._client.agent.service.register(service_name, **kwargs)
+
+    @catch_consul_errors
+    def deregister_service(self, service_id):
+        logger.info('Deregister service %s', service_id)
+        # service_id can contain special characters, but is used as part of uri in deregister request
+        service_id = quote(service_id)
+        return self._client.agent.service.deregister(service_id)
+
+    def _update_service(self, data):
+        service_name = self._service_name
+        role = data['role'].replace('_', '-')
+        state = data['state']
+        api_parts = urlparse(data['api_url'])
+        api_parts = api_parts._replace(path='/{0}'.format(role))
+        conn_parts = urlparse(data['conn_url'])
+        check = base.Check.http(api_parts.geturl(), self._service_check_interval,
+                                deregister='{0}s'.format(self._client.http.ttl * 10))
+        params = {
+            'service_id': '{0}/{1}'.format(self._scope, self._name),
+            'address': conn_parts.hostname,
+            'port': conn_parts.port,
+            'check': check,
+            'tags': [role]
+        }
+
+        if state == 'stopped':
+            return self.deregister_service(params['service_id'])
+
+        if role in ['master', 'replica', 'standby-leader']:
+            if state != 'running':
+                return
+            return self.register_service(service_name, **params)
+
+        logger.warning('Could not register service: unknown role type %s', role)
+
+    @force_if_last_failed
+    def update_service(self, old_data, new_data, force=False):
+        update = False
+
+        for key in ['role', 'api_url', 'conn_url', 'state']:
+            if key not in new_data:
+                logger.warning('Could not register service: not enough params in member data')
+                return
+            if old_data.get(key) != new_data[key]:
+                update = True
+
+        if force or update:
+            return self._update_service(new_data)
+
+    @catch_consul_errors
+    def _do_attempt_to_acquire_leader(self, permanent):
+        try:
+            kwargs = {} if permanent else {'acquire': self._session}
+            return self.retry(self._client.kv.put, self.leader_path, self._name, **kwargs)
+        except InvalidSession:
+            self._session = None
+            logger.error('Our session disappeared from Consul. Will try to get a new one and retry attempt')
+            self.refresh_session()
+            return self.retry(self._client.kv.put, self.leader_path, self._name, acquire=self._session)
+
     def attempt_to_acquire_leader(self, permanent=False):
         if not self._session and not permanent:
             self.refresh_session()
 
-        args = {} if permanent else {'acquire': self._session}
+        ret = self._do_attempt_to_acquire_leader(permanent)
-        ret = self.retry(self._client.kv.put, self.leader_path, self._name, **args)
         if not ret:
             logger.info('Could not take out TTL lock')
+
         return ret
 
     def take_leader(self):
@@ -271,7 +478,7 @@ class Consul(AbstractDCS):
         return self._client.kv.put(self.leader_optime_path, last_operation)
 
     @catch_consul_errors
-    def update_leader(self):
+    def _update_leader(self):
         if self._session:
             self.retry(self._client.session.renew, self._session)
             self._last_session_refresh = time.time()
@@ -290,6 +497,10 @@ class Consul(AbstractDCS):
     def delete_cluster(self):
         return self.retry(self._client.kv.delete, self.client_path(''), recurse=True)
 
+    @catch_consul_errors
+    def set_history_value(self, value):
+        return self._client.kv.put(self.history_path, value)
+
     @catch_consul_errors
     def delete_leader(self):
         cluster = self.cluster
@@ -298,13 +509,14 @@ class Consul(AbstractDCS):
 
     @catch_consul_errors
     def set_sync_state_value(self, value, index=None):
-        return self._client.kv.put(self.sync_path, value, cas=index)
+        return self.retry(self._client.kv.put, self.sync_path, value, cas=index)
 
     @catch_consul_errors
     def delete_sync_state(self, index=None):
-        return self._client.kv.delete(self.sync_path, cas=index)
+        return self.retry(self._client.kv.delete, self.sync_path, cas=index)
 
     def watch(self, leader_index, timeout):
+        self._last_session_refresh = 0
         if self.__do_not_watch:
             self.__do_not_watch = False
             return True
@@ -316,7 +528,7 @@ class Consul(AbstractDCS):
                     idx, _ = self._client.kv.get(self.leader_path, index=leader_index, wait=str(timeout) + 's')
                     return str(idx) != str(leader_index)
                 except (ConsulException, HTTPException, HTTPError, socket.error, socket.timeout):
-                    logging.exception('watch')
+                    logger.exception('watch')
 
                 timeout = end_time - time.time()
 

patroni/dcs/etcd.py

@@ -1,18 +1,20 @@
 from __future__ import absolute_import
 import etcd
+import json
 import logging
 import os
 import urllib3.util.connection
 import random
 import requests
+import six
 import socket
 import time
 
 from dns.exception import DNSException
 from dns import resolver
-from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState
+from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
 from patroni.exceptions import DCSError
-from patroni.utils import Retry, RetryFailedError
+from patroni.utils import Retry, RetryFailedError, split_host_port, uri
 from urllib3.exceptions import HTTPError, ReadTimeoutError
 from requests.exceptions import RequestException
 from six.moves.queue import Queue
@@ -81,9 +83,12 @@ class Client(etcd.Client):
         args = {p: config.get(p) for p in ('host', 'port', 'protocol', 'use_proxies', 'username', 'password',
                                            'cert', 'ca_cert') if config.get(p)}
         super(Client, self).__init__(read_timeout=config['retry_timeout'], **args)
+        # For some reason python3-etcd on debian and ubuntu are not based on the latest version
+        # Workaround for the case when https://github.com/jplana/python-etcd/pull/196 is not applied
+        self.http.connection_pool_kw.pop('ssl_version', None)
         self._config = config
         self._load_machines_cache()
-        self._allow_reconnect = not self._use_proxies
+        self._allow_reconnect = True
 
     def _build_request_parameters(self):
         kwargs = {'headers': self._get_headers(), 'redirect': self.allow_redirect}
@@ -122,8 +127,11 @@ class Client(etcd.Client):
         while True:
             try:
                 response = self.http.request(self._MGET, self._base_uri + self.version_prefix + '/machines', **kwargs)
-                machines = [n.strip() for n in self._handle_server_response(response).data.decode('utf-8').split(',')]
+                data = self._handle_server_response(response).data.decode('utf-8')
+                machines = [m.strip() for m in data.split(',') if m.strip()]
                 logger.debug("Retrieved list of machines: %s", machines)
+                if not machines:
+                    raise etcd.EtcdException
                 random.shuffle(machines)
                 for url in machines:
                     r = urlparse(url)
@@ -180,10 +188,8 @@ class Client(etcd.Client):
         # Update machines_cache if previous attempt of update has failed
         if self._update_machines_cache:
             self._load_machines_cache()
-        elif time.time() - self._machines_cache_updated > self._machines_cache_ttl:
+        elif not self._use_proxies and time.time() - self._machines_cache_updated > self._machines_cache_ttl:
-            self._machines_cache = self.machines
+            self._refresh_machines_cache()
-            if self._base_uri in self._machines_cache:
-                self._machines_cache.remove(self._base_uri)
             self._machines_cache_updated = time.time()
 
         kwargs.update(self._build_request_parameters())
@@ -200,12 +206,13 @@ class Client(etcd.Client):
 
                 if response is False:
                     some_request_failed = True
-            if some_request_failed and not self._use_proxies:
+            if some_request_failed:
-                self._machines_cache = self.machines
+                self._refresh_machines_cache()
-                if self._base_uri in self._machines_cache:
+        except etcd.EtcdConnectionFailed as e:
-                    self._machines_cache.remove(self._base_uri)
+            if isinstance(e, etcd.EtcdWatchTimedOut) and self._machines_cache:
-        except etcd.EtcdConnectionFailed:
+                self._base_uri = self._next_server()
-            self._update_machines_cache = True
+            else:
+                self._update_machines_cache = True
             if not response:
                 raise
         return self._handle_server_response(response)
@@ -227,7 +234,7 @@ class Client(etcd.Client):
             protocol = 'https' if '-ssl' in r else 'http'
             endpoint = '/members' if '-server' in r else ''
             for host, port in self.get_srv_record('_etcd{0}._tcp.{1}'.format(r, srv)):
-                url = '{0}://{1}:{2}{3}'.format(protocol, host, port, endpoint)
+                url = uri(protocol, (host, port), endpoint)
                 if endpoint:
                     try:
                         response = requests.get(url, timeout=self.read_timeout, verify=False)
@@ -249,15 +256,25 @@ class Client(etcd.Client):
     def _get_machines_cache_from_dns(self, host, port):
         """One host might be resolved into multiple ip addresses. We will make list out of it"""
         if self.protocol == 'http':
-            ret = []
+            ret = map(lambda res: uri(self.protocol, res[-1][:2]), self._dns_resolver.resolve(host, port))
-            for af, _, _, _, sa in self._dns_resolver.resolve(host, port):
-                host, port = sa[:2]
-                if af == socket.AF_INET6:
-                    host = '[{0}]'.format(host)
-                ret.append('{0}://{1}:{2}'.format(self.protocol, host, port))
             if ret:
                 return list(set(ret))
-        return ['{0}://{1}:{2}'.format(self.protocol, host, port)]
+        return [uri(self.protocol, (host, port))]
+
+    def _get_machines_cache_from_config(self):
+        if 'proxy' in self._config:
+            return [uri(self.protocol, (self._config['host'], self._config['port']))]
+
+        machines_cache = []
+        if 'srv' in self._config:
+            machines_cache = self._get_machines_cache_from_srv(self._config['srv'])
+
+        if not machines_cache and 'hosts' in self._config:
+            machines_cache = list(self._config['hosts'])
+
+        if not machines_cache and 'host' in self._config:
+            machines_cache = self._get_machines_cache_from_dns(self._config['host'], self._config['port'])
+        return machines_cache
 
     def _load_machines_cache(self):
         """This method should fill up `_machines_cache` from scratch.
@@ -267,19 +284,10 @@ class Client(etcd.Client):
 
         self._update_machines_cache = True
 
-        if 'srv' not in self._config and 'host' not in self._config:
+        if 'srv' not in self._config and 'host' not in self._config and 'hosts' not in self._config:
-            raise Exception('Neither srv nor host url are defined in etcd section of config')
+            raise Exception('Neither srv, hosts, host nor url are defined in etcd section of config')
 
-        if self._use_proxies:
+        self._machines_cache = self._get_machines_cache_from_config()
-            self._machines_cache = ['{0}://{1}:{2}'.format(self.protocol, self._config['host'], self._config['port'])]
-        else:
-            self._machines_cache = []
-
-            if 'srv' in self._config:
-                self._machines_cache = self._get_machines_cache_from_srv(self._config['srv'])
-
-            if not self._machines_cache and 'host' in self._config:
-                self._machines_cache = self._get_machines_cache_from_dns(self._config['host'], self._config['port'])
 
         # Can not bootstrap list of etcd-cluster members, giving up
         if not self._machines_cache:
@@ -287,26 +295,15 @@ class Client(etcd.Client):
 
         # After filling up initial list of machines_cache we should ask etcd-cluster about actual list
         self._base_uri = self._next_server()
-        self._machines_cache = self.machines
+        self._refresh_machines_cache()
-
-        if self._base_uri in self._machines_cache:
-            self._machines_cache.remove(self._base_uri)
 
         self._update_machines_cache = False
         self._machines_cache_updated = time.time()
 
-
+    def _refresh_machines_cache(self):
-def catch_etcd_errors(func):
+        self._machines_cache = self._get_machines_cache_from_config() if self._use_proxies else self.machines
-    def wrapper(*args, **kwargs):
+        if self._base_uri in self._machines_cache:
-        try:
+            self._machines_cache.remove(self._base_uri)
-            return func(*args, **kwargs) is not None
-        except (RetryFailedError, etcd.EtcdException):
-            return False
-        except:
-            logger.exception("")
-            raise EtcdError("unexpected error")
-
-    return wrapper
 
 
 class Etcd(AbstractDCS):
@@ -320,10 +317,36 @@ class Etcd(AbstractDCS):
                                               etcd.EtcdEventIndexCleared))
         self._client = self.get_etcd_client(config)
         self.__do_not_watch = False
+        self._has_failed = False
 
     def retry(self, *args, **kwargs):
         return self._retry.copy()(*args, **kwargs)
 
+    def _handle_exception(self, e, name='', do_sleep=False, raise_ex=None):
+        if not self._has_failed:
+            logger.exception(name)
+        else:
+            logger.error(e)
+            if do_sleep:
+                time.sleep(1)
+        self._has_failed = True
+        if isinstance(raise_ex, Exception):
+            raise raise_ex
+
+    def catch_etcd_errors(func):
+        def wrapper(self, *args, **kwargs):
+            try:
+                retval = func(self, *args, **kwargs) is not None
+                self._has_failed = False
+                return retval
+            except (RetryFailedError, etcd.EtcdException) as e:
+                self._handle_exception(e)
+                return False
+            except Exception as e:
+                self._handle_exception(e, raise_ex=EtcdError('unexpected error'))
+
+        return wrapper
+
     @staticmethod
     def get_etcd_client(config):
         if 'proxy' in config:
@@ -334,8 +357,20 @@ class Etcd(AbstractDCS):
             r = urlparse(config['url'])
             config.update({'protocol': r.scheme, 'host': r.hostname, 'port': r.port or 2379,
                            'username': r.username, 'password': r.password})
+        elif 'hosts' in config:
+            hosts = config.pop('hosts')
+            default_port = config.pop('port', 2379)
+            protocol = config.get('protocol', 'http')
+
+            if isinstance(hosts, six.string_types):
+                hosts = hosts.split(',')
+
+            config['hosts'] = []
+            for value in hosts:
+                if isinstance(value, six.string_types):
+                    config['hosts'].append(uri(protocol, split_host_port(value, default_port)))
         elif 'host' in config:
-            host, port = (config['host'] + ':2379').split(':')[:2]
+            host, port = split_host_port(config['host'], 2379)
             config['host'] = host
             if 'port' not in config:
                 config['port'] = int(port)
@@ -389,6 +424,8 @@ class Etcd(AbstractDCS):
         while not client:
             try:
                 client = Client(config, dns_resolver)
+                if 'use_proxies' in config and not client.machines:
+                    raise etcd.EtcdException
             except etcd.EtcdException:
                 logger.info('waiting on etcd')
                 time.sleep(5)
@@ -400,6 +437,10 @@ class Etcd(AbstractDCS):
         self._ttl = ttl
         self._client.set_machines_cache_ttl(ttl*10)
 
+    @property
+    def ttl(self):
+        return self._ttl
+
     def set_retry_timeout(self, retry_timeout):
         self._retry.deadline = retry_timeout
         self._client.set_read_timeout(retry_timeout)
@@ -409,9 +450,10 @@ class Etcd(AbstractDCS):
         return Member.from_node(node.modifiedIndex, os.path.basename(node.key), node.ttl, node.value)
 
     def _load_cluster(self):
+        cluster = None
         try:
             result = self.retry(self._client.read, self.client_path(''), recursive=True)
-            nodes = {os.path.relpath(node.key, result.key): node for node in result.leaves}
+            nodes = {node.key[len(result.key):].lstrip('/'): node for node in result.leaves}
 
             # get initialize flag
             initialize = nodes.get(self._INITIALIZE)
@@ -421,6 +463,10 @@ class Etcd(AbstractDCS):
             config = nodes.get(self._CONFIG)
             config = config and ClusterConfig.from_node(config.modifiedIndex, config.value)
 
+            # get timeline history
+            history = nodes.get(self._HISTORY)
+            history = history and TimelineHistory.from_node(history.modifiedIndex, history.value)
+
             # get last leader operation
             last_leader_operation = nodes.get(self._LEADER_OPTIME)
             last_leader_operation = 0 if last_leader_operation is None else int(last_leader_operation.value)
@@ -445,16 +491,18 @@ class Etcd(AbstractDCS):
             sync = nodes.get(self._SYNC)
             sync = SyncState.from_node(sync and sync.modifiedIndex, sync and sync.value)
 
-            self._cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync)
+            cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
         except etcd.EtcdKeyNotFound:
-            self._cluster = Cluster(None, None, None, None, [], None, None)
+            cluster = Cluster(None, None, None, None, [], None, None, None)
-        except:
+        except Exception as e:
-            logger.exception('get_cluster')
+            self._handle_exception(e, 'get_cluster', raise_ex=EtcdError('Etcd is not responding properly'))
-            raise EtcdError('Etcd is not responding properly')
+        self._has_failed = False
+        return cluster
 
     @catch_etcd_errors
-    def touch_member(self, data, ttl=None, permanent=False):
+    def touch_member(self, data, permanent=False):
-        return self.retry(self._client.set, self.member_path, data, None if permanent else ttl or self._ttl)
+        data = json.dumps(data, separators=(',', ':'))
+        return self._client.set(self.member_path, data, None if permanent else self._ttl)
 
     @catch_etcd_errors
     def take_leader(self):
@@ -486,7 +534,7 @@ class Etcd(AbstractDCS):
         return self._client.set(self.leader_optime_path, last_operation)
 
     @catch_etcd_errors
-    def update_leader(self):
+    def _update_leader(self):
         return self.retry(self._client.test_and_set, self.leader_path, self._name, self._name, self._ttl)
 
     @catch_etcd_errors
@@ -505,9 +553,13 @@ class Etcd(AbstractDCS):
     def delete_cluster(self):
         return self.retry(self._client.delete, self.client_path(''), recursive=True)
 
+    @catch_etcd_errors
+    def set_history_value(self, value):
+        return self._client.write(self.history_path, value)
+
     @catch_etcd_errors
     def set_sync_state_value(self, value, index=None):
-        return self._client.write(self.sync_path, value, prevIndex=index or 0)
+        return self.retry(self._client.write, self.sync_path, value, prevIndex=index or 0)
 
     @catch_etcd_errors
     def delete_sync_state(self, index=None):
@@ -524,16 +576,19 @@ class Etcd(AbstractDCS):
             while timeout >= 1:  # when timeout is too small urllib3 doesn't have enough time to connect
                 try:
                     self._client.watch(self.leader_path, index=leader_index, timeout=timeout + 0.5)
+                    self._has_failed = False
                     # Synchronous work of all cluster members with etcd is less expensive
                     # than reestablishing http connection every time from every replica.
                     return True
                 except etcd.EtcdWatchTimedOut:
                     self._client.http.clear()
+                    self._has_failed = False
                     return False
                 except (etcd.EtcdEventIndexCleared, etcd.EtcdWatcherCleared):  # Watch failed
+                    self._has_failed = False
                     return True  # leave the loop, because watch with the same parameters will fail anyway
-                except etcd.EtcdException:
+                except etcd.EtcdException as e:
-                    logger.exception('watch')
+                    self._handle_exception(e, 'watch', True)
 
                 timeout = end_time - time.time()
 

patroni/dcs/exhibitor.py

@@ -4,6 +4,7 @@ import requests
 import time
 
 from patroni.dcs.zookeeper import ZooKeeper
+from patroni.utils import uri
 from requests.exceptions import RequestException
 
 logger = logging.getLogger(__name__)
@@ -46,9 +47,8 @@ class ExhibitorEnsembleProvider(object):
     def _query_exhibitors(self, exhibitors):
         random.shuffle(exhibitors)
         for host in exhibitors:
-            uri = 'http://{0}:{1}{2}'.format(host, self._exhibitor_port, self._uri_path)
             try:
-                response = requests.get(uri, timeout=self.TIMEOUT)
+                response = requests.get(uri('http', (host, self._exhibitor_port), self._uri_path), timeout=self.TIMEOUT)
                 return response.json()
             except RequestException:
                 pass

patroni/dcs/kubernetes.py

@@ -0,0 +1,441 @@
+from __future__ import absolute_import
+import datetime
+import functools
+import json
+import logging
+import socket
+import sys
+import time
+
+from kubernetes import client as k8s_client, config as k8s_config, watch as k8s_watch
+from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
+from patroni.exceptions import DCSError
+from patroni.utils import deep_compare, tzutc, Retry, RetryFailedError
+from urllib3.exceptions import HTTPError
+from six.moves.http_client import HTTPException
+
+logger = logging.getLogger(__name__)
+
+
+class KubernetesError(DCSError):
+    pass
+
+
+class KubernetesRetriableException(k8s_client.rest.ApiException):
+
+    def __init__(self, orig):
+        super(KubernetesRetriableException, self).__init__(orig.status, orig.reason)
+        self.body = orig.body
+        self.headers = orig.headers
+
+
+class CoreV1ApiProxy(object):
+
+    def __init__(self, use_endpoints=False):
+        self._api = k8s_client.CoreV1Api()
+        self._request_timeout = None
+        self._use_endpoints = use_endpoints
+
+    def set_timeout(self, timeout):
+        self._request_timeout = (1, timeout / 3.0)
+
+    def __getattr__(self, func):
+        if func.endswith('_kind'):
+            func = func[:-4] + ('endpoints' if self._use_endpoints else 'config_map')
+
+        def wrapper(*args, **kwargs):
+            if '_request_timeout' not in kwargs:
+                kwargs['_request_timeout'] = self._request_timeout
+            try:
+                return getattr(self._api, func)(*args, **kwargs)
+            except k8s_client.rest.ApiException as e:
+                if e.status in (502, 503, 504):  # XXX
+                    raise KubernetesRetriableException(e)
+                raise
+        return wrapper
+
+
+def catch_kubernetes_errors(func):
+    def wrapper(*args, **kwargs):
+        try:
+            return func(*args, **kwargs)
+        except k8s_client.rest.ApiException as e:
+            if e.status == 403:
+                logger.exception('Permission denied')
+            elif e.status != 409:  # Object exists or conflict in resource_version
+                logger.exception('Unexpected error from Kubernetes API')
+            return False
+        except (RetryFailedError, HTTPException, HTTPError, socket.error, socket.timeout):
+            return False
+    return wrapper
+
+
+class Kubernetes(AbstractDCS):
+
+    def __init__(self, config):
+        self._labels = config['labels']
+        self._labels[config.get('scope_label', 'cluster-name')] = config['scope']
+        self._label_selector = ','.join('{0}={1}'.format(k, v) for k, v in self._labels.items())
+        self._namespace = config.get('namespace') or 'default'
+        self._role_label = config.get('role_label', 'role')
+        config['namespace'] = ''
+        super(Kubernetes, self).__init__(config)
+        self._retry = Retry(deadline=config['retry_timeout'], max_delay=1, max_tries=-1,
+                            retry_exceptions=(KubernetesRetriableException, HTTPException,
+                                              HTTPError, socket.error, socket.timeout))
+        self._ttl = None
+        try:
+            k8s_config.load_incluster_config()
+        except k8s_config.ConfigException:
+            k8s_config.load_kube_config(context=config.get('context', 'local'))
+
+        self.__subsets = None
+        use_endpoints = config.get('use_endpoints') and (config.get('patronictl') or 'pod_ip' in config)
+        if use_endpoints:
+            addresses = [k8s_client.V1EndpointAddress(ip='127.0.0.1' if config.get('patronictl') else config['pod_ip'])]
+            ports = []
+            for p in config.get('ports', [{}]):
+                port = {'port': int(p.get('port', '5432'))}
+                port.update({n: p[n] for n in ('name', 'protocol') if p.get(n)})
+                ports.append(k8s_client.V1EndpointPort(**port))
+            self.__subsets = [k8s_client.V1EndpointSubset(addresses=addresses, ports=ports)]
+            self._should_create_config_service = True
+        self._api = CoreV1ApiProxy(use_endpoints)
+        self.set_retry_timeout(config['retry_timeout'])
+        self.set_ttl(config.get('ttl') or 30)
+        self._leader_observed_record = {}
+        self._leader_observed_time = None
+        self._leader_resource_version = None
+        self._leader_observed_subsets = []
+        self._config_resource_version = None
+        self.__do_not_watch = False
+
+    def retry(self, *args, **kwargs):
+        return self._retry.copy()(*args, **kwargs)
+
+    def client_path(self, path):
+        return super(Kubernetes, self).client_path(path)[1:].replace('/', '-')
+
+    @property
+    def leader_path(self):
+        return self._base_path[1:] if self.__subsets else super(Kubernetes, self).leader_path
+
+    def set_ttl(self, ttl):
+        ttl = int(ttl)
+        self.__do_not_watch = self._ttl != ttl
+        self._ttl = ttl
+
+    @property
+    def ttl(self):
+        return self._ttl
+
+    def set_retry_timeout(self, retry_timeout):
+        self._retry.deadline = retry_timeout
+        self._api.set_timeout(retry_timeout)
+
+    @staticmethod
+    def member(pod):
+        annotations = pod.metadata.annotations or {}
+        member = Member.from_node(pod.metadata.resource_version, pod.metadata.name, None, annotations.get('status', ''))
+        member.data['pod_labels'] = pod.metadata.labels
+        return member
+
+    def _load_cluster(self):
+        try:
+            # get list of members
+            response = self.retry(self._api.list_namespaced_pod, self._namespace, label_selector=self._label_selector)
+            members = [self.member(pod) for pod in response.items]
+
+            response = self.retry(self._api.list_namespaced_kind, self._namespace, label_selector=self._label_selector)
+            nodes = {item.metadata.name: item for item in response.items}
+
+            config = nodes.get(self.config_path)
+            metadata = config and config.metadata
+            self._config_resource_version = metadata.resource_version if metadata else None
+            annotations = metadata and metadata.annotations or {}
+
+            # get initialize flag
+            initialize = annotations.get(self._INITIALIZE)
+
+            # get global dynamic configuration
+            config = ClusterConfig.from_node(metadata and metadata.resource_version,
+                                             annotations.get(self._CONFIG) or '{}',
+                                             metadata.resource_version if self._CONFIG in annotations else 0)
+
+            # get timeline history
+            history = TimelineHistory.from_node(metadata and metadata.resource_version,
+                                                annotations.get(self._HISTORY) or '[]')
+
+            leader = nodes.get(self.leader_path)
+            metadata = leader and leader.metadata
+            self._leader_resource_version = metadata.resource_version if metadata else None
+            self._leader_observed_subsets = leader.subsets if self.__subsets and leader and leader.subsets else []
+            annotations = metadata and metadata.annotations or {}
+
+            # get last leader operation
+            last_leader_operation = annotations.get(self._OPTIME)
+            last_leader_operation = 0 if last_leader_operation is None else int(last_leader_operation)
+
+            # get leader
+            leader_record = {n: annotations.get(n) for n in (self._LEADER, 'acquireTime',
+                             'ttl', 'renewTime', 'transitions') if n in annotations}
+            if (leader_record or self._leader_observed_record) and leader_record != self._leader_observed_record:
+                self._leader_observed_record = leader_record
+                self._leader_observed_time = time.time()
+
+            leader = leader_record.get(self._LEADER)
+            try:
+                ttl = int(leader_record.get('ttl')) or self._ttl
+            except (TypeError, ValueError):
+                ttl = self._ttl
+
+            if not metadata or not self._leader_observed_time or self._leader_observed_time + ttl < time.time():
+                leader = None
+
+            if metadata:
+                member = Member(-1, leader, None, {})
+                member = ([m for m in members if m.name == leader] or [member])[0]
+                leader = Leader(response.metadata.resource_version, None, member)
+
+            # failover key
+            failover = nodes.get(self.failover_path)
+            metadata = failover and failover.metadata
+            failover = Failover.from_node(metadata and metadata.resource_version, metadata and metadata.annotations)
+
+            # get synchronization state
+            sync = nodes.get(self.sync_path)
+            metadata = sync and sync.metadata
+            sync = SyncState.from_node(metadata and metadata.resource_version,  metadata and metadata.annotations)
+
+            return Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
+        except Exception:
+            logger.exception('get_cluster')
+            raise KubernetesError('Kubernetes API is not responding properly')
+
+    @staticmethod
+    def compare_ports(p1, p2):
+        return p1.name == p2.name and p1.port == p2.port and (p1.protocol or 'TCP') == (p2.protocol or 'TCP')
+
+    @staticmethod
+    def subsets_changed(last_observed_subsets, subsets):
+        """
+        >>> Kubernetes.subsets_changed([], [])
+        False
+        >>> Kubernetes.subsets_changed([], [k8s_client.V1EndpointSubset()])
+        True
+        >>> s1 = [k8s_client.V1EndpointSubset(addresses=[k8s_client.V1EndpointAddress(ip='1.2.3.4')])]
+        >>> s2 = [k8s_client.V1EndpointSubset(addresses=[k8s_client.V1EndpointAddress(ip='1.2.3.5')])]
+        >>> Kubernetes.subsets_changed(s1, s2)
+        True
+        >>> a = [k8s_client.V1EndpointAddress(ip='1.2.3.4')]
+        >>> s1 = [k8s_client.V1EndpointSubset(addresses=a, ports=[k8s_client.V1EndpointPort(protocol='TCP', port=1)])]
+        >>> s2 = [k8s_client.V1EndpointSubset(addresses=a, ports=[k8s_client.V1EndpointPort(port=5432)])]
+        >>> Kubernetes.subsets_changed(s1, s2)
+        True
+        >>> p1 = k8s_client.V1EndpointPort(name='port1', port=1)
+        >>> p2 = k8s_client.V1EndpointPort(name='port2', port=2)
+        >>> p3 = k8s_client.V1EndpointPort(name='port3', port=3)
+        >>> s1 = [k8s_client.V1EndpointSubset(addresses=a, ports=[p1, p2])]
+        >>> s2 = [k8s_client.V1EndpointSubset(addresses=a, ports=[p2, p3])]
+        >>> Kubernetes.subsets_changed(s1, s2)
+        True
+        >>> s2 = [k8s_client.V1EndpointSubset(addresses=a, ports=[p2, p1])]
+        >>> Kubernetes.subsets_changed(s1, s2)
+        False
+        """
+        if len(last_observed_subsets) != len(subsets):
+            return True
+        if subsets == []:
+            return False
+        if len(last_observed_subsets[0].addresses or []) != 1 or \
+                last_observed_subsets[0].addresses[0].ip != subsets[0].addresses[0].ip or \
+                len(last_observed_subsets[0].ports) != len(subsets[0].ports):
+            return True
+        if len(subsets[0].ports) == 1:
+            return not Kubernetes.compare_ports(last_observed_subsets[0].ports[0], subsets[0].ports[0])
+        observed_ports = {p.name: p for p in last_observed_subsets[0].ports}
+        for p in subsets[0].ports:
+            if p.name not in observed_ports or not Kubernetes.compare_ports(p, observed_ports.pop(p.name)):
+                return True
+        return False
+
+    @catch_kubernetes_errors
+    def patch_or_create(self, name, annotations, resource_version=None, patch=False, retry=True, subsets=None):
+        metadata = {'namespace': self._namespace, 'name': name, 'labels': self._labels, 'annotations': annotations}
+        if patch or resource_version:
+            if resource_version is not None:
+                metadata['resource_version'] = resource_version
+            func = functools.partial(self._api.patch_namespaced_kind, name)
+        else:
+            func = functools.partial(self._api.create_namespaced_kind)
+            # skip annotations with null values
+            metadata['annotations'] = {k: v for k, v in metadata['annotations'].items() if v is not None}
+
+        metadata = k8s_client.V1ObjectMeta(**metadata)
+        if subsets is not None and self.__subsets:
+            endpoints = {'metadata': metadata}
+            if self.subsets_changed(self._leader_observed_subsets, subsets):
+                endpoints['subsets'] = subsets
+            body = k8s_client.V1Endpoints(**endpoints)
+        else:
+            body = k8s_client.V1ConfigMap(metadata=metadata)
+        return self.retry(func, self._namespace, body) if retry else func(self._namespace, body)
+
+    def patch_or_create_config(self, annotations, resource_version=None, patch=False, retry=True):
+        # SCOPE-config endpoint requires corresponding service otherwise it might be "cleaned" by k8s master
+        if self.__subsets and not patch and not resource_version:
+            self._should_create_config_service = True
+            self._create_config_service()
+        ret = self.patch_or_create(self.config_path, annotations, resource_version, patch, retry)
+        if ret:
+            self._config_resource_version = ret.metadata.resource_version
+        return ret
+
+    def _create_config_service(self):
+        metadata = k8s_client.V1ObjectMeta(namespace=self._namespace, name=self.config_path, labels=self._labels)
+        body = k8s_client.V1Service(metadata=metadata, spec=k8s_client.V1ServiceSpec(cluster_ip='None'))
+        try:
+            if not self._api.create_namespaced_service(self._namespace, body):
+                return
+        except Exception as e:
+            if not isinstance(e, k8s_client.rest.ApiException) or e.status != 409:  # Service already exists
+                return logger.exception('create_config_service failed')
+        self._should_create_config_service = False
+
+    def _write_leader_optime(self, last_operation):
+        """Unused"""
+
+    def _update_leader(self):
+        """Unused"""
+
+    def update_leader(self, last_operation, access_is_restricted=False):
+        now = datetime.datetime.now(tzutc).isoformat()
+        annotations = {self._LEADER: self._name, 'ttl': str(self._ttl), 'renewTime': now,
+                       'acquireTime': self._leader_observed_record.get('acquireTime') or now,
+                       'transitions': self._leader_observed_record.get('transitions') or '0'}
+        if last_operation:
+            annotations[self._OPTIME] = last_operation
+
+        subsets = [] if access_is_restricted else self.__subsets
+
+        ret = self.patch_or_create(self.leader_path, annotations, self._leader_resource_version, subsets=subsets)
+        if ret:
+            self._leader_resource_version = ret.metadata.resource_version
+        return ret
+
+    def attempt_to_acquire_leader(self, permanent=False):
+        now = datetime.datetime.now(tzutc).isoformat()
+        annotations = {self._LEADER: self._name, 'ttl': str(sys.maxsize if permanent else self._ttl),
+                       'renewTime': now, 'acquireTime': now, 'transitions': '0'}
+        if self._leader_observed_record:
+            try:
+                transitions = int(self._leader_observed_record.get('transitions'))
+            except (TypeError, ValueError):
+                transitions = 0
+
+            if self._leader_observed_record.get(self._LEADER) != self._name:
+                transitions += 1
+            else:
+                annotations['acquireTime'] = self._leader_observed_record.get('acquireTime') or now
+            annotations['transitions'] = str(transitions)
+        subsets = [] if self.__subsets else None
+        ret = self.patch_or_create(self.leader_path, annotations, self._leader_resource_version, subsets=subsets)
+        if ret:
+            self._leader_resource_version = ret.metadata.resource_version
+        else:
+            logger.info('Could not take out TTL lock')
+        return ret
+
+    def take_leader(self):
+        return self.attempt_to_acquire_leader()
+
+    def set_failover_value(self, value, index=None):
+        """Unused"""
+
+    def manual_failover(self, leader, candidate, scheduled_at=None, index=None):
+        annotations = {'leader': leader or None, 'member': candidate or None, 'scheduled_at': scheduled_at}
+        patch = bool(self.cluster and isinstance(self.cluster.failover, Failover) and self.cluster.failover.index)
+        return self.patch_or_create(self.failover_path, annotations, index, bool(index or patch), False)
+
+    def set_config_value(self, value, index=None):
+        return self.patch_or_create_config({self._CONFIG: value}, index, bool(self._config_resource_version), False)
+
+    @catch_kubernetes_errors
+    def touch_member(self, data, permanent=False):
+        cluster = self.cluster
+        if cluster and cluster.leader and cluster.leader.name == self._name:
+            role = 'promoted' if data['role'] in ('replica', 'promoted') else 'master'
+        elif data['state'] == 'running' and data['role'] != 'master':
+            role = data['role']
+        else:
+            role = None
+
+        member = cluster and cluster.get_member(self._name, fallback_to_leader=False)
+        pod_labels = member and member.data.pop('pod_labels', None)
+        ret = pod_labels is not None and pod_labels.get(self._role_label) == role and deep_compare(data, member.data)
+
+        if not ret:
+            metadata = {'namespace': self._namespace, 'name': self._name, 'labels': {self._role_label: role},
+                        'annotations': {'status': json.dumps(data, separators=(',', ':'))}}
+            body = k8s_client.V1Pod(metadata=k8s_client.V1ObjectMeta(**metadata))
+            ret = self._api.patch_namespaced_pod(self._name, self._namespace, body)
+        if self.__subsets and self._should_create_config_service:
+            self._create_config_service()
+        return ret
+
+    def initialize(self, create_new=True, sysid=""):
+        cluster = self.cluster
+        resource_version = cluster.config.index if cluster and cluster.config and cluster.config.index else None
+        return self.patch_or_create_config({self._INITIALIZE: sysid}, resource_version)
+
+    def delete_leader(self):
+        if self.cluster and isinstance(self.cluster.leader, Leader) and self.cluster.leader.name == self._name:
+            self.patch_or_create(self.leader_path, {self._LEADER: None}, self._leader_resource_version, True, False, [])
+            self.reset_cluster()
+
+    def cancel_initialization(self):
+        self.patch_or_create_config({self._INITIALIZE: None}, self._config_resource_version, True)
+
+    @catch_kubernetes_errors
+    def delete_cluster(self):
+        self.retry(self._api.delete_collection_namespaced_kind, self._namespace, label_selector=self._label_selector)
+
+    def set_history_value(self, value):
+        return self.patch_or_create_config({self._HISTORY: value}, None, bool(self._config_resource_version), False)
+
+    def set_sync_state_value(self, value, index=None):
+        """Unused"""
+
+    def write_sync_state(self, leader, sync_standby, index=None):
+        return self.patch_or_create(self.sync_path, self.sync_state(leader, sync_standby), index, False)
+
+    def delete_sync_state(self, index=None):
+        return self.write_sync_state(None, None, index)
+
+    def watch(self, leader_index, timeout):
+        if self.__do_not_watch:
+            self.__do_not_watch = False
+            return True
+
+        if leader_index:
+            end_time = time.time() + timeout
+            w = k8s_watch.Watch()
+            while timeout >= 1:
+                try:
+                    for event in w.stream(self._api.list_namespaced_kind, self._namespace,
+                                          resource_version=leader_index, timeout_seconds=int(timeout + 0.5),
+                                          field_selector='metadata.name=' + self.leader_path,
+                                          _request_timeout=(1, timeout + 1)):
+                        return event['raw_object'].get('metadata', {}).get('resourceVersion') != leader_index
+                    return False
+                except KeyboardInterrupt:
+                    raise
+                except Exception:
+                    logger.exception('watch')
+
+                timeout = end_time - time.time()
+
+        try:
+            return super(Kubernetes, self).watch(None, timeout)
+        finally:
+            self.event.clear()

patroni/dcs/zookeeper.py

@@ -1,10 +1,14 @@
+import json
 import logging
+import select
+import time
 
-from kazoo.client import KazooClient, KazooState
+from kazoo.client import KazooClient, KazooState, KazooRetry
 from kazoo.exceptions import NoNodeError, NodeExistsError
 from kazoo.handlers.threading import SequentialThreadingHandler
-from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState
+from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
 from patroni.exceptions import DCSError
+from patroni.utils import deep_compare
 
 logger = logging.getLogger(__name__)
 
@@ -34,12 +38,21 @@ class PatroniSequentialThreadingHandler(SequentialThreadingHandler):
                      `connect_timeout` (negotiated session timeout) as the second element."""
 
         args = list(args)
-        if len(args) == 1:
+        if len(args) == 0:  # kazoo 2.6.0 slightly changed the way how it calls create_connection method
+            kwargs['timeout'] = max(self._connect_timeout, kwargs.get('timeout', self._connect_timeout*10)/10.0)
+        elif len(args) == 1:
             args.append(self._connect_timeout)
         else:
             args[1] = max(self._connect_timeout, args[1]/10.0)
         return super(PatroniSequentialThreadingHandler, self).create_connection(*args, **kwargs)
 
+    def select(self, *args, **kwargs):
+        """Python3 raises `ValueError` if socket is closed, because fd == -1"""
+        try:
+            return super(PatroniSequentialThreadingHandler, self).select(*args, **kwargs)
+        except ValueError as e:
+            raise select.error(9, str(e))
+
 
 class ZooKeeper(AbstractDCS):
 
@@ -51,11 +64,11 @@ class ZooKeeper(AbstractDCS):
             hosts = ','.join(hosts)
 
         self._client = KazooClient(hosts, handler=PatroniSequentialThreadingHandler(config['retry_timeout']),
-                                   timeout=config['ttl'], connection_retry={'max_delay': 1, 'max_tries': -1},
+                                   timeout=config['ttl'], connection_retry=KazooRetry(max_delay=1, max_tries=-1,
-                                   command_retry={'deadline': config['retry_timeout'], 'max_delay': 1, 'max_tries': -1})
+                                   sleep_func=time.sleep), command_retry=KazooRetry(deadline=config['retry_timeout'],
+                                   max_delay=1, max_tries=-1, sleep_func=time.sleep))
         self._client.add_listener(self.session_listener)
 
-        self._my_member_data = None
         self._fetch_cluster = True
 
         self._orig_kazoo_connect = self._client._connection._connect
@@ -113,8 +126,13 @@ class ZooKeeper(AbstractDCS):
             self._client.restart()
             return True
 
+    @property
+    def ttl(self):
+        return self._client._session_timeout
+
     def set_retry_timeout(self, retry_timeout):
-        self._client._retry.deadline = retry_timeout
+        retry = self._client.retry if isinstance(self._client.retry, KazooRetry) else self._client._retry
+        retry.deadline = retry_timeout
 
     def get_node(self, key, watch=None):
         try:
@@ -156,6 +174,10 @@ class ZooKeeper(AbstractDCS):
         config = self.get_node(self.config_path, watch=self.cluster_watcher) if self._CONFIG in nodes else None
         config = config and ClusterConfig.from_node(config[1].version, config[0], config[1].mzxid)
 
+        # get timeline history
+        history = self.get_node(self.history_path, watch=self.cluster_watcher) if self._HISTORY in nodes else None
+        history = history and TimelineHistory.from_node(history[1].mzxid, history[0])
+
         # get last leader operation
         last_leader_operation = self._OPTIME in nodes and self._fetch_cluster and self.get_node(self.leader_optime_path)
         last_leader_operation = last_leader_operation and int(last_leader_operation[0]) or 0
@@ -188,84 +210,98 @@ class ZooKeeper(AbstractDCS):
         failover = self.get_node(self.failover_path, watch=self.cluster_watcher) if self._FAILOVER in nodes else None
         failover = failover and Failover.from_node(failover[1].version, failover[0])
 
-        self._cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync)
+        return Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
 
     def _load_cluster(self):
-        if self._fetch_cluster or self._cluster is None:
+        cluster = self.cluster
+        if self._fetch_cluster or cluster is None:
             try:
-                self._client.retry(self._inner_load_cluster)
+                cluster = self._client.retry(self._inner_load_cluster)
             except Exception:
                 logger.exception('get_cluster')
                 self.cluster_watcher(None)
                 raise ZooKeeperError('ZooKeeper in not responding properly')
+        return cluster
 
-    def _create(self, path, value, **kwargs):
+    def _create(self, path, value, retry=False, ephemeral=False):
         try:
-            self._client.retry(self._client.create, path, value.encode('utf-8'), **kwargs)
+            if retry:
+                self._client.retry(self._client.create, path, value, makepath=True, ephemeral=ephemeral)
+            else:
+                self._client.create_async(path, value, makepath=True, ephemeral=ephemeral).get(timeout=1)
             return True
-        except:
+        except Exception:
-            return False
+            logger.exception('Failed to create %s', path)
+        return False
 
     def attempt_to_acquire_leader(self, permanent=False):
-        ret = self._create(self.leader_path, self._name, makepath=True, ephemeral=not permanent)
+        ret = self._create(self.leader_path, self._name.encode('utf-8'), retry=True, ephemeral=not permanent)
         if not ret:
             logger.info('Could not take out TTL lock')
         return ret
 
-    def set_failover_value(self, value, index=None):
+    def _set_or_create(self, key, value, index=None, retry=False, do_not_create_empty=False):
+        value = value.encode('utf-8')
         try:
-            self._client.retry(self._client.set, self.failover_path, value.encode('utf-8'), version=index or -1)
+            if retry:
+                self._client.retry(self._client.set, key, value, version=index or -1)
+            else:
+                self._client.set_async(key, value, version=index or -1).get(timeout=1)
             return True
         except NoNodeError:
-            return value == '' or (index is None and self._create(self.failover_path, value))
+            if do_not_create_empty and not value:
-        except:
+                return True
-            logging.exception('set_failover_value')
+            elif index is None:
-            return False
+                return self._create(key, value, retry)
+            else:
+                return False
+        except Exception:
+            logger.exception('Failed to update %s', key)
+        return False
+
+    def set_failover_value(self, value, index=None):
+        return self._set_or_create(self.failover_path, value, index)
 
     def set_config_value(self, value, index=None):
-        try:
+        return self._set_or_create(self.config_path, value, index, retry=True)
-            self._client.retry(self._client.set, self.config_path, value.encode('utf-8'), version=index or -1)
-            return True
-        except NoNodeError:
-            return index is None and self._create(self.config_path, value)
-        except Exception:
-            logging.exception('set_config_value')
-            return False
 
     def initialize(self, create_new=True, sysid=""):
-        return self._create(self.initialize_path, sysid, makepath=True) if create_new \
+        sysid = sysid.encode('utf-8')
-            else self._client.retry(self._client.set, self.initialize_path, sysid.encode("utf-8"))
+        return self._create(self.initialize_path, sysid, retry=True) if create_new \
+            else self._client.retry(self._client.set, self.initialize_path, sysid)
 
-    def touch_member(self, data, ttl=None, permanent=False):
+    def touch_member(self, data, permanent=False):
         cluster = self.cluster
         member = cluster and cluster.get_member(self._name, fallback_to_leader=False)
-        data = data.encode('utf-8')
+        encoded_data = json.dumps(data, separators=(',', ':')).encode('utf-8')
-        if member and self._client.client_id is not None and member.session != self._client.client_id[0]:
+        if member and (self._client.client_id is not None and member.session != self._client.client_id[0] or
+                       not (deep_compare(member.data.get('tags', {}), data.get('tags', {})) and
+                            member.data.get('version') == data.get('version') and
+                            member.data.get('checkpoint_after_promote') == data.get('checkpoint_after_promote'))):
             try:
                 self._client.delete_async(self.member_path).get(timeout=1)
             except NoNodeError:
                 pass
-            except:
+            except Exception:
                 return False
             member = None
 
         if member:
-            if data == self._my_member_data:
+            if deep_compare(data, member.data):
                 return True
         else:
             try:
-                self._client.create_async(self.member_path, data, makepath=True, ephemeral=not permanent).get(timeout=1)
+                self._client.create_async(self.member_path, encoded_data, makepath=True,
-                self._my_member_data = data
+                                          ephemeral=not permanent).get(timeout=1)
                 return True
             except Exception as e:
                 if not isinstance(e, NodeExistsError):
                     logger.exception('touch_member')
                     return False
         try:
-            self._client.set_async(self.member_path, data).get(timeout=1)
+            self._client.set_async(self.member_path, encoded_data).get(timeout=1)
-            self._my_member_data = data
             return True
-        except:
+        except Exception:
             logger.exception('touch_member')
 
         return False
@@ -274,26 +310,13 @@ class ZooKeeper(AbstractDCS):
         return self.attempt_to_acquire_leader()
 
     def _write_leader_optime(self, last_operation):
-        last_operation = last_operation.encode('utf-8')
+        return self._set_or_create(self.leader_optime_path, last_operation)
-        try:
-            self._client.set_async(self.leader_optime_path, last_operation).get(timeout=1)
-            return True
-        except NoNodeError:
-            try:
-                self._client.create_async(self.leader_optime_path, last_operation, makepath=True).get(timeout=1)
-                return True
-            except:
-                logger.exception('Failed to create %s', self.leader_optime_path)
-        except:
-            logger.exception('Failed to update %s', self.leader_optime_path)
-        return False
 
-    def update_leader(self):
+    def _update_leader(self):
         return True
 
     def delete_leader(self):
         self._client.restart()
-        self._my_member_data = None
         return True
 
     def _cancel_initialization(self):
@@ -304,7 +327,7 @@ class ZooKeeper(AbstractDCS):
     def cancel_initialization(self):
         try:
             self._client.retry(self._cancel_initialization)
-        except:
+        except Exception:
             logger.exception("Unable to delete initialize key")
 
     def delete_cluster(self):
@@ -313,15 +336,11 @@ class ZooKeeper(AbstractDCS):
         except NoNodeError:
             return True
 
+    def set_history_value(self, value):
+        return self._set_or_create(self.history_path, value)
+
     def set_sync_state_value(self, value, index=None):
-        try:
+        return self._set_or_create(self.sync_path, value, index, retry=True, do_not_create_empty=True)
-            self._client.retry(self._client.set, self.sync_path, value.encode('utf-8'), version=index or -1)
-            return True
-        except NoNodeError:
-            return value == '' or (index is None and self._create(self.sync_path, value))
-        except:
-            logging.exception('set_sync_state_value')
-            return False
 
     def delete_sync_state(self, index=None):
         return self.set_sync_state_value("{}", index)

patroni/exceptions.py

@@ -23,3 +23,7 @@ class DCSError(PatroniException):
 
 class PostgresConnectionException(PostgresException):
     pass
+
+
+class WatchdogError(PatroniException):
+    pass

patroni/log.py

@@ -0,0 +1,158 @@
+import logging
+import os
+import sys
+
+from copy import deepcopy
+from logging.handlers import RotatingFileHandler
+from patroni.utils import deep_compare
+from six.moves.queue import Queue, Full
+from threading import Lock, Thread
+
+_LOGGER = logging.getLogger(__name__)
+
+
+class QueueHandler(logging.Handler):
+
+    def __init__(self):
+        logging.Handler.__init__(self)
+        self.queue = Queue()
+        self._records_lost = 0
+
+    def _put_record(self, record):
+        self.format(record)
+        record.msg = record.message
+        record.args = None
+        record.exc_info = None
+        self.queue.put_nowait(record)
+
+    def _try_to_report_lost_records(self):
+        if self._records_lost:
+            try:
+                record = _LOGGER.makeRecord(_LOGGER.name, logging.WARNING, __file__, 0,
+                                            'QueueHandler has lost %s log records',
+                                            (self._records_lost,), None, 'emit')
+                self._put_record(record)
+                self._records_lost = 0
+            except Exception:
+                pass
+
+    def emit(self, record):
+        try:
+            self._put_record(record)
+            self._try_to_report_lost_records()
+        except Exception:
+            self._records_lost += 1
+
+    @property
+    def records_lost(self):
+        return self._records_lost
+
+
+class PatroniLogger(Thread):
+
+    DEFAULT_LEVEL = 'INFO'
+    DEFAULT_FORMAT = '%(asctime)s %(levelname)s: %(message)s'
+
+    NORMAL_LOG_QUEUE_SIZE = 2  # When everything goes normal Patroni writes only 2 messages per HA loop
+    DEFAULT_MAX_QUEUE_SIZE = 1000
+    LOGGING_BROKEN_EXIT_CODE = 5
+
+    def __init__(self):
+        super(PatroniLogger, self).__init__()
+        self.daemon = True
+        self._queue_handler = QueueHandler()
+        self._root_logger = logging.getLogger()
+        self._root_logger.addHandler(self._queue_handler)
+        self._config = None
+        self._log_handler = None
+        self._old_handlers = []
+        self._log_handler_lock = Lock()
+        self.reload_config({'level': 'DEBUG'})
+        self.start()
+
+    def update_loggers(self):
+        loggers = deepcopy(self._config.get('loggers') or {})
+        for name, logger in self._root_logger.manager.loggerDict.items():
+            if not isinstance(logger, logging.PlaceHolder):
+                level = loggers.pop(name, logging.NOTSET)
+                logger.setLevel(level)
+
+        for name, level in loggers.items():
+            logger = self._root_logger.manager.getLogger(name)
+            logger.setLevel(level)
+
+    def reload_config(self, config):
+        if self._config is None or not deep_compare(self._config, config):
+            with self._queue_handler.queue.mutex:
+                self._queue_handler.queue.maxsize = config.get('max_queue_size', self.DEFAULT_MAX_QUEUE_SIZE)
+
+            self._root_logger.setLevel(config.get('level', PatroniLogger.DEFAULT_LEVEL))
+
+            new_handler = None
+            if 'dir' in config:
+                if not isinstance(self._log_handler, RotatingFileHandler):
+                    new_handler = RotatingFileHandler(os.path.join(config['dir'], __name__))
+                handler = new_handler or self._log_handler
+                handler.maxBytes = int(config.get('file_size', 25000000))
+                handler.backupCount = int(config.get('file_num', 4))
+            else:
+                if self._log_handler is None or isinstance(self._log_handler, RotatingFileHandler):
+                    new_handler = logging.StreamHandler()
+                handler = new_handler or self._log_handler
+
+            oldlogformat = (self._config or {}).get('format', PatroniLogger.DEFAULT_FORMAT)
+            logformat = config.get('format', PatroniLogger.DEFAULT_FORMAT)
+
+            olddateformat = (self._config or {}).get('dateformat') or None
+            dateformat = config.get('dateformat') or None  # Convert empty string to `None`
+
+            if oldlogformat != logformat or olddateformat != dateformat or new_handler:
+                handler.setFormatter(logging.Formatter(logformat, dateformat))
+
+            if new_handler:
+                with self._log_handler_lock:
+                    if self._log_handler:
+                        self._old_handlers.append(self._log_handler)
+                    self._log_handler = new_handler
+
+            self._config = config.copy()
+            self.update_loggers()
+
+    def _close_old_handlers(self):
+        while True:
+            with self._log_handler_lock:
+                if not self._old_handlers:
+                    break
+                handler = self._old_handlers.pop()
+            try:
+                handler.close()
+            except Exception:
+                _LOGGER.exception('Failed to close the old log handler %s', handler)
+
+    def run(self):
+        while True:
+            self._close_old_handlers()
+
+            record = self._queue_handler.queue.get(True)
+            if record is None:
+                break
+
+            self._log_handler.handle(record)
+            self._queue_handler.queue.task_done()
+
+    def shutdown(self):
+        try:
+            self._queue_handler.queue.put_nowait(None)
+        except Full:  # Queue is full.
+            # It seems that logging is not working, exiting with non-standard exit-code is the best we can do.
+            sys.exit(self.LOGGING_BROKEN_EXIT_CODE)
+        self.join()
+        logging.shutdown()
+
+    @property
+    def queue_size(self):
+        return self._queue_handler.queue.qsize()
+
+    @property
+    def records_lost(self):
+        return self._queue_handler.records_lost

patroni/postgresql/__init__.py

@@ -0,0 +1,902 @@
+import logging
+import os
+import psycopg2
+import shlex
+import shutil
+import subprocess
+import time
+
+from contextlib import contextmanager
+from copy import deepcopy
+from patroni.postgresql.callback_executor import CallbackExecutor
+from patroni.postgresql.bootstrap import Bootstrap
+from patroni.postgresql.cancellable import CancellableSubprocess
+from patroni.postgresql.config import ConfigHandler
+from patroni.postgresql.connection import Connection, get_connection_cursor
+from patroni.postgresql.misc import parse_history, postgres_major_version_to_int
+from patroni.postgresql.postmaster import PostmasterProcess
+from patroni.postgresql.slots import SlotsHandler
+from patroni.exceptions import PostgresConnectionException
+from patroni.utils import Retry, RetryFailedError, polling_loop
+from patroni.dcs import slot_name_from_member_name, RemoteMember
+from threading import current_thread, Lock
+
+
+logger = logging.getLogger(__name__)
+
+ACTION_ON_START = "on_start"
+ACTION_ON_STOP = "on_stop"
+ACTION_ON_RESTART = "on_restart"
+ACTION_ON_RELOAD = "on_reload"
+ACTION_ON_ROLE_CHANGE = "on_role_change"
+ACTION_NOOP = "noop"
+
+STATE_RUNNING = 'running'
+STATE_REJECT = 'rejecting connections'
+STATE_NO_RESPONSE = 'not responding'
+STATE_UNKNOWN = 'unknown'
+
+STOP_POLLING_INTERVAL = 1
+
+cluster_info_query = ("SELECT CASE WHEN pg_catalog.pg_is_in_recovery() THEN 0 "
+                      "ELSE ('x' || pg_catalog.substr(pg_catalog.pg_{0}file_name("
+                      "pg_catalog.pg_current_{0}_{1}()), 1, 8))::bit(32)::int END, "
+                      "CASE WHEN pg_catalog.pg_is_in_recovery() THEN GREATEST("
+                      " pg_catalog.pg_{0}_{1}_diff(COALESCE("
+                      "pg_catalog.pg_last_{0}_receive_{1}(), '0/0'), '0/0')::bigint,"
+                      " pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_last_{0}_replay_{1}(), '0/0')::bigint)"
+                      "ELSE pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_current_{0}_{1}(), '0/0')::bigint END")
+
+
+@contextmanager
+def null_context():
+    yield
+
+
+class Postgresql(object):
+
+    def __init__(self, config):
+        self.name = config['name']
+        self.scope = config['scope']
+        self._data_dir = config['data_dir']
+        self._database = config.get('database', 'postgres')
+        self._version_file = os.path.join(self._data_dir, 'PG_VERSION')
+        self._major_version = self.get_major_version()
+
+        self._state_lock = Lock()
+        self.set_state('stopped')
+
+        self._pending_restart = False
+        self._connection = Connection()
+        self.config = ConfigHandler(self, config)
+
+        self._bin_dir = config.get('bin_dir') or ''
+        self.bootstrap = Bootstrap(self)
+        self.bootstrapping = False
+        self.__thread_ident = current_thread().ident
+
+        self.slots_handler = SlotsHandler(self)
+
+        self._pgpass = config.get('pgpass') or os.path.join(os.path.expanduser('~'), 'pgpass')
+        self._callback_executor = CallbackExecutor()
+        self.__cb_called = False
+        self.__cb_pending = None
+
+        self.cancellable = CancellableSubprocess()
+
+        self._sysid = None
+        self.retry = Retry(max_tries=-1, deadline=config['retry_timeout']/2.0, max_delay=1,
+                           retry_exceptions=PostgresConnectionException)
+
+        # Retry 'pg_is_in_recovery()' only once
+        self._is_leader_retry = Retry(max_tries=1, deadline=config['retry_timeout']/2.0, max_delay=1,
+                                      retry_exceptions=PostgresConnectionException)
+
+        self._role_lock = Lock()
+        self.set_role(self.get_postgres_role_from_data_directory())
+        self._state_entry_timestamp = None
+
+        self._cluster_info_state = {}
+        self._cached_replica_timeline = None
+
+        # Last known running process
+        self._postmaster_proc = None
+
+        if self.is_running():
+            self.set_state('running')
+            self.set_role('master' if self.is_leader() else 'replica')
+            self.config.write_postgresql_conf()  # we are "joining" already running postgres
+            hba_saved = self.config.replace_pg_hba()
+            ident_saved = self.config.replace_pg_ident()
+            if hba_saved or ident_saved:
+                self.reload()
+        elif self.role == 'master':
+            self.set_role('demoted')
+
+    @property
+    def create_replica_methods(self):
+        return self.config.get('create_replica_methods', []) or self.config.get('create_replica_method', [])
+
+    @property
+    def major_version(self):
+        return self._major_version
+
+    @property
+    def database(self):
+        return self._database
+
+    @property
+    def data_dir(self):
+        return self._data_dir
+
+    @property
+    def callback(self):
+        return self.config.get('callbacks') or {}
+
+    @property
+    def wal_name(self):
+        return 'wal' if self._major_version >= 100000 else 'xlog'
+
+    @property
+    def lsn_name(self):
+        return 'lsn' if self._major_version >= 100000 else 'location'
+
+    def _version_file_exists(self):
+        return not self.data_directory_empty() and os.path.isfile(self._version_file)
+
+    def get_major_version(self):
+        if self._version_file_exists():
+            try:
+                with open(self._version_file) as f:
+                    return postgres_major_version_to_int(f.read().strip())
+            except Exception:
+                logger.exception('Failed to read PG_VERSION from %s', self._data_dir)
+        return 0
+
+    def pgcommand(self, cmd):
+        """Returns path to the specified PostgreSQL command"""
+        return os.path.join(self._bin_dir, cmd)
+
+    def pg_ctl(self, cmd, *args, **kwargs):
+        """Builds and executes pg_ctl command
+
+        :returns: `!True` when return_code == 0, otherwise `!False`"""
+
+        pg_ctl = [self.pgcommand('pg_ctl'), cmd]
+        return subprocess.call(pg_ctl + ['-D', self._data_dir] + list(args), **kwargs) == 0
+
+    def pg_isready(self):
+        """Runs pg_isready to see if PostgreSQL is accepting connections.
+
+        :returns: 'ok' if PostgreSQL is up, 'reject' if starting up, 'no_resopnse' if not up."""
+
+        r = self.config.local_connect_kwargs
+        cmd = [self.pgcommand('pg_isready'), '-p', r['port'], '-d', self._database]
+
+        # Host is not set if we are connecting via default unix socket
+        if 'host' in r:
+            cmd.extend(['-h', r['host']])
+
+        # We only need the username because pg_isready does not try to authenticate
+        if 'user' in r:
+            cmd.extend(['-U', r['user']])
+
+        ret = subprocess.call(cmd)
+        return_codes = {0: STATE_RUNNING,
+                        1: STATE_REJECT,
+                        2: STATE_NO_RESPONSE,
+                        3: STATE_UNKNOWN}
+        return return_codes.get(ret, STATE_UNKNOWN)
+
+    def reload_config(self, config):
+        self.config.reload_config(config)
+        self._is_leader_retry.deadline = self.retry.deadline = config['retry_timeout']/2.0
+
+    @property
+    def pending_restart(self):
+        return self._pending_restart
+
+    def set_pending_restart(self, value):
+        self._pending_restart = value
+
+    @property
+    def sysid(self):
+        if not self._sysid and not self.bootstrapping:
+            data = self.controldata()
+            self._sysid = data.get('Database system identifier', "")
+        return self._sysid
+
+    def get_postgres_role_from_data_directory(self):
+        if self.data_directory_empty():
+            return 'uninitialized'
+        elif self.config.recovery_conf_exists():
+            return 'replica'
+        else:
+            return 'master'
+
+    @property
+    def server_version(self):
+        return self._connection.server_version
+
+    def connection(self):
+        return self._connection.get()
+
+    def set_connection_kwargs(self, kwargs):
+        self._connection.set_conn_kwargs(kwargs)
+
+    def _query(self, sql, *params):
+        """We are always using the same cursor, therefore this method is not thread-safe!!!
+        You can call it from different threads only if you are holding explicit `AsyncExecutor` lock,
+        because the main thread is always holding this lock when running HA cycle."""
+        cursor = None
+        try:
+            cursor = self._connection.cursor()
+            cursor.execute(sql, params)
+            return cursor
+        except psycopg2.Error as e:
+            if cursor and cursor.connection.closed == 0:
+                # When connected via unix socket, psycopg2 can't recoginze 'connection lost'
+                # and leaves `_cursor_holder.connection.closed == 0`, but psycopg2.OperationalError
+                # is still raised (what is correct). It doesn't make sense to continiue with existing
+                # connection and we will close it, to avoid its reuse by the `cursor` method.
+                if isinstance(e, psycopg2.OperationalError):
+                    self._connection.close()
+                else:
+                    raise e
+            if self.state == 'restarting':
+                raise RetryFailedError('cluster is being restarted')
+            raise PostgresConnectionException('connection problems')
+
+    def query(self, sql, *args, **kwargs):
+        if not kwargs.get('retry', True):
+            return self._query(sql, *args)
+        try:
+            return self.retry(self._query, sql, *args)
+        except RetryFailedError as e:
+            raise PostgresConnectionException(str(e))
+
+    def data_directory_empty(self):
+        return not os.path.exists(self._data_dir) or os.listdir(self._data_dir) == []
+
+    def write_pgpass(self, record):
+        if 'user' not in record or 'password' not in record:
+            return os.environ.copy()
+
+        with open(self._pgpass, 'w') as f:
+            if os.name != 'nt':
+                os.fchmod(f.fileno(), 0o600)
+            f.write('{host}:{port}:*:{user}:{password}\n'.format(**record))
+
+        env = os.environ.copy()
+        env['PGPASSFILE'] = self._pgpass
+        return env
+
+    def replica_method_options(self, method):
+        return deepcopy(self.config.get(method, {}))
+
+    def replica_method_can_work_without_replication_connection(self, method):
+        return method != 'basebackup' and self.replica_method_options(method).get('no_master')
+
+    def can_create_replica_without_replication_connection(self, replica_methods=None):
+        """ go through the replication methods to see if there are ones
+            that does not require a working replication connection.
+        """
+        if replica_methods is None:
+            replica_methods = self.create_replica_methods
+        return any(self.replica_method_can_work_without_replication_connection(m) for m in replica_methods)
+
+    def reset_cluster_info_state(self):
+        self._cluster_info_state = {}
+
+    def _cluster_info_state_get(self, name):
+        if not self._cluster_info_state:
+            stmt = cluster_info_query.format(self.wal_name, self.lsn_name)
+            try:
+                result = self._is_leader_retry(self._query, stmt).fetchone()
+                self._cluster_info_state = dict(zip(['timeline', 'wal_position'], result))
+            except RetryFailedError as e:  # SELECT failed two times
+                self._cluster_info_state = {'error': str(e)}
+                if not self.is_starting() and self.pg_isready() == STATE_REJECT:
+                    self.set_state('starting')
+
+        if 'error' in self._cluster_info_state:
+            raise PostgresConnectionException(self._cluster_info_state['error'])
+
+        return self._cluster_info_state.get(name)
+
+    def is_leader(self):
+        return bool(self._cluster_info_state_get('timeline'))
+
+    def is_running(self):
+        """Returns PostmasterProcess if one is running on the data directory or None. If most recently seen process
+        is running updates the cached process based on pid file."""
+        if self._postmaster_proc:
+            if self._postmaster_proc.is_running():
+                return self._postmaster_proc
+            self._postmaster_proc = None
+
+        # we noticed that postgres was restarted, force syncing of replication
+        self.slots_handler.schedule()
+
+        self._postmaster_proc = PostmasterProcess.from_pidfile(self._data_dir)
+        return self._postmaster_proc
+
+    @property
+    def cb_called(self):
+        return self.__cb_called
+
+    def call_nowait(self, cb_name):
+        """ pick a callback command and call it without waiting for it to finish """
+        if self.bootstrapping:
+            return
+        if cb_name in (ACTION_ON_START, ACTION_ON_STOP, ACTION_ON_RESTART, ACTION_ON_ROLE_CHANGE):
+            self.__cb_called = True
+
+        if self.callback and cb_name in self.callback:
+            cmd = self.callback[cb_name]
+            try:
+                cmd = shlex.split(self.callback[cb_name]) + [cb_name, self.role, self.scope]
+                self._callback_executor.call(cmd)
+            except Exception:
+                logger.exception('callback %s %s %s %s failed', cmd, cb_name, self.role, self.scope)
+
+    @property
+    def role(self):
+        with self._role_lock:
+            return self._role
+
+    def set_role(self, value):
+        with self._role_lock:
+            self._role = value
+
+    @property
+    def state(self):
+        with self._state_lock:
+            return self._state
+
+    def set_state(self, value):
+        with self._state_lock:
+            self._state = value
+            self._state_entry_timestamp = time.time()
+
+    def time_in_state(self):
+        return time.time() - self._state_entry_timestamp
+
+    def is_starting(self):
+        return self.state == 'starting'
+
+    def wait_for_port_open(self, postmaster, timeout):
+        """Waits until PostgreSQL opens ports."""
+        for _ in polling_loop(timeout):
+            if self.cancellable.is_cancelled:
+                return False
+
+            if not postmaster.is_running():
+                logger.error('postmaster is not running')
+                self.set_state('start failed')
+                return False
+
+            isready = self.pg_isready()
+            if isready != STATE_NO_RESPONSE:
+                if isready not in [STATE_REJECT, STATE_RUNNING]:
+                    logger.warning("Can't determine PostgreSQL startup status, assuming running")
+                return True
+
+        logger.warning("Timed out waiting for PostgreSQL to start")
+        return False
+
+    def start(self, timeout=None, task=None, block_callbacks=False, role=None):
+        """Start PostgreSQL
+
+        Waits for postmaster to open ports or terminate so pg_isready can be used to check startup completion
+        or failure.
+
+        :returns: True if start was initiated and postmaster ports are open, False if start failed"""
+        # make sure we close all connections established against
+        # the former node, otherwise, we might get a stalled one
+        # after kill -9, which would report incorrect data to
+        # patroni.
+        self._connection.close()
+
+        if self.is_running():
+            logger.error('Cannot start PostgreSQL because one is already running.')
+            self.set_state('starting')
+            return True
+
+        if not block_callbacks:
+            self.__cb_pending = ACTION_ON_START
+
+        self.set_role(role or self.get_postgres_role_from_data_directory())
+
+        self.set_state('starting')
+        self._pending_restart = False
+
+        configuration = self.config.effective_configuration
+        self.config.write_postgresql_conf(configuration)
+        self.config.resolve_connection_addresses()
+        self.config.replace_pg_hba()
+        self.config.replace_pg_ident()
+
+        options = ['--{0}={1}'.format(p, configuration[p]) for p in self.config.CMDLINE_OPTIONS
+                   if p in configuration and p != 'wal_keep_segments']
+
+        if self.cancellable.is_cancelled:
+            return False
+
+        with task or null_context():
+            if task and task.is_cancelled:
+                logger.info("PostgreSQL start cancelled.")
+                return False
+
+            self._postmaster_proc = PostmasterProcess.start(self.pgcommand('postgres'),
+                                                            self._data_dir,
+                                                            self.config.postgresql_conf,
+                                                            options)
+
+            if task:
+                task.complete(self._postmaster_proc)
+
+        start_timeout = timeout
+        if not start_timeout:
+            try:
+                start_timeout = float(self.config.get('pg_ctl_timeout', 60))
+            except ValueError:
+                start_timeout = 60
+
+        # We want postmaster to open ports before we continue
+        if not self._postmaster_proc or not self.wait_for_port_open(self._postmaster_proc, start_timeout):
+            return False
+
+        ret = self.wait_for_startup(start_timeout)
+        if ret is not None:
+            return ret
+        elif timeout is not None:
+            return False
+        else:
+            return None
+
+    def checkpoint(self, connect_kwargs=None):
+        check_not_is_in_recovery = connect_kwargs is not None
+        connect_kwargs = connect_kwargs or self.config.local_connect_kwargs
+        for p in ['connect_timeout', 'options']:
+            connect_kwargs.pop(p, None)
+        try:
+            with get_connection_cursor(**connect_kwargs) as cur:
+                cur.execute("SET statement_timeout = 0")
+                if check_not_is_in_recovery:
+                    cur.execute('SELECT pg_catalog.pg_is_in_recovery()')
+                    if cur.fetchone()[0]:
+                        return 'is_in_recovery=true'
+                return cur.execute('CHECKPOINT')
+        except psycopg2.Error:
+            logger.exception('Exception during CHECKPOINT')
+            return 'not accessible or not healty'
+
+    def stop(self, mode='fast', block_callbacks=False, checkpoint=None, on_safepoint=None):
+        """Stop PostgreSQL
+
+        Supports a callback when a safepoint is reached. A safepoint is when no user backend can return a successful
+        commit to users. Currently this means we wait for user backends to close. But in the future alternate mechanisms
+        could be added.
+
+        :param on_safepoint: This callback is called when no user backends are running.
+        """
+        if checkpoint is None:
+            checkpoint = False if mode == 'immediate' else True
+
+        success, pg_signaled = self._do_stop(mode, block_callbacks, checkpoint, on_safepoint)
+        if success:
+            # block_callbacks is used during restart to avoid
+            # running start/stop callbacks in addition to restart ones
+            if not block_callbacks:
+                self.set_state('stopped')
+                if pg_signaled:
+                    self.call_nowait(ACTION_ON_STOP)
+        else:
+            logger.warning('pg_ctl stop failed')
+            self.set_state('stop failed')
+        return success
+
+    def _do_stop(self, mode, block_callbacks, checkpoint, on_safepoint):
+        postmaster = self.is_running()
+        if not postmaster:
+            if on_safepoint:
+                on_safepoint()
+            return True, False
+
+        if checkpoint and not self.is_starting():
+            self.checkpoint()
+
+        if not block_callbacks:
+            self.set_state('stopping')
+
+        # Send signal to postmaster to stop
+        success = postmaster.signal_stop(mode)
+        if success is not None:
+            if success and on_safepoint:
+                on_safepoint()
+            return success, True
+
+        # We can skip safepoint detection if we don't have a callback
+        if on_safepoint:
+            # Wait for our connection to terminate so we can be sure that no new connections are being initiated
+            self._wait_for_connection_close(postmaster)
+            postmaster.wait_for_user_backends_to_close()
+            on_safepoint()
+
+        postmaster.wait()
+
+        return True, True
+
+    @staticmethod
+    def terminate_starting_postmaster(postmaster):
+        """Terminates a postmaster that has not yet opened ports or possibly even written a pid file. Blocks
+        until the process goes away."""
+        postmaster.signal_stop('immediate')
+        postmaster.wait()
+
+    def _wait_for_connection_close(self, postmaster):
+        try:
+            with self.connection().cursor() as cur:
+                while postmaster.is_running():  # Need a timeout here?
+                    cur.execute("SELECT 1")
+                    time.sleep(STOP_POLLING_INTERVAL)
+        except psycopg2.Error:
+            pass
+
+    def reload(self):
+        ret = self.pg_ctl('reload')
+        if ret:
+            self.call_nowait(ACTION_ON_RELOAD)
+        return ret
+
+    def check_for_startup(self):
+        """Checks PostgreSQL status and returns if PostgreSQL is in the middle of startup."""
+        return self.is_starting() and not self.check_startup_state_changed()
+
+    def check_startup_state_changed(self):
+        """Checks if PostgreSQL has completed starting up or failed or still starting.
+
+        Should only be called when state == 'starting'
+
+        :returns: True if state was changed from 'starting'
+        """
+        ready = self.pg_isready()
+
+        if ready == STATE_REJECT:
+            return False
+        elif ready == STATE_NO_RESPONSE:
+            self.set_state('start failed')
+            self.slots_handler.schedule(False)  # TODO: can remove this?
+            self.config.save_configuration_files(True)  # TODO: maybe remove this?
+            return True
+        else:
+            if ready != STATE_RUNNING:
+                # Bad configuration or unexpected OS error. No idea of PostgreSQL status.
+                # Let the main loop of run cycle clean up the mess.
+                logger.warning("%s status returned from pg_isready",
+                               "Unknown" if ready == STATE_UNKNOWN else "Invalid")
+            self.set_state('running')
+            self.slots_handler.schedule()
+            self.config.save_configuration_files(True)
+            # TODO: __cb_pending can be None here after PostgreSQL restarts on its own. Do we want to call the callback?
+            # Previously we didn't even notice.
+            action = self.__cb_pending or ACTION_ON_START
+            self.call_nowait(action)
+            self.__cb_pending = None
+
+            return True
+
+    def wait_for_startup(self, timeout=None):
+        """Waits for PostgreSQL startup to complete or fail.
+
+        :returns: True if start was successful, False otherwise"""
+        if not self.is_starting():
+            # Should not happen
+            logger.warning("wait_for_startup() called when not in starting state")
+
+        while not self.check_startup_state_changed():
+            if self.cancellable.is_cancelled or timeout and self.time_in_state() > timeout:
+                return None
+            time.sleep(1)
+
+        return self.state == 'running'
+
+    def restart(self, timeout=None, task=None, block_callbacks=False, role=None):
+        """Restarts PostgreSQL.
+
+        When timeout parameter is set the call will block either until PostgreSQL has started, failed to start or
+        timeout arrives.
+
+        :returns: True when restart was successful and timeout did not expire when waiting.
+        """
+        self.set_state('restarting')
+        if not block_callbacks:
+            self.__cb_pending = ACTION_ON_RESTART
+        ret = self.stop(block_callbacks=True) and self.start(timeout, task, True, role)
+        if not ret and not self.is_starting():
+            self.set_state('restart failed ({0})'.format(self.state))
+        return ret
+
+    def is_healthy(self):
+        if not self.is_running():
+            logger.warning('Postgresql is not running.')
+            return False
+        return True
+
+    def controldata(self):
+        """ return the contents of pg_controldata, or non-True value if pg_controldata call failed """
+        result = {}
+        # Don't try to call pg_controldata during backup restore
+        if self._version_file_exists() and self.state != 'creating replica':
+            try:
+                env = {'LANG': 'C', 'LC_ALL': 'C', 'PATH': os.getenv('PATH')}
+                if os.getenv('SYSTEMROOT') is not None:
+                    env['SYSTEMROOT'] = os.getenv('SYSTEMROOT')
+                data = subprocess.check_output([self.pgcommand('pg_controldata'), self._data_dir], env=env)
+                if data:
+                    data = data.decode('utf-8').splitlines()
+                    # pg_controldata output depends on major verion. Some of parameters are prefixed by 'Current '
+                    result = {l.split(':')[0].replace('Current ', '', 1): l.split(':', 1)[1].strip() for l in data
+                              if l and ':' in l}
+            except subprocess.CalledProcessError:
+                logger.exception("Error when calling pg_controldata")
+        return result
+
+    @contextmanager
+    def get_replication_connection_cursor(self, host='localhost', port=5432, database=None, **kwargs):
+        replication = self.config.replication
+        with get_connection_cursor(host=host, port=int(port), database=database or self._database, replication=1,
+                                   user=replication['username'], password=replication.get('password'),
+                                   connect_timeout=3, options='-c statement_timeout=2000') as cur:
+            yield cur
+
+    def get_local_timeline_lsn_from_replication_connection(self):
+        timeline = lsn = None
+        try:
+            with self.get_replication_connection_cursor(**self.config.local_replication_address) as cur:
+                cur.execute('IDENTIFY_SYSTEM')
+                timeline, lsn = cur.fetchone()[1:3]
+        except Exception:
+            logger.exception('Can not fetch local timeline and lsn from replication connection')
+        return timeline, lsn
+
+    def get_replica_timeline(self):
+        return self.get_local_timeline_lsn_from_replication_connection()[0]
+
+    def replica_cached_timeline(self, master_timeline):
+        if not self._cached_replica_timeline or not master_timeline or self._cached_replica_timeline != master_timeline:
+            self._cached_replica_timeline = self.get_replica_timeline()
+        return self._cached_replica_timeline
+
+    def get_master_timeline(self):
+        return self._cluster_info_state_get('timeline')
+
+    def get_history(self, timeline):
+        history_path = 'pg_{0}/{1:08X}.history'.format(self.wal_name, timeline)
+        try:
+            cursor = self._connection.cursor()
+            cursor.execute('SELECT isdir, modification FROM pg_catalog.pg_stat_file(%s)', (history_path,))
+            isdir, modification = cursor.fetchone()
+            if not isdir:
+                cursor.execute('SELECT pg_catalog.pg_read_file(%s)', (history_path,))
+                history = list(parse_history(cursor.fetchone()[0]))
+                if history[-1][0] == timeline - 1:
+                    history[-1].append(modification.isoformat())
+                return history
+        except Exception:
+            logger.exception('Failed to read and parse %s', (history_path,))
+
+    def follow(self, member, role='replica', timeout=None):
+        is_remote_master = isinstance(member, RemoteMember)
+        no_replication_slot = is_remote_master and member.no_replication_slot
+        restore_command = is_remote_master and member.restore_command
+        min_apply_delay = is_remote_master and member.recovery_min_apply_delay
+        archive_cleanup = is_remote_master and member.archive_cleanup_command
+
+        primary_conninfo = self.config.primary_conninfo(member)
+        change_role = self.cb_called and (self.role in ('master', 'demoted') or
+                                          not {'standby_leader', 'replica'} - {self.role, role})
+
+        recovery_params = self.config.get('recovery_conf', {}).copy()
+        recovery_params.update({'standby_mode': 'on', 'recovery_target_timeline': 'latest'})
+        if primary_conninfo:
+            recovery_params['primary_conninfo'] = primary_conninfo
+        if self.slots_handler.use_slots and not no_replication_slot:
+            required_name = is_remote_master and member.data.get('primary_slot_name')
+            name = required_name or slot_name_from_member_name(self.name)
+            recovery_params['primary_slot_name'] = name
+        if restore_command:
+            recovery_params['restore_command'] = restore_command
+        if min_apply_delay:
+            recovery_params['recovery_min_apply_delay'] = min_apply_delay
+        if archive_cleanup:
+            recovery_params['archive_cleanup_command'] = archive_cleanup
+
+        self.config.write_recovery_conf(recovery_params)
+
+        # When we demoting the master or standby_leader to replica or promoting replica to a standby_leader
+        # and we know for sure that postgres was already running before, we will only execute on_role_change
+        # callback and prevent execution of on_restart/on_start callback.
+        # If the role remains the same (replica or standby_leader), we will execute on_start or on_restart
+        if change_role:
+            self.__cb_pending = ACTION_NOOP
+
+        if self.is_running():
+            self.restart(block_callbacks=change_role, role=role)
+        else:
+            self.start(timeout=timeout, block_callbacks=change_role, role=role)
+
+        if change_role:
+            # TODO: postpone this until start completes, or maybe do even earlier
+            self.call_nowait(ACTION_ON_ROLE_CHANGE)
+        return True
+
+    def _wait_promote(self, wait_seconds):
+        for _ in polling_loop(wait_seconds):
+            data = self.controldata()
+            if data.get('Database cluster state') == 'in production':
+                return True
+
+    def promote(self, wait_seconds, on_success=None, access_is_restricted=False):
+        if self.role == 'master':
+            return True
+        ret = self.pg_ctl('promote', '-W')
+        if ret:
+            self.set_role('master')
+            if on_success is not None:
+                on_success()
+            if not access_is_restricted:
+                self.call_nowait(ACTION_ON_ROLE_CHANGE)
+            ret = self._wait_promote(wait_seconds)
+        return ret
+
+    def timeline_wal_position(self):
+        # This method could be called from different threads (simultaneously with some other `_query` calls).
+        # If it is called not from main thread we will create a new cursor to execute statement.
+        if current_thread().ident == self.__thread_ident:
+            return self._cluster_info_state_get('timeline'), self._cluster_info_state_get('wal_position')
+
+        with self.connection().cursor() as cursor:
+            cursor.execute(cluster_info_query.format(self.wal_name, self.lsn_name))
+            return cursor.fetchone()[:2]
+
+    def postmaster_start_time(self):
+        try:
+            cursor = self.query("SELECT pg_catalog.to_char(pg_catalog.pg_postmaster_start_time(),"
+                                " 'YYYY-MM-DD HH24:MI:SS.MS TZ')")
+            return cursor.fetchone()[0]
+        except psycopg2.Error:
+            return None
+
+    def last_operation(self):
+        return str(self._cluster_info_state_get('wal_position'))
+
+    def configure_server_parameters(self):
+        self._major_version = self.get_major_version()
+        self.config.setup_server_parameters()
+        return True
+
+    def move_data_directory(self):
+        if os.path.isdir(self._data_dir) and not self.is_running():
+            try:
+                new_name = '{0}_{1}'.format(self._data_dir, time.strftime('%Y-%m-%d-%H-%M-%S'))
+                logger.info('renaming data directory to %s', new_name)
+                os.rename(self._data_dir, new_name)
+            except OSError:
+                logger.exception("Could not rename data directory %s", self._data_dir)
+
+    def remove_data_directory(self):
+        self.set_role('uninitialized')
+        logger.info('Removing data directory: %s', self._data_dir)
+        try:
+            if os.path.islink(self._data_dir):
+                os.unlink(self._data_dir)
+            elif not os.path.exists(self._data_dir):
+                return
+            elif os.path.isfile(self._data_dir):
+                os.remove(self._data_dir)
+            elif os.path.isdir(self._data_dir):
+
+                # let's see if pg_xlog|pg_wal is a symlink, in this case we
+                # should clean the target
+                for pg_wal_dir in ('pg_xlog', 'pg_wal'):
+                    pg_wal_path = os.path.join(self._data_dir, pg_wal_dir)
+                    if os.path.exists(pg_wal_path) and os.path.islink(pg_wal_path):
+                        pg_wal_realpath = os.path.realpath(pg_wal_path)
+                        logger.info('Removing WAL directory: %s', pg_wal_realpath)
+                        shutil.rmtree(pg_wal_realpath)
+
+                shutil.rmtree(self._data_dir)
+        except (IOError, OSError):
+            logger.exception('Could not remove data directory %s', self._data_dir)
+            self.move_data_directory()
+
+    def pick_synchronous_standby(self, cluster):
+        """Finds the best candidate to be the synchronous standby.
+
+        Current synchronous standby is always preferred, unless it has disconnected or does not want to be a
+        synchronous standby any longer.
+
+        :returns tuple of candidate name or None, and bool showing if the member is the active synchronous standby.
+        """
+        current = cluster.sync.sync_standby
+        current = current.lower() if current else current
+        members = {m.name.lower(): m for m in cluster.members}
+        candidates = []
+        # Pick candidates based on who has flushed WAL farthest.
+        # TODO: for synchronous_commit = remote_write we actually want to order on write_location
+        for app_name, state, sync_state in self.query(
+                "SELECT pg_catalog.lower(application_name), state, sync_state"
+                " FROM pg_catalog.pg_stat_replication"
+                " ORDER BY flush_{0} DESC".format(self.lsn_name)):
+            member = members.get(app_name)
+            if state != 'streaming' or not member or member.tags.get('nosync', False):
+                continue
+            if sync_state == 'sync':
+                return app_name, True
+            if sync_state == 'potential' and app_name == current:
+                # Prefer current even if not the best one any more to avoid indecisivness and spurious swaps.
+                return current, False
+            if sync_state in ('async', 'potential'):
+                candidates.append(app_name)
+
+        if candidates:
+            return candidates[0], False
+        return None, False
+
+    def read_postmaster_opts(self):
+        """returns the list of option names/values from postgres.opts, Empty dict if read failed or no file"""
+        result = {}
+        try:
+            with open(os.path.join(self._data_dir, 'postmaster.opts')) as f:
+                data = f.read()
+                for opt in data.split('" "'):
+                    if '=' in opt and opt.startswith('--'):
+                        name, val = opt.split('=', 1)
+                        result[name.strip('-')] = val.rstrip('"\n')
+        except IOError:
+            logger.exception('Error when reading postmaster.opts')
+        return result
+
+    def single_user_mode(self, command=None, options=None):
+        """run a given command in a single-user mode. If the command is empty - then just start and stop"""
+        cmd = [self.pgcommand('postgres'), '--single', '-D', self._data_dir]
+        for opt, val in sorted((options or {}).items()):
+            cmd.extend(['-c', '{0}={1}'.format(opt, val)])
+        # need a database name to connect
+        cmd.append(self._database)
+        return self.cancellable.call(cmd, communicate_input=command)
+
+    def cleanup_archive_status(self):
+        status_dir = os.path.join(self._data_dir, 'pg_' + self.wal_name, 'archive_status')
+        try:
+            for f in os.listdir(status_dir):
+                path = os.path.join(status_dir, f)
+                try:
+                    if os.path.islink(path):
+                        os.unlink(path)
+                    elif os.path.isfile(path):
+                        os.remove(path)
+                except OSError:
+                    logger.exception('Unable to remove %s', path)
+        except OSError:
+            logger.exception('Unable to list %s', status_dir)
+
+    def fix_cluster_state(self):
+        self.cleanup_archive_status()
+
+        # Start in a single user mode and stop to produce a clean shutdown
+        opts = self.read_postmaster_opts()
+        opts.update({'archive_mode': 'on', 'archive_command': 'false'})
+        self.config.remove_recovery_conf()
+        return self.single_user_mode(options=opts) == 0 or None
+
+    def schedule_sanity_checks_after_pause(self):
+        """
+            After coming out of pause we have to:
+            1. sync replication slots, because it might happen that slots were removed
+            2. get new 'Database system identifier' to make sure that it wasn't changed
+        """
+        self.slots_handler.schedule()
+        self._sysid = None

patroni/postgresql/bootstrap.py

@@ -0,0 +1,385 @@
+import logging
+import os
+import shlex
+import tempfile
+import time
+
+from patroni.dcs import RemoteMember
+from patroni.utils import deep_compare, uri
+from six import string_types
+from six.moves.urllib.parse import quote_plus
+
+logger = logging.getLogger(__name__)
+
+
+class Bootstrap(object):
+
+    def __init__(self, postgresql):
+        self._postgresql = postgresql
+        self._running_custom_bootstrap = False
+
+    @property
+    def running_custom_bootstrap(self):
+        return self._running_custom_bootstrap
+
+    @property
+    def keep_existing_recovery_conf(self):
+        return self._running_custom_bootstrap and self._keep_existing_recovery_conf
+
+    @staticmethod
+    def process_user_options(tool, options, not_allowed_options, error_handler):
+        user_options = []
+
+        def option_is_allowed(name):
+            ret = name not in not_allowed_options
+            if not ret:
+                error_handler('{0} option for {1} is not allowed'.format(name, tool))
+            return ret
+
+        if isinstance(options, dict):
+            for k, v in options.items():
+                if k and v:
+                    user_options.append('--{0}={1}'.format(k, v))
+        elif isinstance(options, list):
+            for opt in options:
+                if isinstance(opt, string_types) and option_is_allowed(opt):
+                    user_options.append('--{0}'.format(opt))
+                elif isinstance(opt, dict):
+                    keys = list(opt.keys())
+                    if len(keys) != 1 or not isinstance(opt[keys[0]], string_types) or not option_is_allowed(keys[0]):
+                        error_handler('Error when parsing {0} key-value option {1}: only one key-value is allowed'
+                                      ' and value should be a string'.format(tool, opt[keys[0]]))
+                    user_options.append('--{0}={1}'.format(keys[0], opt[keys[0]]))
+                else:
+                    error_handler('Error when parsing {0} option {1}: value should be string value'
+                                  ' or a single key-value pair'.format(tool, opt))
+        else:
+            error_handler('{0} options must be list ot dict'.format(tool))
+        return user_options
+
+    def _initdb(self, config):
+        self._postgresql.set_state('initalizing new cluster')
+        not_allowed_options = ('pgdata', 'nosync', 'pwfile', 'sync-only', 'version')
+
+        def error_handler(e):
+            raise Exception(e)
+
+        options = self.process_user_options('initdb', config or [], not_allowed_options, error_handler)
+        pwfile = None
+
+        if self._postgresql.config.superuser:
+            if 'username' in self._postgresql.config.superuser:
+                options.append('--username={0}'.format(self._postgresql.config.superuser['username']))
+            if 'password' in self._postgresql.config.superuser:
+                (fd, pwfile) = tempfile.mkstemp()
+                os.write(fd, self._postgresql.config.superuser['password'].encode('utf-8'))
+                os.close(fd)
+                options.append('--pwfile={0}'.format(pwfile))
+        options = ['-o', ' '.join(options)] if options else []
+
+        ret = self._postgresql.pg_ctl('initdb', *options)
+        if pwfile:
+            os.remove(pwfile)
+        if ret:
+            self._postgresql.configure_server_parameters()
+        else:
+            self._postgresql.set_state('initdb failed')
+        return ret
+
+    def _post_restore(self):
+        self._postgresql.config.restore_configuration_files()
+        self._postgresql.configure_server_parameters()
+
+        # make sure there is no trigger file or postgres will be automatically promoted
+        trigger_file = 'promote_trigger_file' if self._postgresql.major_version >= 120000 else 'trigger_file'
+        trigger_file = self._postgresql.config.get('recovery_conf', {}).get(trigger_file) or 'promote'
+        trigger_file = os.path.abspath(os.path.join(self._postgresql.data_dir, trigger_file))
+        if os.path.exists(trigger_file):
+            os.unlink(trigger_file)
+
+    def _custom_bootstrap(self, config):
+        self._postgresql.set_state('running custom bootstrap script')
+        params = ['--scope=' + self._postgresql.scope, '--datadir=' + self._postgresql.data_dir]
+        try:
+            logger.info('Running custom bootstrap script: %s', config['command'])
+            if self._postgresql.cancellable.call(shlex.split(config['command']) + params) != 0:
+                self._postgresql.set_state('custom bootstrap failed')
+                return False
+        except Exception:
+            logger.exception('Exception during custom bootstrap')
+            return False
+        self._post_restore()
+
+        if 'recovery_conf' in config:
+            self._postgresql.config.write_recovery_conf(config['recovery_conf'])
+        elif not self.keep_existing_recovery_conf:
+            self._postgresql.config.remove_recovery_conf()
+        return True
+
+    def call_post_bootstrap(self, config):
+        """
+        runs a script after initdb or custom bootstrap script is called and waits until completion.
+        """
+        cmd = config.get('post_bootstrap') or config.get('post_init')
+        if cmd:
+            r = self._postgresql.config.local_connect_kwargs
+
+            if 'host' in r:
+                # '/tmp' => '%2Ftmp' for unix socket path
+                host = quote_plus(r['host']) if r['host'].startswith('/') else r['host']
+            else:
+                host = ''
+
+                # https://www.postgresql.org/docs/current/static/libpq-pgpass.html
+                # A host name of localhost matches both TCP (host name localhost) and Unix domain socket
+                # (pghost empty or the default socket directory) connections coming from the local machine.
+                r['host'] = 'localhost'  # set it to localhost to write into pgpass
+
+            if 'user' in r:
+                user = r['user']
+            else:
+                user = ''
+                if 'password' in r:
+                    import getpass
+                    r.setdefault('user', os.environ.get('PGUSER', getpass.getuser()))
+
+            connstring = uri('postgres', (host, r['port']), r['database'], user)
+            env = self._postgresql.write_pgpass(r) if 'password' in r else None
+
+            try:
+                ret = self._postgresql.cancellable.call(shlex.split(cmd) + [connstring], env=env)
+            except OSError:
+                logger.error('post_init script %s failed', cmd)
+                return False
+            if ret != 0:
+                logger.error('post_init script %s returned non-zero code %d', cmd, ret)
+                return False
+        return True
+
+    def create_replica(self, clone_member):
+        """
+            create the replica according to the replica_method
+            defined by the user.  this is a list, so we need to
+            loop through all methods the user supplies
+        """
+
+        self._postgresql.set_state('creating replica')
+        self._postgresql.schedule_sanity_checks_after_pause()
+
+        is_remote_master = isinstance(clone_member, RemoteMember)
+
+        # get list of replica methods either from clone member or from
+        # the config. If there is no configuration key, or no value is
+        # specified, use basebackup
+        replica_methods = (clone_member.create_replica_methods if is_remote_master
+                           else self._postgresql.create_replica_methods) or ['basebackup']
+
+        if clone_member and clone_member.conn_url:
+            r = clone_member.conn_kwargs(self._postgresql.config.replication)
+            connstring = uri('postgres', (r['host'], r['port']), r['database'], r['user'])
+            # add the credentials to connect to the replica origin to pgpass.
+            env = self._postgresql.write_pgpass(r)
+        else:
+            connstring = ''
+            env = os.environ.copy()
+            # if we don't have any source, leave only replica methods that work without it
+            replica_methods = [r for r in replica_methods
+                               if self._postgresql.replica_method_can_work_without_replication_connection(r)]
+
+        # go through them in priority order
+        ret = 1
+        for replica_method in replica_methods:
+            if self._postgresql.cancellable.is_cancelled:
+                break
+
+            method_config = self._postgresql.replica_method_options(replica_method)
+
+            # if the method is basebackup, then use the built-in
+            if replica_method == "basebackup":
+                ret = self.basebackup(connstring, env, method_config)
+                if ret == 0:
+                    logger.info("replica has been created using basebackup")
+                    # if basebackup succeeds, exit with success
+                    break
+            else:
+                if not self._postgresql.data_directory_empty():
+                    if method_config.get('keep_data', False):
+                        logger.info('Leaving data directory uncleaned')
+                    else:
+                        self._postgresql.remove_data_directory()
+
+                cmd = replica_method
+                # user-defined method; check for configuration
+                # not required, actually
+                if method_config:
+                    # look to see if the user has supplied a full command path
+                    # if not, use the method name as the command
+                    cmd = method_config.pop('command', cmd)
+
+                # add the default parameters
+                if not method_config.get('no_params', False):
+                    method_config.update({"scope": self._postgresql.scope,
+                                          "role": "replica",
+                                          "datadir": self._postgresql.data_dir,
+                                          "connstring": connstring})
+                else:
+                    for param in ('no_params', 'no_master', 'keep_data'):
+                        method_config.pop(param, None)
+                params = ["--{0}={1}".format(arg, val) for arg, val in method_config.items()]
+                try:
+                    # call script with the full set of parameters
+                    ret = self._postgresql.cancellable.call(shlex.split(cmd) + params, env=env)
+                    # if we succeeded, stop
+                    if ret == 0:
+                        logger.info('replica has been created using %s', replica_method)
+                        break
+                    else:
+                        logger.error('Error creating replica using method %s: %s exited with code=%s',
+                                     replica_method, cmd, ret)
+                except Exception:
+                    logger.exception('Error creating replica using method %s', replica_method)
+                    ret = 1
+
+        self._postgresql.set_state('stopped')
+        return ret
+
+    def basebackup(self, conn_url, env, options):
+        # creates a replica data dir using pg_basebackup.
+        # this is the default, built-in create_replica_methods
+        # tries twice, then returns failure (as 1)
+        # uses "stream" as the xlog-method to avoid sync issues
+        # supports additional user-supplied options, those are not validated
+        maxfailures = 2
+        ret = 1
+        not_allowed_options = ('pgdata', 'format', 'wal-method', 'xlog-method', 'gzip',
+                               'version', 'compress', 'dbname', 'host', 'port', 'username', 'password')
+        user_options = self.process_user_options('basebackup', options, not_allowed_options, logger.error)
+
+        for bbfailures in range(0, maxfailures):
+            if self._postgresql.cancellable.is_cancelled:
+                break
+            if not self._postgresql.data_directory_empty():
+                self._postgresql.remove_data_directory()
+            try:
+                ret = self._postgresql.cancellable.call([self._postgresql.pgcommand('pg_basebackup'),
+                                                         '--pgdata=' + self._postgresql.data_dir, '-X', 'stream',
+                                                         '--dbname=' + conn_url] + user_options, env=env)
+                if ret == 0:
+                    break
+                else:
+                    logger.error('Error when fetching backup: pg_basebackup exited with code=%s', ret)
+
+            except Exception as e:
+                logger.error('Error when fetching backup with pg_basebackup: %s', e)
+
+            if bbfailures < maxfailures - 1:
+                logger.warning('Trying again in 5 seconds')
+                time.sleep(5)
+
+        return ret
+
+    def clone(self, clone_member):
+        """
+             - initialize the replica from an existing member (master or replica)
+             - initialize the replica using the replica creation method that
+               works without the replication connection (i.e. restore from on-disk
+               base backup)
+        """
+
+        ret = self.create_replica(clone_member) == 0
+        if ret:
+            self._post_restore()
+        return ret
+
+    def bootstrap(self, config):
+        """ Initialize a new node from scratch and start it. """
+        pg_hba = config.get('pg_hba', [])
+        method = config.get('method') or 'initdb'
+        if method != 'initdb' and method in config and 'command' in config[method]:
+            self._keep_existing_recovery_conf = config[method].get('keep_existing_recovery_conf')
+            self._running_custom_bootstrap = True
+            do_initialize = self._custom_bootstrap
+        else:
+            method = 'initdb'
+            do_initialize = self._initdb
+        return do_initialize(config.get(method)) and self._postgresql.config.append_pg_hba(pg_hba) \
+            and self._postgresql.config.save_configuration_files() and self._postgresql.start()
+
+    def create_or_update_role(self, name, password, options):
+        options = list(map(str.upper, options))
+        if 'NOLOGIN' not in options and 'LOGIN' not in options:
+            options.append('LOGIN')
+
+        params = [name]
+        if password:
+            options.extend(['PASSWORD', '%s'])
+            params.extend([password, password])
+
+        sql = """DO $$
+BEGIN
+    SET local synchronous_commit = 'local';
+    PERFORM * FROM pg_authid WHERE rolname = %s;
+    IF FOUND THEN
+        ALTER ROLE "{0}" WITH {1};
+    ELSE
+        CREATE ROLE "{0}" WITH {1};
+    END IF;
+END;$$""".format(name, ' '.join(options))
+        self._postgresql.query(sql, *params)
+
+    def post_bootstrap(self, config, task):
+        try:
+            postgresql = self._postgresql
+            superuser = postgresql.config.superuser
+            if 'username' in superuser and 'password' in superuser:
+                self.create_or_update_role(superuser['username'], superuser['password'], ['SUPERUSER'])
+
+            task.complete(self.call_post_bootstrap(config))
+            if task.result:
+                replication = postgresql.config.replication
+                self.create_or_update_role(replication['username'], replication.get('password'), ['REPLICATION'])
+
+                rewind = postgresql.config.rewind_credentials
+                if not deep_compare(rewind, superuser):
+                    self.create_or_update_role(rewind['username'], rewind.get('password'), [])
+                    for f in ('pg_ls_dir(text, boolean, boolean)', 'pg_stat_file(text, boolean)',
+                              'pg_read_binary_file(text)', 'pg_read_binary_file(text, bigint, bigint, boolean)'):
+                        sql = """DO $$
+BEGIN
+    SET local synchronous_commit = 'local';
+    GRANT EXECUTE ON function pg_catalog.{0} TO "{1}";
+END;$$""".format(f, rewind['username'])
+                        postgresql.query(sql)
+
+                for name, value in (config.get('users') or {}).items():
+                    if all(name != a.get('username') for a in (superuser, replication, rewind)):
+                        self.create_or_update_role(name, value.get('password'), value.get('options', []))
+
+                # We were doing a custom bootstrap instead of running initdb, therefore we opened trust
+                # access from certain addresses to be able to reach cluster and change password
+                if self._running_custom_bootstrap:
+                    self._running_custom_bootstrap = False
+                    # If we don't have custom configuration for pg_hba.conf we need to restore original file
+                    if not postgresql.config.get('pg_hba'):
+                        os.unlink(postgresql.config.pg_hba_conf)
+                        postgresql.config.restore_configuration_files()
+                    postgresql.config.write_postgresql_conf()
+                    postgresql.config.replace_pg_ident()
+
+                    # at this point there should be no recovery.conf
+                    postgresql.config.remove_recovery_conf()
+
+                    if postgresql.config.hba_file and postgresql.config.hba_file != postgresql.config.pg_hba_conf:
+                        postgresql.restart()
+                    else:
+                        postgresql.config.replace_pg_hba()
+                        if postgresql.pending_restart:
+                            postgresql.restart()
+                        else:
+                            postgresql.reload()
+                            time.sleep(1)  # give a time to postgres to "reload" configuration files
+                            postgresql.connection().close()  # close connection to reconnect with a new password
+        except Exception:
+            logger.exception('post_bootstrap')
+            task.complete(False)
+        return task.result

patroni/postgresql/callback_executor.py

@@ -19,8 +19,11 @@ class CallbackExecutor(Thread):
     def call(self, cmd):
         with self._lock:
             if self._process and self._process.poll() is None:
-                self._process.kill()
+                try:
-                logger.warning('Killed the old callback process because it was still running: %s', self._cmd)
+                    self._process.kill()
+                    logger.warning('Killed the old callback process because it was still running: %s', self._cmd)
+                except OSError:
+                    logger.exception('Failed to kill the old callback')
         self._cmd = cmd
         self._callback_event.set()
 

patroni/postgresql/cancellable.py

@@ -0,0 +1,76 @@
+import logging
+import os
+import subprocess
+
+from patroni.exceptions import PostgresException
+from patroni.utils import polling_loop
+from six import string_types
+from threading import Lock
+
+logger = logging.getLogger(__name__)
+
+
+class CancellableSubprocess(object):
+
+    def __init__(self):
+        self._is_cancelled = False
+        self._process = None
+        self._lock = Lock()
+
+    def call(self, *args, **kwargs):
+        for s in ('stdin', 'stdout', 'stderr'):
+            kwargs.pop(s, None)
+
+        communicate_input = 'communicate_input' in kwargs
+        if communicate_input:
+            input_data = kwargs.pop('communicate_input', None)
+            if not isinstance(input_data, string_types):
+                input_data = ''
+            if input_data and input_data[-1] != '\n':
+                input_data += '\n'
+            kwargs['stdin'] = subprocess.PIPE
+            kwargs['stdout'] = open(os.devnull, 'w')
+            kwargs['stderr'] = subprocess.STDOUT
+
+        try:
+            with self._lock:
+                if self._is_cancelled:
+                    raise PostgresException('cancelled')
+
+                self._is_cancelled = False
+                self._process = subprocess.Popen(*args, **kwargs)
+
+            if communicate_input:
+                if input_data:
+                    self._process.communicate(input_data)
+                self._process.stdin.close()
+
+            return self._process.wait()
+        finally:
+            with self._lock:
+                self._process = None
+
+    def reset_is_cancelled(self):
+        with self._lock:
+            self._is_cancelled = False
+
+    @property
+    def is_cancelled(self):
+        with self._lock:
+            return self._is_cancelled
+
+    def cancel(self):
+        with self._lock:
+            self._is_cancelled = True
+            if self._process is None or self._process.returncode is not None:
+                return
+            self._process.terminate()
+
+        for _ in polling_loop(10):
+            with self._lock:
+                if self._process is None or self._process.returncode is not None:
+                    return
+
+        with self._lock:
+            if self._process is not None and self._process.returncode is None:
+                self._process.kill()

patroni/postgresql/config.py

@@ -0,0 +1,763 @@
+import logging
+import os
+import re
+import shutil
+import socket
+import stat
+
+from requests.structures import CaseInsensitiveDict
+from six.moves.urllib_parse import urlparse, parse_qsl, unquote
+
+from ..utils import compare_values, parse_bool, parse_int, split_host_port, uri
+
+logger = logging.getLogger(__name__)
+
+SYNC_STANDBY_NAME_RE = re.compile(r'^[A-Za-z_][A-Za-z_0-9\$]*$')
+PARAMETER_RE = re.compile(r'([a-z_]+)\s*=\s*')
+
+
+def quote_ident(value):
+    """Very simplified version of quote_ident"""
+    return value if SYNC_STANDBY_NAME_RE.match(value) else '"' + value + '"'
+
+
+def conninfo_uri_parse(dsn):
+    ret = {}
+    r = urlparse(dsn)
+    if r.username:
+        ret['user'] = r.username
+    if r.password:
+        ret['password'] = r.password
+    if r.path[1:]:
+        ret['dbname'] = r.path[1:]
+    hosts = []
+    ports = []
+    for netloc in r.netloc.split('@')[-1].split(','):
+        host = port = None
+        if '[' in netloc and ']' in netloc:
+            host = netloc.split(']')[0][1:]
+        tmp = netloc.split(':', 1)
+        if host is None:
+            host = tmp[0]
+        if len(tmp) == 2:
+            host, port = tmp
+        if host is not None:
+            hosts.append(host)
+        if port is not None:
+            ports.append(port)
+    if hosts:
+        ret['host'] = ','.join(hosts)
+    if ports:
+        ret['port'] = ','.join(ports)
+    ret = {name: unquote(value) for name, value in ret.items()}
+    ret.update({name: value for name, value in parse_qsl(r.query)})
+    if ret.get('ssl') == 'true':
+        del ret['ssl']
+        ret['sslmode'] = 'require'
+    return ret
+
+
+def read_param_value(value, is_quoted=False):
+    length = len(value)
+    ret = ''
+    i = 0
+    while i < length:
+        if is_quoted:
+            if value[i] == "'":
+                return ret, i
+        elif value[i].isspace():
+            break
+        if value[i] == '\\':
+            i += 1
+            if i >= length:
+                break
+        ret += value[i]
+        i += 1
+    return (None, None) if is_quoted else (ret, i)
+
+
+def conninfo_parse(dsn):
+    ret = {}
+    length = len(dsn)
+    i = 0
+    while i < length:
+        if dsn[i].isspace():
+            i += 1
+            continue
+
+        param_match = PARAMETER_RE.match(dsn[i:])
+        if not param_match:
+            return
+
+        param = param_match.group(1)
+        i += param_match.end()
+
+        if i >= length:
+            return
+
+        is_quoted = dsn[i] == "'"
+        i += int(is_quoted)
+        value, end = read_param_value(dsn[i:], is_quoted)
+        if value is None:
+            return
+        i += end + int(is_quoted)
+        ret[param] = value
+    return ret
+
+
+def parse_dsn(value):
+    """
+    Very simple equivalent of `psycopg2.extensions.parse_dsn` introduced in 2.7.0.
+    We are not using psycopg2 function in order to remain compatible with 2.5.4+.
+    There is one minor difference though, this function removes `dbname` from the result
+    and sets the sslmode` to `prefer` if it is not present in the connection string.
+    This is necessary to simplify comparison of the old and the new values.
+
+    >>> r = parse_dsn('postgresql://u%2Fse:pass@:%2f123,[%2Fhost2]/db%2Fsdf?application_name=mya%2Fpp&ssl=true')
+    >>> r == {'application_name': 'mya/pp', 'host': ',/host2', 'sslmode': 'require',\
+              'password': 'pass', 'port': '/123', 'user': 'u/se'}
+    True
+    >>> r = parse_dsn(" host = 'host' dbname = db\\ name requiressl=1 ")
+    >>> r == {'host': 'host', 'sslmode': 'require'}
+    True
+    >>> parse_dsn('requiressl = 0\\\\') == {'sslmode': 'prefer'}
+    True
+    >>> parse_dsn("host=a foo = '") is None
+    True
+    >>> parse_dsn("host=a foo = ") is None
+    True
+    >>> parse_dsn("1") is None
+    True
+    """
+    if value.startswith('postgres://') or value.startswith('postgresql://'):
+        ret = conninfo_uri_parse(value)
+    else:
+        ret = conninfo_parse(value)
+
+    if ret:
+        if 'sslmode' not in ret:  # allow sslmode to take precedence over requiressl
+            requiressl = ret.pop('requiressl', None)
+            if requiressl == '1':
+                ret['sslmode'] = 'require'
+            elif requiressl is not None:
+                ret['sslmode'] = 'prefer'
+            ret.setdefault('sslmode', 'prefer')
+        if 'dbname' in ret:
+            del ret['dbname']
+    return ret
+
+
+def mtime(filename):
+    try:
+        return os.stat(filename).st_mtime
+    except OSError:
+        return None
+
+
+class ConfigHandler(object):
+
+    # List of parameters which must be always passed to postmaster as command line options
+    # to make it not possible to change them with 'ALTER SYSTEM'.
+    # Some of these parameters have sane default value assigned and Patroni doesn't allow
+    # to decrease this value. E.g. 'wal_level' can't be lower then 'hot_standby' and so on.
+    # These parameters could be changed only globally, i.e. via DCS.
+    # P.S. 'listen_addresses' and 'port' are added here just for convenience, to mark them
+    # as a parameters which should always be passed through command line.
+    #
+    # Format:
+    #  key - parameter name
+    #  value - tuple(default_value, check_function, min_version)
+    #    default_value -- some sane default value
+    #    check_function -- if the new value is not correct must return `!False`
+    #    min_version -- major version of PostgreSQL when parameter was introduced
+    CMDLINE_OPTIONS = CaseInsensitiveDict({
+        'listen_addresses': (None, lambda _: False, 90100),
+        'port': (None, lambda _: False, 90100),
+        'cluster_name': (None, lambda _: False, 90500),
+        'wal_level': ('hot_standby', lambda v: v.lower() in ('hot_standby', 'replica', 'logical'), 90100),
+        'hot_standby': ('on', lambda _: False, 90100),
+        'max_connections': (100, lambda v: int(v) >= 25, 90100),
+        'max_wal_senders': (10, lambda v: int(v) >= 3, 90100),
+        'wal_keep_segments': (8, lambda v: int(v) >= 1, 90100),
+        'max_prepared_transactions': (0, lambda v: int(v) >= 0, 90100),
+        'max_locks_per_transaction': (64, lambda v: int(v) >= 32, 90100),
+        'track_commit_timestamp': ('off', lambda v: parse_bool(v) is not None, 90500),
+        'max_replication_slots': (10, lambda v: int(v) >= 4, 90400),
+        'max_worker_processes': (8, lambda v: int(v) >= 2, 90400),
+        'wal_log_hints': ('on', lambda _: False, 90400)
+    })
+
+    _RECOVERY_PARAMETERS = {
+        'archive_cleanup_command',
+        'restore_command',
+        'recovery_end_command',
+        'recovery_target',
+        'recovery_target_name',
+        'recovery_target_time',
+        'recovery_target_xid',
+        'recovery_target_lsn',
+        'recovery_target_inclusive',
+        'recovery_target_timeline',
+        'recovery_target_action',
+        'recovery_min_apply_delay',
+        'primary_conninfo',
+        'primary_slot_name',
+        'promote_trigger_file',
+        'trigger_file'
+    }
+
+    _CONFIG_WARNING_HEADER = '# Do not edit this file manually!\n# It will be overwritten by Patroni!\n'
+
+    def __init__(self, postgresql, config):
+        self._postgresql = postgresql
+        self._config_dir = os.path.abspath(config.get('config_dir') or postgresql.data_dir)
+        config_base_name = config.get('config_base_name', 'postgresql')
+        self._postgresql_conf = os.path.join(self._config_dir, config_base_name + '.conf')
+        self._postgresql_conf_mtime = None
+        self._postgresql_base_conf_name = config_base_name + '.base.conf'
+        self._postgresql_base_conf = os.path.join(self._config_dir, self._postgresql_base_conf_name)
+        self._pg_hba_conf = os.path.join(self._config_dir, 'pg_hba.conf')
+        self._pg_ident_conf = os.path.join(self._config_dir, 'pg_ident.conf')
+        self._recovery_conf = os.path.join(postgresql.data_dir, 'recovery.conf')
+        self._recovery_conf_mtime = None
+        self._recovery_signal = os.path.join(postgresql.data_dir, 'recovery.signal')
+        self._standby_signal = os.path.join(postgresql.data_dir, 'standby.signal')
+        self._auto_conf = os.path.join(postgresql.data_dir, 'postgresql.auto.conf')
+        self._auto_conf_mtime = None
+        self._synchronous_standby_names = None
+        self._postmaster_ctime = None
+        self._primary_conninfo = None
+        self._config = {}
+        self._recovery_params = {}
+        self.reload_config(config)
+
+    def setup_server_parameters(self):
+        self._server_parameters = self.get_server_parameters(self._config)
+        self._adjust_recovery_parameters()
+
+    @property
+    def _configuration_to_save(self):
+        configuration = [os.path.basename(self._postgresql_conf)]
+        if 'custom_conf' not in self._config:
+            configuration.append(os.path.basename(self._postgresql_base_conf_name))
+        if not self.hba_file:
+            configuration.append('pg_hba.conf')
+        if not self._server_parameters.get('ident_file'):
+            configuration.append('pg_ident.conf')
+        return configuration
+
+    def save_configuration_files(self, check_custom_bootstrap=False):
+        """
+            copy postgresql.conf to postgresql.conf.backup to be able to retrive configuration files
+            - originally stored as symlinks, those are normally skipped by pg_basebackup
+            - in case of WAL-E basebackup (see http://comments.gmane.org/gmane.comp.db.postgresql.wal-e/239)
+        """
+        if not (check_custom_bootstrap and self._postgresql.bootstrap.running_custom_bootstrap):
+            try:
+                for f in self._configuration_to_save:
+                    config_file = os.path.join(self._config_dir, f)
+                    backup_file = os.path.join(self._postgresql.data_dir, f + '.backup')
+                    if os.path.isfile(config_file):
+                        shutil.copy(config_file, backup_file)
+            except IOError:
+                logger.exception('unable to create backup copies of configuration files')
+        return True
+
+    def restore_configuration_files(self):
+        """ restore a previously saved postgresql.conf """
+        try:
+            for f in self._configuration_to_save:
+                config_file = os.path.join(self._config_dir, f)
+                backup_file = os.path.join(self._postgresql.data_dir, f + '.backup')
+                if not os.path.isfile(config_file):
+                    if os.path.isfile(backup_file):
+                        shutil.copy(backup_file, config_file)
+                    # Previously we didn't backup pg_ident.conf, if file is missing just create empty
+                    elif f == 'pg_ident.conf':
+                        open(config_file, 'w').close()
+        except IOError:
+            logger.exception('unable to restore configuration files from backup')
+
+    def write_postgresql_conf(self, configuration=None):
+        # rename the original configuration if it is necessary
+        if 'custom_conf' not in self._config and not os.path.exists(self._postgresql_base_conf):
+            os.rename(self._postgresql_conf, self._postgresql_base_conf)
+
+        with open(self._postgresql_conf, 'w') as f:
+            os.chmod(self._postgresql_conf, stat.S_IWRITE | stat.S_IREAD)
+            f.write(self._CONFIG_WARNING_HEADER)
+            f.write("include '{0}'\n\n".format(self._config.get('custom_conf') or self._postgresql_base_conf_name))
+            for name, value in sorted((configuration or self._server_parameters).items()):
+                if (not self._postgresql.bootstrap.running_custom_bootstrap or name != 'hba_file') \
+                        and name not in self._RECOVERY_PARAMETERS:
+                    f.write("{0} = '{1}'\n".format(name, value))
+            # when we are doing custom bootstrap we assume that we don't know superuser password
+            # and in order to be able to change it, we are opening trust access from a certain address
+            # therefore we need to make sure that hba_file is not overriden
+            # after changing superuser password we will "revert" all these "changes"
+            if self._postgresql.bootstrap.running_custom_bootstrap or 'hba_file' not in self._server_parameters:
+                f.write("hba_file = '{0}'\n".format(self._pg_hba_conf.replace('\\', '\\\\')))
+            if 'ident_file' not in self._server_parameters:
+                f.write("ident_file = '{0}'\n".format(self._pg_ident_conf.replace('\\', '\\\\')))
+
+            if self._postgresql.major_version >= 120000:
+                if self._recovery_params:
+                    f.write('\n# recovery.conf\n')
+                    for name, value in sorted(self._recovery_params.items()):
+                        f.write("{0} = '{1}'\n".format(name, value))
+
+                if not self._postgresql.bootstrap.keep_existing_recovery_conf:
+                    self._sanitize_auto_conf()
+
+    def append_pg_hba(self, config):
+        if not self.hba_file and not self._config.get('pg_hba'):
+            with open(self._pg_hba_conf, 'a') as f:
+                f.write('\n{}\n'.format('\n'.join(config)))
+        return True
+
+    def replace_pg_hba(self):
+        """
+        Replace pg_hba.conf content in the PGDATA if hba_file is not defined in the
+        `postgresql.parameters` and pg_hba is defined in `postgresql` configuration section.
+
+        :returns: True if pg_hba.conf was rewritten.
+        """
+
+        # when we are doing custom bootstrap we assume that we don't know superuser password
+        # and in order to be able to change it, we are opening trust access from a certain address
+        if self._postgresql.bootstrap.running_custom_bootstrap:
+            addresses = {'': 'local'}
+            if 'host' in self.local_replication_address and not self.local_replication_address['host'].startswith('/'):
+                addresses.update({sa[0] + '/32': 'host' for _, _, _, _, sa in socket.getaddrinfo(
+                                  self.local_replication_address['host'], self.local_replication_address['port'],
+                                  0, socket.SOCK_STREAM, socket.IPPROTO_TCP)})
+
+            with open(self._pg_hba_conf, 'w') as f:
+                f.write(self._CONFIG_WARNING_HEADER)
+                for address, t in addresses.items():
+                    f.write((
+                        '{0}\treplication\t{1}\t{3}\ttrust\n'
+                        '{0}\tall\t{2}\t{3}\ttrust\n'
+                    ).format(t, self.replication['username'], self._superuser.get('username') or 'all', address))
+        elif not self.hba_file and self._config.get('pg_hba'):
+            with open(self._pg_hba_conf, 'w') as f:
+                f.write(self._CONFIG_WARNING_HEADER)
+                for line in self._config['pg_hba']:
+                    f.write('{0}\n'.format(line))
+            return True
+
+    def replace_pg_ident(self):
+        """
+        Replace pg_ident.conf content in the PGDATA if ident_file is not defined in the
+        `postgresql.parameters` and pg_ident is defined in the `postgresql` section.
+
+        :returns: True if pg_ident.conf was rewritten.
+        """
+
+        if not self._server_parameters.get('ident_file') and self._config.get('pg_ident'):
+            with open(self._pg_ident_conf, 'w') as f:
+                f.write(self._CONFIG_WARNING_HEADER)
+                for line in self._config['pg_ident']:
+                    f.write('{0}\n'.format(line))
+            return True
+
+    def primary_conninfo_params(self, member):
+        name = self._postgresql.name
+        if not (member and member.conn_url) or member.name == name:
+            return None
+        ret = member.conn_kwargs(self.replication)
+        ret.update(application_name=name, sslmode='prefer')
+        if self._krbsrvname:
+            ret['krbsrvname'] = self._krbsrvname
+        if 'database' in ret:
+            del ret['database']
+        return ret
+
+    def primary_conninfo(self, member):
+        r = self.primary_conninfo_params(member)
+        if not r:
+            return None
+        keywords = 'user password host port sslmode application_name krbsrvname'.split()
+        return ' '.join('{0}={{{0}}}'.format(kw) for kw in keywords if r.get(kw)).format(**r)
+
+    def recovery_conf_exists(self):
+        if self._postgresql.major_version >= 120000:
+            return os.path.exists(self._standby_signal) or os.path.exists(self._recovery_signal)
+        return os.path.exists(self._recovery_conf)
+
+    def _read_primary_conninfo(self):
+        pg_conf_mtime = mtime(self._postgresql_conf)
+        auto_conf_mtime = mtime(self._auto_conf)
+        postmaster_ctime = self._postgresql.is_running()
+        if postmaster_ctime:
+            postmaster_ctime = postmaster_ctime.create_time()
+
+        if self._postgresql_conf_mtime == pg_conf_mtime and self._auto_conf_mtime == auto_conf_mtime \
+                and self._postmaster_ctime == postmaster_ctime:
+            return None, False
+
+        try:
+            primary_conninfo = self._postgresql.query('SHOW primary_conninfo').fetchone()[0]
+            self._postgresql_conf_mtime = pg_conf_mtime
+            self._auto_conf_mtime = auto_conf_mtime
+            self._postmaster_ctime = postmaster_ctime
+        except Exception:
+            primary_conninfo = None
+        return primary_conninfo, True
+
+    def _read_primary_conninfo_pre_v12(self):
+        recovery_conf_mtime = mtime(self._recovery_conf)
+        if recovery_conf_mtime == self._recovery_conf_mtime:
+            return None, False
+
+        primary_conninfo = ''
+        with open(self._recovery_conf, 'r') as f:
+            for line in f:
+                match = PARAMETER_RE.match(line)
+                if match and match.group(1) == 'primary_conninfo':
+                    i = match.end()
+                    if i < len(line):
+                        is_quoted = line[i] == "'"
+                        i += int(is_quoted)
+                        primary_conninfo, _ = read_param_value(line[i:], is_quoted)
+            self._recovery_conf_mtime = recovery_conf_mtime
+        return primary_conninfo, True
+
+    def check_recovery_conf(self, member):  # Name is confusing. In fact it checks the value of primary_conninfo
+        # TODO: recovery.conf could be stale, would be nice to detect that.
+        if self._postgresql.major_version >= 120000:
+            if not os.path.exists(self._standby_signal):
+                return False
+
+            _read_primary_conninfo = self._read_primary_conninfo
+        else:
+            if not self.recovery_conf_exists():
+                return False
+
+            _read_primary_conninfo = self._read_primary_conninfo_pre_v12
+
+        primary_conninfo, updated = _read_primary_conninfo()
+        # updated indicates that mtime of postgresql.conf, postgresql.auto.conf, or recovery.conf was changed
+        # and the primary_conninfo value was read either from config or from the database connection.
+        if updated:
+            # primary_conninfo is one of:
+            # - None (exception or unparsable config)
+            # - '' (not in config)
+            # - or the actual dsn value
+            self._primary_conninfo = primary_conninfo
+            if primary_conninfo:
+                # We will cache parsed value until the next config change.
+                self._primary_conninfo = parse_dsn(primary_conninfo)
+                # If we failed to parse non-empty connection string this indicates that config if broken.
+                if not self._primary_conninfo:
+                    return False
+            elif primary_conninfo is not None:
+                self._primary_conninfo = {}
+            else:  # primary_conninfo is None, config is probably broken
+                return False
+
+        wanted_primary_conninfo = self.primary_conninfo_params(member)
+        # first we will cover corner cases, when we are replicating from somewhere while shouldn't
+        # or there is no primary_conninfo but we should replicate from some specific node.
+        if not wanted_primary_conninfo:
+            return not self._primary_conninfo
+        elif not self._primary_conninfo:
+            return False
+
+        return all(self._primary_conninfo.get(p) == str(v) for p, v in wanted_primary_conninfo.items())
+
+    @staticmethod
+    def _remove_file_if_exists(name):
+        if os.path.isfile(name) or os.path.islink(name):
+            os.unlink(name)
+
+    def write_recovery_conf(self, recovery_params):
+        if self._postgresql.major_version >= 120000:
+            if parse_bool(recovery_params.pop('standby_mode', None)):
+                open(self._standby_signal, 'w').close()
+            else:
+                self._remove_file_if_exists(self._standby_signal)
+                open(self._recovery_signal, 'w').close()
+            self._recovery_params = recovery_params
+        else:
+            with open(self._recovery_conf, 'w') as f:
+                os.chmod(self._recovery_conf, stat.S_IWRITE | stat.S_IREAD)
+                for name, value in recovery_params.items():
+                    f.write("{0} = '{1}'\n".format(name, value))
+
+    def remove_recovery_conf(self):
+        for name in (self._recovery_conf, self._standby_signal, self._recovery_signal):
+            self._remove_file_if_exists(name)
+        self._recovery_params = {}
+
+    def _sanitize_auto_conf(self):
+        overwrite = False
+        lines = []
+
+        if os.path.exists(self._auto_conf):
+            try:
+                with open(self._auto_conf) as f:
+                    for raw_line in f:
+                        line = raw_line.strip()
+                        match = PARAMETER_RE.match(line)
+                        if match and match.group(1).lower() in self._RECOVERY_PARAMETERS:
+                            overwrite = True
+                        else:
+                            lines.append(raw_line)
+            except Exception:
+                logger.info('Failed to read %s', self._auto_conf)
+
+        if overwrite:
+            try:
+                with open(self._auto_conf, 'w') as f:
+                    for raw_line in lines:
+                        f.write(raw_line)
+            except Exception:
+                logger.exception('Failed to remove some unwanted parameters from %s', self._auto_conf)
+
+    def _adjust_recovery_parameters(self):
+        # It is not strictly necessary, but we can make patroni configs crossi-compatible with all postgres versions.
+        recovery_conf = {n: v for n, v in self._server_parameters.items() if n.lower() in self._RECOVERY_PARAMETERS}
+        if recovery_conf:
+            self._config['recovery_conf'] = recovery_conf
+
+        if self.get('recovery_conf'):
+            good_name, bad_name = 'trigger_file', 'promote_trigger_file'
+            if self._postgresql.major_version >= 120000:
+                good_name, bad_name = bad_name, good_name
+
+            value = self._config['recovery_conf'].pop(bad_name, None)
+            if good_name not in self._config['recovery_conf'] and value:
+                self._config['recovery_conf'][good_name] = value
+
+    def get_server_parameters(self, config):
+        parameters = config['parameters'].copy()
+        listen_addresses, port = split_host_port(config['listen'], 5432)
+        parameters.update(cluster_name=self._postgresql.scope, listen_addresses=listen_addresses, port=str(port))
+        if config.get('synchronous_mode', False):
+            if self._synchronous_standby_names is None:
+                if config.get('synchronous_mode_strict', False):
+                    parameters['synchronous_standby_names'] = '*'
+                else:
+                    parameters.pop('synchronous_standby_names', None)
+            else:
+                parameters['synchronous_standby_names'] = self._synchronous_standby_names
+        if self._postgresql.major_version >= 90600 and parameters['wal_level'] == 'hot_standby':
+            parameters['wal_level'] = 'replica'
+        ret = CaseInsensitiveDict({k: v for k, v in parameters.items() if not self._postgresql.major_version or
+                                   self._postgresql.major_version >= self.CMDLINE_OPTIONS.get(k, (0, 1, 90100))[2]})
+        ret.update({k: os.path.join(self._config_dir, ret[k]) for k in ('hba_file', 'ident_file') if k in ret})
+        return ret
+
+    @staticmethod
+    def _get_unix_local_address(unix_socket_directories):
+        for d in unix_socket_directories.split(','):
+            d = d.strip()
+            if d.startswith('/'):  # Only absolute path can be used to connect via unix-socket
+                return d
+        return ''
+
+    def _get_tcp_local_address(self):
+        listen_addresses = self._server_parameters['listen_addresses'].split(',')
+
+        for la in listen_addresses:
+            if la.strip().lower() in ('*', '0.0.0.0', '127.0.0.1', 'localhost'):  # we are listening on '*' or localhost
+                return 'localhost'  # connection via localhost is preferred
+        return listen_addresses[0].strip()  # can't use localhost, take first address from listen_addresses
+
+    @property
+    def local_connect_kwargs(self):
+        ret = self._local_address.copy()
+        ret.update({'database': self._postgresql.database,
+                    'fallback_application_name': 'Patroni',
+                    'connect_timeout': 3,
+                    'options': '-c statement_timeout=2000'})
+        if 'username' in self._superuser:
+            ret['user'] = self._superuser['username']
+        if 'password' in self._superuser:
+            ret['password'] = self._superuser['password']
+        return ret
+
+    def resolve_connection_addresses(self):
+        port = self._server_parameters['port']
+        tcp_local_address = self._get_tcp_local_address()
+
+        local_address = {'port': port}
+        if self._config.get('use_unix_socket'):
+            unix_socket_directories = self._server_parameters.get('unix_socket_directories')
+            if unix_socket_directories is not None:
+                # fallback to tcp if unix_socket_directories is set, but there are no sutable values
+                local_address['host'] = self._get_unix_local_address(unix_socket_directories) or tcp_local_address
+
+            # if unix_socket_directories is not specified, but use_unix_socket is set to true - do our best
+            # to use default value, i.e. don't specify a host neither in connection url nor arguments
+        else:
+            local_address['host'] = tcp_local_address
+
+        self._local_address = local_address
+        self.local_replication_address = {'host': tcp_local_address, 'port': port}
+
+        netloc = self._config.get('connect_address') or tcp_local_address + ':' + port
+        self._postgresql.connection_string = uri('postgres', netloc, self._postgresql.database)
+
+        self._postgresql.set_connection_kwargs(self.local_connect_kwargs)
+
+    def reload_config(self, config):
+        self._superuser = config['authentication'].get('superuser', {})
+        server_parameters = self.get_server_parameters(config)
+
+        conf_changed = hba_changed = ident_changed = local_connection_address_changed = pending_restart = False
+        if self._postgresql.state == 'running':
+            changes = CaseInsensitiveDict({p: v for p, v in server_parameters.items() if '.' not in p})
+            changes.update({p: None for p in self._server_parameters.keys() if not ('.' in p or p in changes)})
+            if changes:
+                # XXX: query can raise an exception
+                for r in self._postgresql.query(('SELECT name, setting, unit, vartype, context '
+                                                 + 'FROM pg_catalog.pg_settings ' +
+                                                 ' WHERE pg_catalog.lower(name) IN ('
+                                                 + ', '.join(['%s'] * len(changes)) +
+                                                 ')'), *(k.lower() for k in changes.keys())):
+                    if r[4] != 'internal' and r[0] in changes:
+                        new_value = changes.pop(r[0])
+                        if new_value is None or not compare_values(r[3], r[2], r[1], new_value):
+                            if r[4] == 'postmaster':
+                                pending_restart = True
+                                logger.info('Changed %s from %s to %s (restart required)', r[0], r[1], new_value)
+                                if config.get('use_unix_socket') and r[0] == 'unix_socket_directories'\
+                                        or r[0] in ('listen_addresses', 'port'):
+                                    local_connection_address_changed = True
+                            else:
+                                logger.info('Changed %s from %s to %s', r[0], r[1], new_value)
+                                conf_changed = True
+                for param in changes:
+                    if param in server_parameters:
+                        logger.warning('Removing invalid parameter `%s` from postgresql.parameters', param)
+                        server_parameters.pop(param)
+
+            # Check that user-defined-paramters have changed (parameters with period in name)
+            if not conf_changed:
+                for p, v in server_parameters.items():
+                    if '.' in p and (p not in self._server_parameters or str(v) != str(self._server_parameters[p])):
+                        logger.info('Changed %s from %s to %s', p, self._server_parameters.get(p), v)
+                        conf_changed = True
+                        break
+                if not conf_changed:
+                    for p, v in self._server_parameters.items():
+                        if '.' in p and (p not in server_parameters or str(v) != str(server_parameters[p])):
+                            logger.info('Changed %s from %s to %s', p, v, server_parameters.get(p))
+                            conf_changed = True
+                            break
+
+            if not server_parameters.get('hba_file') and config.get('pg_hba'):
+                hba_changed = self._config.get('pg_hba', []) != config['pg_hba']
+
+            if not server_parameters.get('ident_file') and config.get('pg_ident'):
+                ident_changed = self._config.get('pg_ident', []) != config['pg_ident']
+
+        self._config = config
+        self._postgresql.set_pending_restart(pending_restart)
+        self._server_parameters = server_parameters
+        self._adjust_recovery_parameters()
+        self._connect_address = config.get('connect_address')
+        self._krbsrvname = config.get('krbsrvname')
+
+        # for not so obvious connection attempts that may happen outside of pyscopg2
+        if self._krbsrvname:
+            os.environ['PGKRBSRVNAME'] = self._krbsrvname
+
+        if not local_connection_address_changed:
+            self.resolve_connection_addresses()
+
+        if conf_changed:
+            self.write_postgresql_conf()
+
+        if hba_changed:
+            self.replace_pg_hba()
+
+        if ident_changed:
+            self.replace_pg_ident()
+
+        if conf_changed or hba_changed or ident_changed:
+            logger.info('PostgreSQL configuration items changed, reloading configuration.')
+            self._postgresql.reload()
+        elif not pending_restart:
+            logger.info('No PostgreSQL configuration items changed, nothing to reload.')
+
+    def set_synchronous_standby(self, name):
+        """Sets a node to be synchronous standby and if changed does a reload for PostgreSQL."""
+        if name and name != '*':
+            name = quote_ident(name)
+        if name != self._synchronous_standby_names:
+            if name is None:
+                self._server_parameters.pop('synchronous_standby_names', None)
+            else:
+                self._server_parameters['synchronous_standby_names'] = name
+            self._synchronous_standby_names = name
+            if self._postgresql.state == 'running':
+                self.write_postgresql_conf()
+                self._postgresql.reload()
+
+    @property
+    def effective_configuration(self):
+        """It might happen that the current value of one (or more) below parameters stored in
+        the controldata is higher than the value stored in the global cluster configuration.
+
+        Example: max_connections in global configuration is 100, but in controldata
+        `Current max_connections setting: 200`. If we try to start postgres with
+        max_connections=100, it will immediately exit.
+        As a workaround we will start it with the values from controldata and set `pending_restart`
+        to true as an indicator that current values of parameters are not matching expectations."""
+
+        if self._postgresql.role == 'master':
+            return self._server_parameters
+
+        options_mapping = {
+            'max_connections': 'max_connections setting',
+            'max_prepared_transactions': 'max_prepared_xacts setting',
+            'max_locks_per_transaction': 'max_locks_per_xact setting'
+        }
+
+        if self._postgresql.major_version >= 90400:
+            options_mapping['max_worker_processes'] = 'max_worker_processes setting'
+
+        if self._postgresql.major_version >= 120000:
+            options_mapping['max_wal_senders'] = 'max_wal_senders setting'
+
+        data = self._postgresql.controldata()
+        effective_configuration = self._server_parameters.copy()
+
+        for name, cname in options_mapping.items():
+            value = parse_int(effective_configuration[name])
+            cvalue = parse_int(data[cname])
+            if cvalue > value:
+                effective_configuration[name] = cvalue
+                self._postgresql.set_pending_restart(True)
+        return effective_configuration
+
+    @property
+    def replication(self):
+        return self._config['authentication']['replication']
+
+    @property
+    def superuser(self):
+        return self._superuser
+
+    @property
+    def rewind_credentials(self):
+        return self._config['authentication'].get('rewind', self._superuser) \
+                if self._postgresql.major_version >= 110000 else self._superuser
+
+    @property
+    def hba_file(self):
+        return self._server_parameters.get('hba_file')
+
+    @property
+    def pg_hba_conf(self):
+        return self._pg_hba_conf
+
+    @property
+    def postgresql_conf(self):
+        return self._postgresql_conf
+
+    def get(self, key, default=None):
+        return self._config.get(key, default)

patroni/postgresql/connection.py

@@ -0,0 +1,46 @@
+import logging
+import psycopg2
+
+from contextlib import contextmanager
+from threading import Lock
+
+logger = logging.getLogger(__name__)
+
+
+class Connection(object):
+
+    def __init__(self):
+        self._lock = Lock()
+        self._connection = None
+        self._cursor_holder = None
+
+    def set_conn_kwargs(self, conn_kwargs):
+        self._conn_kwargs = conn_kwargs
+
+    def get(self):
+        with self._lock:
+            if not self._connection or self._connection.closed != 0:
+                self._connection = psycopg2.connect(**self._conn_kwargs)
+                self._connection.autocommit = True
+                self.server_version = self._connection.server_version
+        return self._connection
+
+    def cursor(self):
+        if not self._cursor_holder or self._cursor_holder.closed or self._cursor_holder.connection.closed != 0:
+            logger.info("establishing a new patroni connection to the postgres cluster")
+            self._cursor_holder = self.get().cursor()
+        return self._cursor_holder
+
+    def close(self):
+        if self._connection and self._connection.closed == 0:
+            self._connection.close()
+            logger.info("closed patroni connection to the postgresql cluster")
+        self._cursor_holder = self._connection = None
+
+
+@contextmanager
+def get_connection_cursor(**kwargs):
+    with psycopg2.connect(**kwargs) as conn:
+        conn.autocommit = True
+        with conn.cursor() as cur:
+            yield cur

patroni/postgresql/misc.py

@@ -0,0 +1,70 @@
+import logging
+
+from patroni.exceptions import PostgresException
+
+logger = logging.getLogger(__name__)
+
+
+def postgres_version_to_int(pg_version):
+    """Convert the server_version to integer
+
+    >>> postgres_version_to_int('9.5.3')
+    90503
+    >>> postgres_version_to_int('9.3.13')
+    90313
+    >>> postgres_version_to_int('10.1')
+    100001
+    >>> postgres_version_to_int('10')  # doctest: +IGNORE_EXCEPTION_DETAIL
+    Traceback (most recent call last):
+        ...
+    PostgresException: 'Invalid PostgreSQL version format: X.Y or X.Y.Z is accepted: 10'
+    >>> postgres_version_to_int('9.6')  # doctest: +IGNORE_EXCEPTION_DETAIL
+    Traceback (most recent call last):
+        ...
+    PostgresException: 'Invalid PostgreSQL version format: X.Y or X.Y.Z is accepted: 9.6'
+    >>> postgres_version_to_int('a.b.c')  # doctest: +IGNORE_EXCEPTION_DETAIL
+    Traceback (most recent call last):
+        ...
+    PostgresException: 'Invalid PostgreSQL version: a.b.c'
+    """
+
+    try:
+        components = list(map(int, pg_version.split('.')))
+    except ValueError:
+        raise PostgresException('Invalid PostgreSQL version: {0}'.format(pg_version))
+
+    if len(components) < 2 or len(components) == 2 and components[0] < 10 or len(components) > 3:
+        raise PostgresException('Invalid PostgreSQL version format: X.Y or X.Y.Z is accepted: {0}'.format(pg_version))
+
+    if len(components) == 2:
+        # new style verion numbers, i.e. 10.1 becomes 100001
+        components.insert(1, 0)
+
+    return int(''.join('{0:02d}'.format(c) for c in components))
+
+
+def postgres_major_version_to_int(pg_version):
+    """
+    >>> postgres_major_version_to_int('10')
+    100000
+    >>> postgres_major_version_to_int('9.6')
+    90600
+    """
+    return postgres_version_to_int(pg_version + '.0')
+
+
+def parse_lsn(lsn):
+    t = lsn.split('/')
+    return int(t[0], 16) * 0x100000000 + int(t[1], 16)
+
+
+def parse_history(data):
+    for line in data.split('\n'):
+        values = line.strip().split('\t')
+        if len(values) == 3:
+            try:
+                values[0] = int(values[0])
+                values[1] = parse_lsn(values[1])
+                yield values
+            except (IndexError, ValueError):
+                logger.exception('Exception when parsing timeline history line "%s"', values)

patroni/postgresql/postmaster.py

@@ -0,0 +1,185 @@
+import logging
+import multiprocessing
+import os
+import psutil
+import re
+import signal
+import subprocess
+
+logger = logging.getLogger(__name__)
+
+STOP_SIGNALS = {
+    'smart': signal.SIGTERM,
+    'fast': signal.SIGINT,
+    'immediate': signal.SIGQUIT if os.name != 'nt' else signal.SIGABRT,
+}
+
+
+def pg_ctl_start(conn, cmdline, env):
+    if os.name != 'nt':
+        os.setsid()
+    try:
+        postmaster = subprocess.Popen(cmdline, close_fds=True, env=env)
+        conn.send(postmaster.pid)
+    except Exception:
+        logger.exception('Failed to execute %s', cmdline)
+        conn.send(None)
+    conn.close()
+
+
+class PostmasterProcess(psutil.Process):
+
+    def __init__(self, pid):
+        self.is_single_user = False
+        if pid < 0:
+            pid = -pid
+            self.is_single_user = True
+        super(PostmasterProcess, self).__init__(pid)
+
+    @staticmethod
+    def _read_postmaster_pidfile(data_dir):
+        """Reads and parses postmaster.pid from the data directory
+
+        :returns dictionary of values if successful, empty dictionary otherwise
+        """
+        pid_line_names = ['pid', 'data_dir', 'start_time', 'port', 'socket_dir', 'listen_addr', 'shmem_key']
+        try:
+            with open(os.path.join(data_dir, 'postmaster.pid')) as f:
+                return {name: line.rstrip('\n') for name, line in zip(pid_line_names, f)}
+        except IOError:
+            return {}
+
+    def _is_postmaster_process(self):
+        try:
+            start_time = int(self._postmaster_pid.get('start_time', 0))
+            if start_time and abs(self.create_time() - start_time) > 3:
+                logger.info('Process %s is not postmaster, too much difference between PID file start time %s and '
+                            'process start time %s', self.pid, self.create_time(), start_time)
+                return False
+        except ValueError:
+            logger.warning('Garbage start time value in pid file: %r', self._postmaster_pid.get('start_time'))
+
+        # Extra safety check. The process can't be ourselves, our parent or our direct child.
+        if self.pid == os.getpid() or self.pid == os.getppid() or self.ppid() == os.getpid():
+            logger.info('Patroni (pid=%s, ppid=%s), "fake postmaster" (pid=%s, ppid=%s)',
+                        os.getpid(), os.getppid(), self.pid, self.ppid())
+            return False
+
+        return True
+
+    @classmethod
+    def _from_pidfile(cls, data_dir):
+        postmaster_pid = PostmasterProcess._read_postmaster_pidfile(data_dir)
+        try:
+            pid = int(postmaster_pid.get('pid', 0))
+            if pid:
+                proc = cls(pid)
+                proc._postmaster_pid = postmaster_pid
+                return proc
+        except ValueError:
+            pass
+
+    @staticmethod
+    def from_pidfile(data_dir):
+        try:
+            proc = PostmasterProcess._from_pidfile(data_dir)
+            return proc if proc and proc._is_postmaster_process() else None
+        except psutil.NoSuchProcess:
+            return None
+
+    @classmethod
+    def from_pid(cls, pid):
+        try:
+            return cls(pid)
+        except psutil.NoSuchProcess:
+            return None
+
+    def signal_stop(self, mode):
+        """Signal postmaster process to stop
+
+        :returns None if signaled, True if process is already gone, False if error
+        """
+        if self.is_single_user:
+            logger.warning("Cannot stop server; single-user server is running (PID: {0})".format(self.pid))
+            return False
+        try:
+            self.send_signal(STOP_SIGNALS[mode])
+        except psutil.NoSuchProcess:
+            return True
+        except psutil.AccessDenied as e:
+            logger.warning("Could not send stop signal to PostgreSQL (error: {0})".format(e))
+            return False
+
+        return None
+
+    def wait_for_user_backends_to_close(self):
+        # These regexps are cross checked against versions PostgreSQL 9.1 .. 11
+        aux_proc_re = re.compile("(?:postgres:)( .*:)? (?:(?:archiver|startup|autovacuum launcher|autovacuum worker|"
+                                 "checkpointer|logger|stats collector|wal receiver|wal writer|writer)(?: process  )?|"
+                                 "walreceiver|wal sender process|walsender|walwriter|background writer|"
+                                 "logical replication launcher|logical replication worker for|bgworker:) ")
+
+        try:
+            children = self.children()
+        except psutil.Error:
+            return logger.debug('Failed to get list of postmaster children')
+
+        user_backends = []
+        user_backends_cmdlines = []
+        for child in children:
+            try:
+                cmdline = child.cmdline()[0]
+                if not aux_proc_re.match(cmdline):
+                    user_backends.append(child)
+                    user_backends_cmdlines.append(cmdline)
+            except psutil.NoSuchProcess:
+                pass
+        if user_backends:
+            logger.debug('Waiting for user backends %s to close', ', '.join(user_backends_cmdlines))
+            psutil.wait_procs(user_backends)
+        logger.debug("Backends closed")
+
+    @staticmethod
+    def start(pgcommand, data_dir, conf, options):
+        # Unfortunately `pg_ctl start` does not return postmaster pid to us. Without this information
+        # it is hard to know the current state of postgres startup, so we had to reimplement pg_ctl start
+        # in python. It will start postgres, wait for port to be open and wait until postgres will start
+        # accepting connections.
+        # Important!!! We can't just start postgres using subprocess.Popen, because in this case it
+        # will be our child for the rest of our live and we will have to take care of it (`waitpid`).
+        # So we will use the same approach as pg_ctl uses: start a new process, which will start postgres.
+        # This process will write postmaster pid to stdout and exit immediately. Now it's responsibility
+        # of init process to take care about postmaster.
+        # In order to make everything portable we can't use fork&exec approach here, so  we will call
+        # ourselves and pass list of arguments which must be used to start postgres.
+        # On Windows, in order to run a side-by-side assembly the specified env must include a valid SYSTEMROOT.
+        env = {p: os.environ[p] for p in ('PATH', 'LD_LIBRARY_PATH', 'LC_ALL', 'LANG', 'SYSTEMROOT') if p in os.environ}
+        try:
+            proc = PostmasterProcess._from_pidfile(data_dir)
+            if proc and not proc._is_postmaster_process():
+                # Upon start postmaster process performs various safety checks if there is a postmaster.pid
+                # file in the data directory. Although Patroni already detected that the running process
+                # corresponding to the postmaster.pid is not a postmaster, the new postmaster might fail
+                # to start, because it thinks that postmaster.pid is already locked.
+                # Important!!! Unlink of postmaster.pid isn't an option, because it has a lot of nasty race conditions.
+                # Luckily there is a workaround to this problem, we can pass the pid from postmaster.pid
+                # in the `PG_GRANDPARENT_PID` environment variable and postmaster will ignore it.
+                logger.info("Telling pg_ctl that it is safe to ignore postmaster.pid for process %s", proc.pid)
+                env['PG_GRANDPARENT_PID'] = str(proc.pid)
+        except psutil.NoSuchProcess:
+            pass
+        cmdline = [pgcommand, '-D', data_dir, '--config-file={}'.format(conf)] + options
+        logger.debug("Starting postgres: %s", " ".join(cmdline))
+        parent_conn, child_conn = multiprocessing.Pipe(False)
+        proc = multiprocessing.Process(target=pg_ctl_start, args=(child_conn, cmdline, env))
+        proc.start()
+        pid = parent_conn.recv()
+        proc.join()
+        if pid is None:
+            return
+        logger.info('postmaster pid=%s', pid)
+
+        # TODO: In an extremely unlikely case, the process could have exited and the pid reassigned. The start
+        # initiation time is not accurate enough to compare to create time as start time would also likely
+        # be relatively close. We need the subprocess extract pid+start_time in a race free manner.
+        return PostmasterProcess.from_pid(pid)

patroni/postgresql/rewind.py

@@ -0,0 +1,217 @@
+import logging
+import os
+import subprocess
+
+from patroni.dcs import Leader
+from patroni.postgresql.connection import get_connection_cursor
+from patroni.postgresql.misc import parse_history, parse_lsn
+
+logger = logging.getLogger(__name__)
+
+REWIND_STATUS = type('Enum', (), {'INITIAL': 0, 'CHECKPOINT': 1, 'CHECK': 2, 'NEED': 3,
+                                  'NOT_NEED': 4, 'SUCCESS': 5, 'FAILED': 6})
+
+
+class Rewind(object):
+
+    def __init__(self, postgresql):
+        self._postgresql = postgresql
+        self.reset_state()
+
+    @staticmethod
+    def configuration_allows_rewind(data):
+        return data.get('wal_log_hints setting', 'off') == 'on' or data.get('Data page checksum version', '0') != '0'
+
+    @property
+    def can_rewind(self):
+        """ check if pg_rewind executable is there and that pg_controldata indicates
+            we have either wal_log_hints or checksums turned on
+        """
+        # low-hanging fruit: check if pg_rewind configuration is there
+        if not self._postgresql.config.get('use_pg_rewind'):
+            return False
+
+        cmd = [self._postgresql.pgcommand('pg_rewind'), '--help']
+        try:
+            ret = subprocess.call(cmd, stdout=open(os.devnull, 'w'), stderr=subprocess.STDOUT)
+            if ret != 0:  # pg_rewind is not there, close up the shop and go home
+                return False
+        except OSError:
+            return False
+        return self.configuration_allows_rewind(self._postgresql.controldata())
+
+    @property
+    def can_rewind_or_reinitialize_allowed(self):
+        return self._postgresql.config.get('remove_data_directory_on_diverged_timelines') or self.can_rewind
+
+    def trigger_check_diverged_lsn(self):
+        if self.can_rewind_or_reinitialize_allowed and self._state != REWIND_STATUS.NEED:
+            self._state = REWIND_STATUS.CHECK
+
+    def check_leader_is_not_in_recovery(self, **kwargs):
+        if not kwargs.get('database'):
+            kwargs['database'] = self._postgresql.database
+        try:
+            with get_connection_cursor(connect_timeout=3, options='-c statement_timeout=2000', **kwargs) as cur:
+                cur.execute('SELECT pg_catalog.pg_is_in_recovery()')
+                if not cur.fetchone()[0]:
+                    return True
+                logger.info('Leader is still in_recovery and therefore can\'t be used for rewind')
+        except Exception:
+            return logger.exception('Exception when working with leader')
+
+    def _get_local_timeline_lsn_from_controldata(self):
+        timeline = lsn = None
+        data = self._postgresql.controldata()
+        try:
+            if data.get('Database cluster state') == 'shut down in recovery':
+                lsn = data.get('Minimum recovery ending location')
+                timeline = int(data.get("Min recovery ending loc's timeline"))
+                if lsn == '0/0' or timeline == 0:  # it was a master when it crashed
+                    data['Database cluster state'] = 'shut down'
+            if data.get('Database cluster state') == 'shut down':
+                lsn = data.get('Latest checkpoint location')
+                timeline = int(data.get("Latest checkpoint's TimeLineID"))
+        except (TypeError, ValueError):
+            logger.exception('Failed to get local timeline and lsn from pg_controldata output')
+        return timeline, lsn
+
+    def _get_local_timeline_lsn(self):
+        if self._postgresql.is_running():  # if postgres is running - get timeline and lsn from replication connection
+            timeline, lsn = self._postgresql.get_local_timeline_lsn_from_replication_connection()
+        else:  # otherwise analyze pg_controldata output
+            timeline, lsn = self._get_local_timeline_lsn_from_controldata()
+        logger.info('Local timeline=%s lsn=%s', timeline, lsn)
+        return timeline, lsn
+
+    def _check_timeline_and_lsn(self, leader):
+        local_timeline, local_lsn = self._get_local_timeline_lsn()
+        if local_timeline is None or local_lsn is None:
+            return
+
+        if isinstance(leader, Leader):
+            if leader.member.data.get('role') != 'master':
+                return
+        # standby cluster
+        elif not self.check_leader_is_not_in_recovery(**leader.conn_kwargs(self._postgresql.config.replication)):
+            return
+
+        history = need_rewind = None
+        try:
+            with self._postgresql.get_replication_connection_cursor(**leader.conn_kwargs()) as cur:
+                cur.execute('IDENTIFY_SYSTEM')
+                master_timeline = cur.fetchone()[1]
+                logger.info('master_timeline=%s', master_timeline)
+                if local_timeline > master_timeline:  # Not always supported by pg_rewind
+                    need_rewind = True
+                elif master_timeline > 1:
+                    cur.execute('TIMELINE_HISTORY %s', (master_timeline,))
+                    history = bytes(cur.fetchone()[1]).decode('utf-8')
+                    logger.info('master: history=%s', history)
+                else:  # local_timeline == master_timeline == 1
+                    need_rewind = False
+        except Exception:
+            return logger.exception('Exception when working with master via replication connection')
+
+        if history is not None:
+            for parent_timeline, switchpoint, _ in parse_history(history):
+                if parent_timeline == local_timeline:
+                    try:
+                        need_rewind = parse_lsn(local_lsn) >= switchpoint
+                    except (IndexError, ValueError):
+                        logger.exception('Exception when parsing lsn')
+                    break
+                elif parent_timeline > local_timeline:
+                    break
+
+        self._state = need_rewind and REWIND_STATUS.NEED or REWIND_STATUS.NOT_NEED
+
+    def rewind_or_reinitialize_needed_and_possible(self, leader):
+        if leader and leader.name != self._postgresql.name and leader.conn_url and self._state == REWIND_STATUS.CHECK:
+            self._check_timeline_and_lsn(leader)
+        return leader and leader.conn_url and self._state == REWIND_STATUS.NEED
+
+    def check_for_checkpoint_after_promote(self):
+        if self._state == REWIND_STATUS.INITIAL and self._postgresql.is_leader():
+            try:
+                timeline = int(self._postgresql.controldata().get("Latest checkpoint's TimeLineID"))
+                if self._postgresql.get_master_timeline() == timeline:
+                    self._state = REWIND_STATUS.CHECKPOINT
+            except (TypeError, ValueError):
+                logger.exception('Failed to parse timeline from pg_controldata output')
+
+    def checkpoint_after_promote(self):
+        return self._state == REWIND_STATUS.CHECKPOINT
+
+    def pg_rewind(self, r):
+        # prepare pg_rewind connection
+        env = self._postgresql.write_pgpass(r)
+        env['PGOPTIONS'] = '-c statement_timeout=0'
+        dsn_attrs = [
+            ('user', r.get('user')),
+            ('host', r.get('host')),
+            ('port', r.get('port')),
+            ('dbname', r.get('database') or self._postgresql.database),
+            ('sslmode', 'prefer'),
+            ('sslcompression', '1'),
+        ]
+        dsn = " ".join("{0}={1}".format(k, v) for k, v in dsn_attrs if v is not None)
+        logger.info('running pg_rewind from %s', dsn)
+        try:
+            return self._postgresql.cancellable.call([self._postgresql.pgcommand('pg_rewind'), '-D',
+                                                      self._postgresql.data_dir, '--source-server', dsn], env=env) == 0
+        except OSError:
+            return False
+
+    def execute(self, leader):
+        if self._postgresql.is_running() and not self._postgresql.stop(checkpoint=False):
+            return logger.warning('Can not run pg_rewind because postgres is still running')
+
+        # prepare pg_rewind connection
+        r = leader.conn_kwargs(self._postgresql.config.rewind_credentials)
+
+        # 1. make sure that we are really trying to rewind from the master
+        # 2. make sure that pg_control contains the new timeline by:
+        #   running a checkpoint or
+        #   waiting until Patroni on the master will expose checkpoint_after_promote=True
+        checkpoint_status = leader.checkpoint_after_promote if isinstance(leader, Leader) else None
+        if checkpoint_status is None:  # master still runs the old Patroni
+            leader_status = self._postgresql.checkpoint(leader.conn_kwargs(self._postgresql.config.superuser))
+            if leader_status:
+                return logger.warning('Can not use %s for rewind: %s', leader.name, leader_status)
+        elif not checkpoint_status:
+            return logger.info('Waiting for checkpoint on %s before rewind', leader.name)
+        elif not self.check_leader_is_not_in_recovery(**r):
+            return
+
+        if self.pg_rewind(r):
+            self._state = REWIND_STATUS.SUCCESS
+        elif not self.check_leader_is_not_in_recovery(**r):
+            logger.warning('Failed to rewind because master %s become unreachable', leader.name)
+        else:
+            logger.error('Failed to rewind from healty master: %s', leader.name)
+
+            for name in ('remove_data_directory_on_rewind_failure', 'remove_data_directory_on_diverged_timelines'):
+                if self._postgresql.config.get(name):
+                    logger.warning('%s is set. removing...', name)
+                    self._postgresql.remove_data_directory()
+                    self._state = REWIND_STATUS.INITIAL
+                    break
+            else:
+                self._state = REWIND_STATUS.FAILED
+        return False
+
+    def reset_state(self):
+        self._state = REWIND_STATUS.INITIAL
+
+    @property
+    def is_needed(self):
+        return self._state in (REWIND_STATUS.CHECK, REWIND_STATUS.NEED)
+
+    @property
+    def executed(self):
+        return self._state > REWIND_STATUS.NOT_NEED
+
+    @property
+    def failed(self):
+        return self._state == REWIND_STATUS.FAILED

patroni/postgresql/slots.py

@@ -0,0 +1,108 @@
+import logging
+
+from patroni.postgresql.connection import get_connection_cursor
+from collections import defaultdict
+
+logger = logging.getLogger(__name__)
+
+
+def compare_slots(s1, s2):
+    return s1['type'] == s2['type'] and (s1['type'] == 'physical' or
+                                         s1['database'] == s2['database'] and s1['plugin'] == s2['plugin'])
+
+
+class SlotsHandler(object):
+
+    def __init__(self, postgresql):
+        self._postgresql = postgresql
+        self._use_slots = postgresql.config.get('use_slots', True)
+        self._replication_slots = {}  # already existing replication slots
+        self.schedule()
+
+    @property
+    def use_slots(self):
+        return self._use_slots and self._postgresql.major_version >= 90400
+
+    def _query(self, sql, *params):
+        return self._postgresql.query(sql, *params, retry=False)
+
+    def load_replication_slots(self):
+        if self.use_slots and self._schedule_load_slots:
+            replication_slots = {}
+            cursor = self._query('SELECT slot_name, slot_type, plugin, database FROM pg_catalog.pg_replication_slots')
+            for r in cursor:
+                value = {'type': r[1]}
+                if r[1] == 'logical':
+                    value.update({'plugin': r[2], 'database': r[3]})
+                replication_slots[r[0]] = value
+            self._replication_slots = replication_slots
+            self._schedule_load_slots = False
+
+    def drop_replication_slot(self, name):
+        cursor = self._query(('SELECT pg_catalog.pg_drop_replication_slot(%s) WHERE EXISTS (SELECT 1 ' +
+                              'FROM pg_catalog.pg_replication_slots WHERE slot_name = %s AND NOT active)'), name, name)
+        # In normal situation rowcount should be 1, otherwise either slot doesn't exists or it is still active
+        return cursor.rowcount == 1
+
+    def sync_replication_slots(self, cluster):
+        if self.use_slots:
+            try:
+                self.load_replication_slots()
+
+                slots = cluster.get_replication_slots(self._postgresql.name, self._postgresql.role)
+
+                # drop old replication slots which are not presented in desired slots
+                for name in set(self._replication_slots) - set(slots):
+                    if not self.drop_replication_slot(name):
+                        logger.error("Failed to drop replication slot '%s'", name)
+                        self._schedule_load_slots = True
+
+                immediately_reserve = ', true' if self._postgresql.major_version >= 90600 else ''
+
+                logical_slots = defaultdict(dict)
+                for name, value in slots.items():
+                    if name in self._replication_slots and not compare_slots(value, self._replication_slots[name]):
+                        logger.info("Trying to drop replication slot '%s' because value is changing from %s to %s",
+                                    name, self._replication_slots[name], value)
+                        if not self.drop_replication_slot(name):
+                            logger.error("Failed to drop replication slot '%s'", name)
+                            self._schedule_load_slots = True
+                            continue
+                        self._replication_slots.pop(name)
+                    if name not in self._replication_slots:
+                        if value['type'] == 'physical':
+                            try:
+                                self._query(("SELECT pg_catalog.pg_create_physical_replication_slot(%s{0})" +
+                                             " WHERE NOT EXISTS (SELECT 1 FROM pg_catalog.pg_replication_slots" +
+                                             " WHERE slot_type = 'physical' AND slot_name = %s)").format(
+                                                 immediately_reserve), name, name)
+                            except Exception:
+                                logger.exception("Failed to create physical replication slot '%s'", name)
+                                self._schedule_load_slots = True
+                        elif value['type'] == 'logical' and name not in self._replication_slots:
+                            logical_slots[value['database']][name] = value
+
+                # create new logical slots
+                for database, values in logical_slots.items():
+                    conn_kwargs = self._postgresql.config.local_connect_kwargs
+                    conn_kwargs['database'] = database
+                    with get_connection_cursor(**conn_kwargs) as cur:
+                        for name, value in values.items():
+                            try:
+                                cur.execute("SELECT pg_catalog.pg_create_logical_replication_slot(%s, %s)" +
+                                            " WHERE NOT EXISTS (SELECT 1 FROM pg_catalog.pg_replication_slots" +
+                                            " WHERE slot_type = 'logical' AND slot_name = %s)",
+                                            (name, value['plugin'], name))
+                            except Exception:
+                                logger.exception("Failed to create logical replication slot '%s' plugin='%s'",
+                                                 name, value['plugin'])
+                                self._schedule_load_slots = True
+                self._replication_slots = slots
+            except Exception:
+                logger.exception('Exception when changing replication slots')
+                self._schedule_load_slots = True
+
+    def schedule(self, value=None):
+        if value is None:
+            value = self.use_slots
+        self._schedule_load_slots = value

patroni/scripts/aws.py

@@ -6,63 +6,64 @@ from requests.exceptions import RequestException
 import sys
 import boto.ec2
 
+from patroni.utils import Retry, RetryFailedError
+
 logger = logging.getLogger(__name__)
 
 
 class AWSConnection(object):
+
     def __init__(self, cluster_name):
         self.available = False
         self.cluster_name = cluster_name if cluster_name is not None else 'unknown'
+        self._retry = Retry(deadline=300, max_delay=30, max_tries=-1, retry_exceptions=(boto.exception.StandardError,))
         try:
             # get the instance id
-            r = requests.get('http://169.254.169.254/latest/dynamic/instance-identity/document', timeout=0.1)
+            r = requests.get('http://169.254.169.254/latest/dynamic/instance-identity/document', timeout=2.1)
         except RequestException:
-            logger.info("cannot query AWS meta-data")
+            logger.error('cannot query AWS meta-data')
             return
+
         if r.ok:
             try:
                 content = r.json()
                 self.instance_id = content['instanceId']
                 self.region = content['region']
-            except Exception as e:
+            except Exception:
-                logger.info('unable to fetch instance id and region from AWS meta-data: {}'.format(e))
+                logger.exception('unable to fetch instance id and region from AWS meta-data')
                 return
             self.available = True
 
+    def retry(self, *args, **kwargs):
+        return self._retry.copy()(*args, **kwargs)
+
     def aws_available(self):
         return self.available
 
-    def _tag_ebs(self, role):
+    def _tag_ebs(self, conn, role):
         """ set tags, carrying the cluster name, instance role and instance id for the EBS storage """
-        if not self.available:
-            return False
-
         tags = {'Name': 'spilo_' + self.cluster_name, 'Role': role, 'Instance': self.instance_id}
-        try:
+        volumes = conn.get_all_volumes(filters={'attachment.instance-id': self.instance_id})
-            conn = boto.ec2.connect_to_region(self.region)
+        conn.create_tags([v.id for v in volumes], tags)
-            volumes = conn.get_all_volumes(filters={'attachment.instance-id': self.instance_id})
-            conn.create_tags([v.id for v in volumes], tags)
-        except Exception as e:
-            logger.info('could not set tags for EBS storage devices attached: {}'.format(e))
-            return False
-        return True
 
-    def _tag_ec2(self, role):
+    def _tag_ec2(self, conn, role):
         """ tag the current EC2 instance with a cluster role """
-        if not self.available:
-            return False
         tags = {'Role': role}
-        try:
+        conn.create_tags([self.instance_id], tags)
-            conn = boto.ec2.connect_to_region(self.region)
-            conn.create_tags([self.instance_id], tags)
-        except Exception as e:
-            logger.info("could not set tags for EC2 instance %s: %s", self.instance_id, e)
-            return False
-        return True
 
     def on_role_change(self, new_role):
-        ret = self._tag_ec2(new_role)
+        if not self.available:
-        return self._tag_ebs(new_role) and ret
+            return False
+        try:
+            conn = self.retry(boto.ec2.connect_to_region, self.region)
+            self.retry(self._tag_ec2, conn, new_role)
+            self.retry(self._tag_ebs, conn, new_role)
+        except RetryFailedError:
+            logger.warning("Unable to communicate to AWS "
+                           "when setting tags for the EC2 instance {0} "
+                           "and attached EBS volumes".format(self.instance_id))
+            return False
+        return True
 
 
 def main():

patroni/scripts/wale_restore.py

@@ -23,19 +23,29 @@
 # currently also requires that you configure the restore_command to use wal_e, example:
 #       recovery_conf:
 #               restore_command: envdir /etc/wal-e.d/env wal-e wal-fetch "%f" "%p" -p 1
-
+import argparse
-from collections import namedtuple
+import csv
 import logging
 import os
 import psycopg2
 import subprocess
 import sys
 import time
-import argparse
+
+from collections import namedtuple
 
 logger = logging.getLogger(__name__)
 
 RETRY_SLEEP_INTERVAL = 1
+si_prefixes = ['K', 'M', 'G', 'T', 'P', 'E', 'Z', 'Y']
+
+
+# Meaningful names to the exit codes used by WALERestore
+ExitCode = type('Enum', (), {
+    'SUCCESS': 0,  #: Succeeded
+    'RETRY_LATER': 1,  #: External issue, retry later
+    'FAIL': 2  #: Don't try again unless configuration changes
+})
 
 
 # We need to know the current PG version in order to figure out the correct WAL directory name
@@ -50,66 +60,141 @@ def get_major_version(data_dir):
     return 0.0
 
 
+def repr_size(n_bytes):
+    """
+    >>> repr_size(1000)
+    '1000 Bytes'
+    >>> repr_size(8257332324597)
+    '7.5 TiB'
+    """
+    if n_bytes < 1024:
+        return '{0} Bytes'.format(n_bytes)
+    i = -1
+    while n_bytes > 1023:
+        n_bytes /= 1024.0
+        i += 1
+    return '{0} {1}iB'.format(round(n_bytes, 1), si_prefixes[i])
+
+
+def size_as_bytes(size_, prefix):
+    """
+    >>> size_as_bytes(7.5, 'T')
+    8246337208320
+    """
+    prefix = prefix.upper()
+
+    assert prefix in si_prefixes
+
+    exponent = si_prefixes.index(prefix) + 1
+
+    return int(size_ * (1024.0 ** exponent))
+
+
+WALEConfig = namedtuple(
+    'WALEConfig',
+    [
+        'env_dir',
+        'threshold_mb',
+        'threshold_pct',
+        'cmd',
+    ]
+)
+
+
 class WALERestore(object):
-    def __init__(self, scope, datadir, connstring, env_dir, threshold_mb, threshold_pct, use_iam, no_master, retries):
+    def __init__(self, scope, datadir, connstring, env_dir, threshold_mb,
+                 threshold_pct, use_iam, no_master, retries):
         self.scope = scope
         self.master_connection = connstring
         self.data_dir = datadir
-        self.wal_e = namedtuple('wale', 'dir,threshold_mb,threshold_pct,iam_string,cmd')
-        self.wal_e.dir = env_dir
-        self.wal_e.threshold_mb = threshold_mb
-        self.wal_e.threshold_pct = threshold_pct
-        self.wal_e.iam_string = ' --aws-instance-profile ' if use_iam == 1 else ''
         self.no_master = no_master
-        self.wal_e.cmd = 'envdir {0} wal-e {1} '.format(self.wal_e.dir, self.wal_e.iam_string)
+
-        self.init_error = (not os.path.exists(self.wal_e.dir))
+        wale_cmd = [
+            'envdir',
+            env_dir,
+            'wal-e',
+        ]
+
+        if use_iam == 1:
+            wale_cmd += ['--aws-instance-profile']
+
+        self.wal_e = WALEConfig(
+            env_dir=env_dir,
+            threshold_mb=threshold_mb,
+            threshold_pct=threshold_pct,
+            cmd=wale_cmd,
+        )
+
+        self.init_error = (not os.path.exists(self.wal_e.env_dir))
         self.retries = retries
 
     def run(self):
-        """ creates a new replica using WAL-E """
+        """
-        if not self.init_error:
+        Creates a new replica using WAL-E
-            try:
+
-                ret = self.should_use_s3_to_create_replica()
+        Returns
-                if ret:
+        -------
-                    return self.create_replica_with_s3()
+        ExitCode
-                elif ret is None:  # caught an exception, need to retry
+            0 = Success
-                    return 1
+            1 = Error, try again
-            except Exception:
+            2 = Error, don't try again
-                logger.exception("Exception when running WAL-E restore")
+
-        return 2
+        """
+        if self.init_error:
+            logger.error('init error: %r did not exist at initialization time',
+                         self.wal_e.env_dir)
+            return ExitCode.FAIL
+
+        try:
+            should_use_s3 = self.should_use_s3_to_create_replica()
+            if should_use_s3 is None:  # Need to retry
+                return ExitCode.RETRY_LATER
+            elif should_use_s3:
+                return self.create_replica_with_s3()
+            elif not should_use_s3:
+                return ExitCode.FAIL
+        except Exception:
+            logger.exception("Unhandled exception when running WAL-E restore")
+        return ExitCode.FAIL
 
     def should_use_s3_to_create_replica(self):
         """ determine whether it makes sense to use S3 and not pg_basebackup """
 
         threshold_megabytes = self.wal_e.threshold_mb
-        threshold_backup_size_percentage = self.wal_e.threshold_pct
+        threshold_percent = self.wal_e.threshold_pct
 
         try:
-            latest_backup = subprocess.check_output(self.wal_e.cmd.split() + ['backup-list', '--detail', 'LATEST'])
+            cmd = self.wal_e.cmd + ['backup-list', '--detail', 'LATEST']
-            # name    last_modified   expanded_size_bytes wal_segment_backup_start    wal_segment_offset_backup_start
+
-            #                                                   wal_segment_backup_stop wal_segment_offset_backup_stop
+            logger.debug('calling %r', cmd)
-            # base_00000001000000000000007F_00000040  2015-05-18T10:13:25.000Z
+            wale_output = subprocess.check_output(cmd)
-            # 20310671    00000001000000000000007F    00000040
+
-            # 00000001000000000000007F    00000240
+            reader = csv.DictReader(wale_output.decode('utf-8').splitlines(),
-            backup_strings = latest_backup.decode('utf-8').splitlines() if latest_backup else ()
+                                    dialect='excel-tab')
-            if len(backup_strings) != 2:
+            rows = list(reader)
+            if not len(rows):
+                logger.warning('wal-e did not find any backups')
                 return False
 
-            names = backup_strings[0].split()
+            # This check might not add much, it was performed in the previous
-            vals = backup_strings[1].split()
+            # version of this code. since the old version rolled CSV parsing the
-            if (len(names) != len(vals)) or (len(names) != 7):
+            # check may have been part of the CSV parsing.
+            if len(rows) > 1:
+                logger.warning(
+                    'wal-e returned more than one row of backups: %r',
+                    rows)
                 return False
 
-            backup_info = dict(zip(names, vals))
+            backup_info = rows[0]
         except subprocess.CalledProcessError:
             logger.exception("could not query wal-e latest backup")
             return None
 
         try:
-            backup_size = backup_info['expanded_size_bytes']
+            backup_size = int(backup_info['expanded_size_bytes'])
             backup_start_segment = backup_info['wal_segment_backup_start']
             backup_start_offset = backup_info['wal_segment_offset_backup_start']
-        except Exception:
+        except KeyError:
             logger.exception("unable to get some of WALE backup parameters")
             return None
 
@@ -124,24 +209,28 @@ class WALERestore(object):
         # construct the LSN from the segment and offset
         backup_start_lsn = '{0}/{1}'.format(lsn_segment, lsn_offset)
 
-        diff_in_bytes = int(backup_size)
+        diff_in_bytes = backup_size
         attempts_no = 0
         while True:
             if self.master_connection:
                 try:
                     # get the difference in bytes between the current WAL location and the backup start offset
                     with psycopg2.connect(self.master_connection) as con:
+                        if con.server_version >= 100000:
+                            wal_name = 'wal'
+                            lsn_name = 'lsn'
+                        else:
+                            wal_name = 'xlog'
+                            lsn_name = 'location'
                         con.autocommit = True
                         with con.cursor() as cur:
-                            cur.execute("""SELECT CASE WHEN pg_is_in_recovery()
+                            cur.execute(("SELECT CASE WHEN pg_catalog.pg_is_in_recovery()"
-                                                       THEN GREATEST(
+                                         " THEN GREATEST(pg_catalog.pg_{0}_{1}_diff(COALESCE("
-                                                                pg_xlog_location_diff(COALESCE(
+                                         "pg_last_{0}_receive_{1}(), '0/0'), %s)::bigint, "
-                                                                  pg_last_xlog_receive_location(), '0/0'), %s)::bigint,
+                                         "pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_last_{0}_replay_{1}(), %s)::bigint)"
-                                                                pg_xlog_location_diff(
+                                         " ELSE pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_current_{0}_{1}(), %s)::bigint"
-                                                                  pg_last_xlog_replay_location(), %s)::bigint)
+                                         " END").format(wal_name, lsn_name),
-                                                       ELSE pg_xlog_location_diff(
+                                        (backup_start_lsn, backup_start_lsn, backup_start_lsn))
-                                                                pg_current_xlog_location(), %s)::bigint
-                                                   END""", (backup_start_lsn, backup_start_lsn, backup_start_lsn))
 
                             diff_in_bytes = int(cur.fetchone()[0])
                 except psycopg2.Error:
@@ -163,8 +252,47 @@ class WALERestore(object):
 
         # if the size of the accumulated WAL segments is more than a certan percentage of the backup size
         # or exceeds the pre-determined size - pg_basebackup is chosen instead.
-        return (diff_in_bytes < int(threshold_megabytes) * 1048576) and\
+        is_size_thresh_ok = diff_in_bytes < int(threshold_megabytes) * 1048576
-            (diff_in_bytes < int(backup_size) * float(threshold_backup_size_percentage) / 100)
+        threshold_pct_bytes = backup_size * threshold_percent / 100.0
+        is_percentage_thresh_ok = float(diff_in_bytes) < int(threshold_pct_bytes)
+        are_thresholds_ok = is_size_thresh_ok and is_percentage_thresh_ok
+
+        class Size(object):
+            def __init__(self, n_bytes, prefix=None):
+                self.n_bytes = n_bytes
+                self.prefix = prefix
+
+            def __repr__(self):
+                if self.prefix is not None:
+                    n_bytes = size_as_bytes(self.n_bytes, self.prefix)
+                else:
+                    n_bytes = self.n_bytes
+                return repr_size(n_bytes)
+
+        class HumanContext(object):
+            def __init__(self, items):
+                self.items = items
+
+            def __repr__(self):
+                return ', '.join('{}={!r}'.format(key, value)
+                                 for key, value in self.items)
+
+        human_context = repr(HumanContext([
+            ('threshold_size', Size(threshold_megabytes, 'M')),
+            ('threshold_percent', threshold_percent),
+            ('threshold_percent_size', Size(threshold_pct_bytes)),
+            ('backup_size', Size(backup_size)),
+            ('backup_diff', Size(diff_in_bytes)),
+            ('is_size_thresh_ok', is_size_thresh_ok),
+            ('is_percentage_thresh_ok', is_percentage_thresh_ok),
+        ]))
+
+        if not are_thresholds_ok:
+            logger.info('wal-e backup size diff is over threshold, falling back '
+                        'to other means of restore: %s', human_context)
+        else:
+            logger.info('Thresholds are OK, using wal-e basebackup: %s', human_context)
+        return are_thresholds_ok
 
     def fix_subdirectory_path_if_broken(self, dirname):
         # in case it is a symlink pointing to a non-existing location, remove it and create the actual directory
@@ -187,15 +315,19 @@ class WALERestore(object):
     def create_replica_with_s3(self):
         # if we're set up, restore the replica using fetch latest
         try:
-            ret = subprocess.call(self.wal_e.cmd.split() + ['backup-fetch', '{}'.format(self.data_dir), 'LATEST'])
+            cmd = self.wal_e.cmd + ['backup-fetch',
+                                    '{}'.format(self.data_dir),
+                                    'LATEST']
+            logger.debug('calling: %r', cmd)
+            exit_code = subprocess.call(cmd)
         except Exception as e:
             logger.error('Error when fetching backup with WAL-E: {0}'.format(e))
-            return 1
+            return ExitCode.RETRY_LATER
 
-        if (ret == 0 and not
+        if (exit_code == 0 and not
-           self.fix_subdirectory_path_if_broken('pg_xlog' if get_major_version(self.data_dir) < 10.0 else 'pg_wal')):
+           self.fix_subdirectory_path_if_broken('pg_xlog' if get_major_version(self.data_dir) < 10 else 'pg_wal')):
-            return 2
+            return ExitCode.FAIL
-        return ret
+        return exit_code
 
 
 def main():
@@ -213,6 +345,9 @@ def main():
     parser.add_argument('--no_master', type=int, default=0)
     args = parser.parse_args()
 
+    exit_code = None
+    assert args.retries >= 0
+
     # Retry cloning in a loop. We do separate retries for the master
     # connection attempt inside should_use_s3_to_create_replica,
     # because we need to differentiate between the last attempt and
@@ -223,12 +358,13 @@ def main():
                               env_dir=args.envdir, threshold_mb=args.threshold_megabytes,
                               threshold_pct=args.threshold_backup_size_percentage, use_iam=args.use_iam,
                               no_master=args.no_master, retries=args.retries)
-        ret = restore.run()
+        exit_code = restore.run()
-        if ret != 1:  # only WAL-E failures lead to the retry
+        if not exit_code == ExitCode.RETRY_LATER:  # only WAL-E failures lead to the retry
+            logger.debug('exit_code is %r, not retrying', exit_code)
             break
         time.sleep(RETRY_SLEEP_INTERVAL)
 
-    return ret
+    return exit_code
 
 
 if __name__ == '__main__':

patroni/utils.py

@@ -1,12 +1,20 @@
+import logging
 import random
-import time
 import re
+import time
 
 from dateutil import tz
 from patroni.exceptions import PatroniException
 
 tzutc = tz.tzutc()
 
+logger = logging.getLogger(__name__)
+
+OCT_RE = re.compile(r'^[-+]?0[0-7]*')
+DEC_RE = re.compile(r'^[-+]?(0|[1-9][0-9]*)')
+HEX_RE = re.compile(r'^[-+]?0x[0-9a-fA-F]+')
+DBL_RE = re.compile(r'^[-+]?[0-9]*\.?[0-9]+([eE][-+]?[0-9]+)?')
+
 
 def deep_compare(obj1, obj2):
     """
@@ -89,42 +97,77 @@ def strtol(value, strict=True):
     True
     >>> strtol(' d ') == (None, 'd')
     True
+    >>> strtol(' 1 d ') == (1, ' d')
+    True
     >>> strtol('9s', False) == (9, 's')
     True
     >>> strtol(' s ', False) == (1, 's')
     True
     """
     value = str(value).strip()
-    l = len(value)
+    for regex, base in ((HEX_RE, 16), (OCT_RE, 8), (DEC_RE, 10)):
-    i = 0
+        match = regex.match(value)
-    # skip sign:
+        if match:
-    if i < l and value[i] in ('-', '+'):
+            end = match.end()
-        i += 1
+            return int(value[:end], base), value[end:]
+    return (None if strict else 1), value
 
-    # we always expect to get digit in the beginning
-    if i < l and value[i].isdigit():
-        if value[i] == '0':
-            i += 1
-            if i < l and value[i] in ('x', 'X'):  # '0' followed by 'x': HEX
-                base = 16
-                i += 1
-            else:  # just starts with '0': OCT
-                base = 8
-        else:  # any other digit: DEC
-            base = 10
 
-        ret = None
+def strtod(value):
-        while i <= l:
+    """As most as possible close equivalent of strtod(3) function used by postgres to parse parameter values.
-            try:  # try to find maximally long number
+    >>> strtod(' A ') == (None, 'A')
-                i += 1  # by giving to `int` longer and longer strings
+    True
-                ret = int(value[:i], base)
+    """
-            except ValueError:  # until we will not get an exception or end of the string
+    value = str(value).strip()
-                i -= 1
+    match = DBL_RE.match(value)
-                break
+    if match:
-        if ret is not None:  # yay! there is a number in the beginning of the string
+        end = match.end()
-            return ret, value[i:].strip()  # return the number and the "rest"
+        return float(value[:end]), value[end:]
+    return None, value
 
-    return (None if strict else 1), value.strip()
+
+def rint(value):
+    """
+    >>> rint(0.5) == 0
+    True
+    >>> rint(0.501) == 1
+    True
+    >>> rint(1.5) == 2
+    True
+    """
+
+    ret = round(value)
+    return 2.0 * round(value / 2.0) if abs(ret - value) == 0.5 else ret
+
+
+def convert_to_base_unit(value, unit, base_unit):
+    convert = {
+        'B': {'B': 1, 'kB': 1024, 'MB': 1024 * 1024, 'GB': 1024 * 1024 * 1024, 'TB': 1024 * 1024 * 1024 * 1024},
+        'kB': {'B': 1.0 / 1024, 'kB': 1, 'MB': 1024, 'GB': 1024 * 1024, 'TB': 1024 * 1024 * 1024},
+        'MB': {'B': 1.0 / (1024 * 1024), 'kB': 1.0 / 1024, 'MB': 1, 'GB': 1024, 'TB': 1024 * 1024},
+        'ms': {'us': 1.0 / 1000, 'ms': 1, 's': 1000, 'min': 1000 * 60, 'h': 1000 * 60 * 60, 'd': 1000 * 60 * 60 * 24},
+        's': {'us': 1.0 / (1000 * 1000), 'ms': 1.0 / 1000, 's': 1, 'min': 60, 'h': 60 * 60, 'd': 60 * 60 * 24},
+        'min': {'us': 1.0 / (1000 * 1000 * 60), 'ms': 1.0 / (1000 * 60), 's': 1.0 / 60, 'min': 1, 'h': 60, 'd': 60 * 24}
+    }
+
+    round_order = {
+        'TB': 'GB', 'GB': 'MB', 'MB': 'kB', 'kB': 'B',
+        'd': 'h', 'h': 'min', 'min': 's', 's': 'ms', 'ms': 'us'
+    }
+
+    if base_unit and base_unit not in convert:
+        base_value, base_unit = strtol(base_unit, False)
+    else:
+        base_value = 1
+
+    if base_unit in convert and unit in convert[base_unit]:
+        value *= convert[base_unit][unit] / float(base_value)
+
+        if unit in round_order:
+            multiplier = convert[base_unit][round_order[unit]]
+            value = rint(value / float(multiplier)) * multiplier
+
+        return value
 
 
 def parse_int(value, base_unit=None):
@@ -137,56 +180,71 @@ def parse_int(value, base_unit=None):
     True
     >>> parse_int('1000 ms', 's') == 1
     True
-    >>> parse_int('1GB', 'MB') is None
+    >>> parse_int('1TB', 'GB') is None
     True
     >>> parse_int(0) == 0
     True
+    >>> parse_int('6GB', '16MB') == 384
+    True
+    >>> parse_int('4097.4kB', 'kB') == 4097
+    True
+    >>> parse_int('4097.5kB', 'kB') == 4098
+    True
     """
 
-    convert = {
+    val, unit = strtol(value)
-        'kB': {'kB': 1, 'MB': 1024, 'GB': 1024 * 1024, 'TB': 1024 * 1024 * 1024},
+    if val is None and unit.startswith('.') or unit and unit[0] in ('.', 'e', 'E'):
-        'ms': {'ms': 1, 's': 1000, 'min': 1000 * 60, 'h': 1000 * 60 * 60, 'd': 1000 * 60 * 60 * 24},
+        val, unit = strtod(value)
-        's': {'ms': -1000, 's': 1, 'min': 60, 'h': 60 * 60, 'd': 60 * 60 * 24},
-        'min': {'ms': -1000 * 60, 's': -60, 'min': 1, 'h': 60, 'd': 60 * 24}
-    }
 
-    value, unit = strtol(value)
+    if val is not None:
-    if value is not None:
+        unit = unit.strip()
         if not unit:
-            return value
+            return int(rint(val))
 
-        if base_unit and base_unit not in convert:
+        val = convert_to_base_unit(val, unit, base_unit)
-            base_value, base_unit = strtol(base_unit, False)
+        if val is not None:
-        else:
+            return int(rint(val))
-            base_value = 1
+
-        if base_unit in convert and unit in convert[base_unit]:
+
-            multiplier = convert[base_unit][unit]
+def parse_real(value, base_unit=None):
-            if multiplier < 0:
+    """
-                value /= -multiplier
+    >>> parse_real(' +0.0005 ') == 0.0005
-            else:
+    True
-                value *= multiplier
+    >>> parse_real('0.0005ms', 'ms') == 0.0
-            return int(value/base_value)
+    True
+    >>> parse_real('0.00051ms', 'ms') == 0.001
+    True
+    """
+    val, unit = strtod(value)
+
+    if val is not None:
+        unit = unit.strip()
+        if not unit:
+            return val
+
+        return convert_to_base_unit(val, unit, base_unit)
 
 
 def compare_values(vartype, unit, old_value, new_value):
     """
     >>> compare_values('enum', None, 'remote_write', 'REMOTE_WRITE')
     True
-    >>> compare_values('real', None, '1.23', 1.23)
+    >>> compare_values('real', None, '1e-06', 0.000001)
     True
     """
 
-    # if the integer or bool new_value is not correct this function will return False
+    converters = {
-    if vartype == 'bool':
+        'bool': lambda v1, v2: parse_bool(v1),
-        old_value = parse_bool(old_value)
+        'integer': parse_int,
-        new_value = parse_bool(new_value)
+        'real': parse_real,
-    elif vartype == 'integer':
+        'enum': lambda v1, v2: str(v1).lower(),
-        old_value = parse_int(old_value)
+        'string': lambda v1, v2: str(v1)
-        new_value = parse_int(new_value, unit)
+    }
-    elif vartype == 'enum':
+
-        return str(old_value).lower() == str(new_value).lower()
+    convert = converters.get(vartype) or converters['string']
-    else:  # ('string', 'real')
+    old_value = convert(old_value, None)
-        return str(old_value) == str(new_value)
+    new_value = convert(new_value, unit)
+
     return old_value is not None and new_value is not None and old_value == new_value
 
 
@@ -194,10 +252,6 @@ def _sleep(interval):
     time.sleep(interval)
 
 
-def is_valid_pg_version(version):
-    return re.match(r'[1-9][0-9]?(\.(0|([1-9][0-9]?))){2}$', version)
-
-
 class RetryFailedError(PatroniException):
 
     """Raised when retrying an operation ultimately failed, after retrying the maximum number of attempts."""
@@ -257,17 +311,19 @@ class Retry(object):
                 if self.deadline is not None and self._cur_stoptime is None:
                     self._cur_stoptime = time.time() + self.deadline
                 return func(*args, **kwargs)
-            except self.retry_exceptions:
+            except self.retry_exceptions as e:
                 # Note: max_tries == -1 means infinite tries.
                 if self._attempts == self.max_tries:
+                    logger.warning('Retry got exception: %s', e)
                     raise RetryFailedError("Too many retry attempts")
                 self._attempts += 1
                 sleeptime = self._cur_delay + (random.randint(0, self.max_jitter) / 100.0)
 
                 if self._cur_stoptime is not None and time.time() + sleeptime >= self._cur_stoptime:
+                    logger.warning('Retry got exception: %s', e)
                     raise RetryFailedError("Exceeded retry deadline")
-                else:
+                logger.debug('Retry got exception: %s', e)
-                    self.sleep_func(sleeptime)
+                self.sleep_func(sleeptime)
                 self._cur_delay = min(self._cur_delay * self.backoff, self.max_delay)
 
 
@@ -280,3 +336,19 @@ def polling_loop(timeout, interval=1):
         yield iteration
         iteration += 1
         time.sleep(interval)
+
+
+def split_host_port(value, default_port):
+    t = value.rsplit(':', 1)
+    t.append(default_port)
+    return t[0], int(t[1])
+
+
+def uri(proto, netloc, path='', user=None):
+    host, port = netloc if isinstance(netloc, (list, tuple)) else split_host_port(netloc, 0)
+    if host and ':' in host and host[0] != '[' and host[-1] != ']':
+        host = '[{0}]'.format(host)
+    port = ':{0}'.format(port) if port else ''
+    path = '/{0}'.format(path) if path and not path.startswith('/') else path
+    user = '{0}@'.format(user) if user else ''
+    return '{0}://{1}{2}{3}{4}'.format(proto, user, host, port, path)

patroni/version.py

@@ -1 +1 @@
-__version__ = '1.2.4'
+__version__ = '1.6.0'

patroni/watchdog/__init__.py

@@ -0,0 +1,2 @@
+from patroni.watchdog.base import WatchdogError, Watchdog
+__all__ = ['WatchdogError', 'Watchdog']

patroni/watchdog/base.py

@@ -0,0 +1,313 @@
+import abc
+import logging
+import platform
+import six
+import sys
+from threading import RLock
+
+from patroni.exceptions import WatchdogError
+
+__all__ = ['WatchdogError', 'Watchdog']
+
+logger = logging.getLogger(__name__)
+
+MODE_REQUIRED = 'required'    # Will not run if a watchdog is not available
+MODE_AUTOMATIC = 'automatic'  # Will use a watchdog if one is available
+MODE_OFF = 'off'              # Will not try to use a watchdog
+
+
+def parse_mode(mode):
+    if mode is False:
+        return MODE_OFF
+    mode = mode.lower()
+    if mode in ['require', 'required']:
+        return MODE_REQUIRED
+    elif mode in ['auto', 'automatic']:
+        return MODE_AUTOMATIC
+    else:
+        if mode not in ['off', 'disable', 'disabled']:
+            logger.warning("Watchdog mode {0} not recognized, disabling watchdog".format(mode))
+        return MODE_OFF
+
+
+def synchronized(func):
+    def wrapped(self, *args, **kwargs):
+        with self._lock:
+            return func(self, *args, **kwargs)
+    return wrapped
+
+
+class WatchdogConfig(object):
+    """Helper to contain a snapshot of configuration"""
+    def __init__(self, config):
+        self.mode = parse_mode(config['watchdog'].get('mode', 'automatic'))
+        self.ttl = config['ttl']
+        self.loop_wait = config['loop_wait']
+        self.safety_margin = config['watchdog'].get('safety_margin', 5)
+        self.driver = config['watchdog'].get('driver', 'default')
+        self.driver_config = dict((k, v) for k, v in config['watchdog'].items()
+                                  if k not in ['mode', 'safety_margin', 'driver'])
+
+    def __eq__(self, other):
+        return isinstance(other, WatchdogConfig) and \
+            all(getattr(self, attr) == getattr(other, attr) for attr in
+                ['mode', 'ttl', 'loop_wait', 'safety_margin', 'driver', 'driver_config'])
+
+    def __ne__(self, other):
+        return not self == other
+
+    def get_impl(self):
+        if self.driver == 'testing':
+            from patroni.watchdog.linux import TestingWatchdogDevice
+            return TestingWatchdogDevice.from_config(self.driver_config)
+        elif platform.system() == 'Linux' and self.driver == 'default':
+            from patroni.watchdog.linux import LinuxWatchdogDevice
+            return LinuxWatchdogDevice.from_config(self.driver_config)
+        else:
+            return NullWatchdog()
+
+    @property
+    def timeout(self):
+        if self.safety_margin == -1:
+            return int(self.ttl // 2)
+        else:
+            return self.ttl - self.safety_margin
+
+    @property
+    def timing_slack(self):
+        return self.timeout - self.loop_wait
+
+
+class Watchdog(object):
+    """Facade to dynamically manage watchdog implementations and handle config changes.
+
+    When activation fails underlying implementation will be switched to a Null implementation. To avoid log spam
+    activation will only be retried when watchdog configuration is changed."""
+    def __init__(self, config):
+        self.active_config = self.config = WatchdogConfig(config)
+        self._lock = RLock()
+        self.active = False
+
+        if self.config.mode == MODE_OFF:
+            self.impl = NullWatchdog()
+        else:
+            self.impl = self.config.get_impl()
+            if self.config.mode == MODE_REQUIRED and self.impl.is_null:
+                logger.error("Configuration requires a watchdog, but watchdog is not supported on this platform.")
+                sys.exit(1)
+
+    @synchronized
+    def reload_config(self, config):
+        self.config = WatchdogConfig(config)
+        # Turning a watchdog off can always be done immediately
+        if self.config.mode == MODE_OFF:
+            if self.active:
+                self._disable()
+            self.active_config = self.config
+            self.impl = NullWatchdog()
+        # If watchdog is not active we can apply config immediately to show any warnings early. Otherwise we need to
+        # delay until next time a keepalive is sent so timeout matches up with leader key update.
+        if not self.active:
+            if self.config.driver != self.active_config.driver or \
+               self.config.driver_config != self.active_config.driver_config:
+                self.impl = self.config.get_impl()
+            self.active_config = self.config
+
+    @synchronized
+    def activate(self):
+        """Activates the watchdog device with suitable timeouts. While watchdog is active keepalive needs
+        to be called every time loop_wait expires.
+
+        :returns False if a safe watchdog could not be configured, but is required.
+        """
+        self.active = True
+        return self._activate()
+
+    def _activate(self):
+        self.active_config = self.config
+
+        if self.config.timing_slack < 0:
+            logger.warning('Watchdog not supported because leader TTL {0} is less than 2x loop_wait {1}'
+                           .format(self.config.ttl, self.config.loop_wait))
+            self.impl = NullWatchdog()
+
+        try:
+            self.impl.open()
+            actual_timeout = self._set_timeout()
+        except WatchdogError as e:
+            logger.warning("Could not activate %s: %s", self.impl.describe(), e)
+            self.impl = NullWatchdog()
+
+        if self.impl.is_running and not self.impl.can_be_disabled:
+            logger.warning("Watchdog implementation can't be disabled."
+                           " Watchdog will trigger after Patroni loses leader key.")
+
+        if not self.impl.is_running or actual_timeout > self.config.timeout:
+            if self.config.mode == MODE_REQUIRED:
+                if self.impl.is_null:
+                    logger.error("Configuration requires watchdog, but watchdog could not be configured.")
+                else:
+                    logger.error("Configuration requires watchdog, but a safe watchdog timeout {0} could"
+                                 " not be configured. Watchdog timeout is {1}.".format(
+                                     self.config.timeout, actual_timeout))
+                return False
+            else:
+                if not self.impl.is_null:
+                    logger.warning("Watchdog timeout {0} seconds does not ensure safe termination within {1} seconds"
+                                   .format(actual_timeout, self.config.timeout))
+
+        if self.is_running:
+            logger.info("{0} activated with {1} second timeout, timing slack {2} seconds"
+                        .format(self.impl.describe(), actual_timeout, self.config.timing_slack))
+        else:
+            if self.config.mode == MODE_REQUIRED:
+                logger.error("Configuration requires watchdog, but watchdog could not be activated")
+                return False
+
+        return True
+
+    def _set_timeout(self):
+        if self.impl.has_set_timeout():
+            self.impl.set_timeout(self.config.timeout)
+
+        # Safety checks for watchdog implementations that don't support configurable timeouts
+        actual_timeout = self.impl.get_timeout()
+        if self.impl.is_running and actual_timeout < self.config.loop_wait:
+            logger.error('loop_wait of {0} seconds is too long for watchdog {1} second timeout'
+                         .format(self.config.loop_wait, actual_timeout))
+            if self.impl.can_be_disabled:
+                logger.info('Disabling watchdog due to unsafe timeout.')
+                self.impl.close()
+                self.impl = NullWatchdog()
+                return None
+        return actual_timeout
+
+    @synchronized
+    def disable(self):
+        self._disable()
+        self.active = False
+
+    def _disable(self):
+        try:
+            if self.impl.is_running and not self.impl.can_be_disabled:
+                # Give sysadmin some extra time to clean stuff up.
+                self.impl.keepalive()
+                logger.warning("Watchdog implementation can't be disabled. System will reboot after "
+                               "{0} seconds when watchdog times out.".format(self.impl.get_timeout()))
+            self.impl.close()
+        except WatchdogError as e:
+            logger.error("Error while disabling watchdog: %s", e)
+
+    @synchronized
+    def keepalive(self):
+        try:
+            if self.active:
+                self.impl.keepalive()
+            # In case there are any pending configuration changes apply them now.
+            if self.active and self.config != self.active_config:
+                if self.config.mode != MODE_OFF and self.active_config.mode == MODE_OFF:
+                    self.impl = self.config.get_impl()
+                    self._activate()
+                if self.config.driver != self.active_config.driver \
+                   or self.config.driver_config != self.active_config.driver_config:
+                    self._disable()
+                    self.impl = self.config.get_impl()
+                    self._activate()
+                if self.config.timeout != self.active_config.timeout:
+                    self.impl.set_timeout(self.config.timeout)
+        except WatchdogError as e:
+            logger.error("Error while sending keepalive: %s", e)
+
+    @property
+    @synchronized
+    def is_running(self):
+        return self.impl.is_running
+
+    @property
+    @synchronized
+    def is_healthy(self):
+        if self.config.mode != MODE_REQUIRED:
+            return True
+        return self.config.timing_slack >= 0 and self.impl.is_healthy
+
+
+@six.add_metaclass(abc.ABCMeta)
+class WatchdogBase(object):
+    """A watchdog object when opened requires periodic calls to keepalive.
+    When keepalive is not called within a timeout the system will be terminated."""
+    is_null = False
+
+    @property
+    def is_running(self):
+        """Returns True when watchdog is activated and capable of performing it's task."""
+        return False
+
+    @property
+    def is_healthy(self):
+        """Returns False when calling open() is known to fail."""
+        return False
+
+    @property
+    def can_be_disabled(self):
+        """Returns True when watchdog will be disabled by calling close(). Some watchdog devices
+        will keep running no matter what once activated. May raise WatchdogError if called without
+        calling open() first."""
+        return True
+
+    @abc.abstractmethod
+    def open(self):
+        """Open watchdog device.
+
+        When watchdog is opened keepalive must be called. Returns nothing on success
+        or raises WatchdogError if the device could not be opened."""
+
+    @abc.abstractmethod
+    def close(self):
+        """Gracefully close watchdog device."""
+
+    @abc.abstractmethod
+    def keepalive(self):
+        """Resets the watchdog timer.
+
+        Watchdog must be open when keepalive is called."""
+
+    @abc.abstractmethod
+    def get_timeout(self):
+        """Returns the current keepalive timeout in effect."""
+
+    @staticmethod
+    def has_set_timeout():
+        """Returns True if setting a timeout is supported."""
+        return False
+
+    def set_timeout(self, timeout):
+        """Set the watchdog timer timeout.
+
+        :param timeout: watchdog timeout in seconds"""
+        raise WatchdogError("Setting timeout is not supported on {0}".format(self.describe()))
+
+    def describe(self):
+        """Human readable name for this device"""
+        return self.__class__.__name__
+
+    @classmethod
+    def from_config(cls, config):
+        return cls()
+
+
+class NullWatchdog(WatchdogBase):
+    """Null implementation when watchdog is not supported."""
+    is_null = True
+
+    def open(self):
+        return
+
+    def close(self):
+        return
+
+    def keepalive(self):
+        return
+
+    def get_timeout(self):
+        # A big enough number to not matter
+        return 1000000000

patroni/watchdog/linux.py

@@ -0,0 +1,235 @@
+import collections
+import ctypes
+import os
+import platform
+from patroni.watchdog.base import WatchdogBase, WatchdogError
+
+# Pythonification of linux/ioctl.h
+IOC_NONE = 0
+IOC_WRITE = 1
+IOC_READ = 2
+
+IOC_NRBITS = 8
+IOC_TYPEBITS = 8
+IOC_SIZEBITS = 14
+IOC_DIRBITS = 2
+
+# Non-generic platform special cases
+machine = platform.machine()
+if machine in ['mips', 'sparc', 'powerpc', 'ppc64']:
+    IOC_SIZEBITS = 13
+    IOC_DIRBITS = 3
+    IOC_NONE, IOC_WRITE, IOC_READ = 1, 2, 4
+elif machine == 'parisc':
+    IOC_WRITE, IOC_READ = 2, 1
+
+IOC_NRSHIFT = 0
+IOC_TYPESHIFT = IOC_NRSHIFT + IOC_NRBITS
+IOC_SIZESHIFT = IOC_TYPESHIFT + IOC_TYPEBITS
+IOC_DIRSHIFT = IOC_SIZESHIFT + IOC_SIZEBITS
+
+
+def IOW(type_, nr, size):
+    return IOC(IOC_WRITE, type_, nr, size)
+
+
+def IOR(type_, nr, size):
+    return IOC(IOC_READ, type_, nr, size)
+
+
+def IOWR(type_, nr, size):
+    return IOC(IOC_READ | IOC_WRITE, type_, nr, size)
+
+
+def IOC(dir_, type_, nr, size):
+    return (dir_ << IOC_DIRSHIFT) \
+        | (ord(type_) << IOC_TYPESHIFT) \
+        | (nr << IOC_NRSHIFT) \
+        | (size << IOC_SIZESHIFT)
+
+
+# Pythonification of linux/watchdog.h
+
+WATCHDOG_IOCTL_BASE = 'W'
+
+
+class watchdog_info(ctypes.Structure):
+    _fields_ = [
+        ('options', ctypes.c_uint32),           # Options the card/driver supports
+        ('firmware_version', ctypes.c_uint32),  # Firmware version of the card
+        ('identity', ctypes.c_uint8 * 32),      # Identity of the board
+    ]
+
+
+struct_watchdog_info_size = ctypes.sizeof(watchdog_info)
+int_size = ctypes.sizeof(ctypes.c_int)
+
+WDIOC_GETSUPPORT = IOR(WATCHDOG_IOCTL_BASE, 0, struct_watchdog_info_size)
+WDIOC_GETSTATUS = IOR(WATCHDOG_IOCTL_BASE, 1, int_size)
+WDIOC_GETBOOTSTATUS = IOR(WATCHDOG_IOCTL_BASE, 2, int_size)
+WDIOC_GETTEMP = IOR(WATCHDOG_IOCTL_BASE, 3, int_size)
+WDIOC_SETOPTIONS = IOR(WATCHDOG_IOCTL_BASE, 4, int_size)
+WDIOC_KEEPALIVE = IOR(WATCHDOG_IOCTL_BASE, 5, int_size)
+WDIOC_SETTIMEOUT = IOWR(WATCHDOG_IOCTL_BASE, 6, int_size)
+WDIOC_GETTIMEOUT = IOR(WATCHDOG_IOCTL_BASE, 7, int_size)
+WDIOC_SETPRETIMEOUT = IOWR(WATCHDOG_IOCTL_BASE, 8, int_size)
+WDIOC_GETPRETIMEOUT = IOR(WATCHDOG_IOCTL_BASE, 9, int_size)
+WDIOC_GETTIMELEFT = IOR(WATCHDOG_IOCTL_BASE, 10, int_size)
+
+
+WDIOF_UNKNOWN = -1  # Unknown flag error
+WDIOS_UNKNOWN = -1  # Unknown status error
+
+WDIOF = {
+    "OVERHEAT": 0x0001,       # Reset due to CPU overheat
+    "FANFAULT": 0x0002,       # Fan failed
+    "EXTERN1": 0x0004,        # External relay 1
+    "EXTERN2": 0x0008,        # External relay 2
+    "POWERUNDER": 0x0010,     # Power bad/power fault
+    "CARDRESET": 0x0020,      # Card previously reset the CPU
+    "POWEROVER": 0x0040,      # Power over voltage
+    "SETTIMEOUT": 0x0080,     # Set timeout (in seconds)
+    "MAGICCLOSE": 0x0100,     # Supports magic close char
+    "PRETIMEOUT": 0x0200,     # Pretimeout (in seconds), get/set
+    "ALARMONLY": 0x0400,      # Watchdog triggers a management or other external alarm not a reboot
+    "KEEPALIVEPING": 0x8000,  # Keep alive ping reply
+}
+
+WDIOS = {
+    "DISABLECARD": 0x0001,    # Turn off the watchdog timer
+    "ENABLECARD": 0x0002,     # Turn on the watchdog timer
+    "TEMPPANIC": 0x0004,      # Kernel panic on temperature trip
+}
+
+# Implementation
+
+
+class WatchdogInfo(collections.namedtuple('WatchdogInfo', 'options,version,identity')):
+    """Watchdog descriptor from the kernel"""
+    def __getattr__(self, name):
+        """Convenience has_XYZ attributes for checking WDIOF bits in options"""
+        if name.startswith('has_') and name[4:] in WDIOF:
+            return bool(self.options & WDIOF[name[4:]])
+
+        raise AttributeError("WatchdogInfo instance has no attribute '{0}'".format(name))
+
+
+class LinuxWatchdogDevice(WatchdogBase):
+    DEFAULT_DEVICE = '/dev/watchdog'
+
+    def __init__(self, device):
+        self.device = device
+        self._support_cache = None
+        self._fd = None
+
+    @classmethod
+    def from_config(cls, config):
+        device = config.get('device', cls.DEFAULT_DEVICE)
+        return cls(device)
+
+    @property
+    def is_running(self):
+        return self._fd is not None
+
+    @property
+    def is_healthy(self):
+        return os.path.exists(self.device) and os.access(self.device, os.W_OK)
+
+    def open(self):
+        try:
+            self._fd = os.open(self.device, os.O_WRONLY)
+        except OSError as e:
+            raise WatchdogError("Can't open watchdog device: {0}".format(e))
+
+    def close(self):
+        if self.is_running:
+            try:
+                os.write(self._fd, b'V')
+                os.close(self._fd)
+                self._fd = None
+            except OSError as e:
+                raise WatchdogError("Error while closing {0}: {1}".format(self.describe(), e))
+
+    @property
+    def can_be_disabled(self):
+        return self.get_support().has_MAGICCLOSE
+
+    def _ioctl(self, func, arg):
+        """Runs the specified ioctl on the underlying fd.
+
+        Raises WatchdogError if the device is closed.
+        Raises OSError or IOError (Python 2) when the ioctl fails."""
+        if self._fd is None:
+            raise WatchdogError("Watchdog device is closed")
+        if os.name != 'nt':
+            import fcntl
+            fcntl.ioctl(self._fd, func, arg, True)
+
+    def get_support(self):
+        if self._support_cache is None:
+            info = watchdog_info()
+            try:
+                self._ioctl(WDIOC_GETSUPPORT, info)
+            except (WatchdogError, OSError, IOError) as e:
+                raise WatchdogError("Could not get information about watchdog device: {}".format(e))
+            self._support_cache = WatchdogInfo(info.options,
+                                               info.firmware_version,
+                                               bytearray(info.identity).decode(errors='ignore').rstrip('\x00'))
+        return self._support_cache
+
+    def describe(self):
+        dev_str = " at {0}".format(self.device) if self.device != self.DEFAULT_DEVICE else ""
+        ver_str = ""
+        identity = "Linux watchdog device"
+        if self._fd:
+            try:
+                _, version, identity = self.get_support()
+                ver_str = " (firmware {0})".format(version) if version else ""
+            except WatchdogError:
+                pass
+
+        return identity + ver_str + dev_str
+
+    def keepalive(self):
+        try:
+            os.write(self._fd, b'1')
+        except OSError as e:
+            raise WatchdogError("Could not send watchdog keepalive: {0}".format(e))
+
+    def has_set_timeout(self):
+        """Returns True if setting a timeout is supported."""
+        return self.get_support().has_SETTIMEOUT
+
+    def set_timeout(self, timeout):
+        timeout = int(timeout)
+        if not 0 < timeout < 0xFFFF:
+            raise WatchdogError("Invalid timeout {0}. Supported values are between 1 and 65535".format(timeout))
+        try:
+            self._ioctl(WDIOC_SETTIMEOUT, ctypes.c_int(timeout))
+        except (WatchdogError, OSError, IOError) as e:
+            raise WatchdogError("Could not set timeout on watchdog device: {}".format(e))
+
+    def get_timeout(self):
+        timeout = ctypes.c_int()
+        try:
+            self._ioctl(WDIOC_GETTIMEOUT, timeout)
+        except (WatchdogError, OSError, IOError) as e:
+            raise WatchdogError("Could not get timeout on watchdog device: {}".format(e))
+        return timeout.value
+
+
+class TestingWatchdogDevice(LinuxWatchdogDevice):
+    """Converts timeout ioctls to regular writes that can be intercepted from a named pipe."""
+    timeout = 60
+
+    def get_support(self):
+        return WatchdogInfo(WDIOF['MAGICCLOSE'] | WDIOF['SETTIMEOUT'], 0, "Watchdog test harness")
+
+    def set_timeout(self, timeout):
+        buf = "Ctimeout={0}\n".format(timeout).encode('utf8')
+        while len(buf):
+            buf = buf[os.write(self._fd, buf):]
+        self.timeout = timeout
+
+    def get_timeout(self):
+        return self.timeout

postgres0.yml

@@ -11,6 +11,11 @@ restapi:
 #    username: username
 #    password: password
 
+# ctl:
+#   insecure: false # Allow connections to SSL sites without certs
+#   certfile: /etc/ssl/certs/ssl-cert-snakeoil.pem
+#   cacert: /etc/ssl/certs/ssl-cacert-snakeoil.pem
+
 etcd:
   host: 127.0.0.1:2379
 
@@ -24,6 +29,10 @@ bootstrap:
     maximum_lag_on_failover: 1048576
 #    master_start_timeout: 300
 #    synchronous_mode: false
+    #standby_cluster:
+      #host: 127.0.0.1
+      #port: 1111
+      #primary_slot_name: patroni
     postgresql:
       use_pg_rewind: true
 #      use_slots: true
@@ -31,8 +40,8 @@ bootstrap:
 #        wal_level: hot_standby
 #        hot_standby: "on"
 #        wal_keep_segments: 8
-#        max_wal_senders: 5
+#        max_wal_senders: 10
-#        max_replication_slots: 5
+#        max_replication_slots: 10
 #        wal_log_hints: "on"
 #        archive_mode: "on"
 #        archive_timeout: 1800s
@@ -46,6 +55,9 @@ bootstrap:
   - data-checksums
 
   pg_hba:  # Add following lines to pg_hba.conf after running 'initdb'
+  # For kerberos gss based connectivity (discard @.*$)
+  #- host replication replicator 127.0.0.1/32 gss include_realm=0
+  #- host all all 0.0.0.0/0 gss include_realm=0
   - host replication replicator 127.0.0.1/32 md5
   - host all all 0.0.0.0/0 md5
 #  - hostssl all all 0.0.0.0/0 md5
@@ -66,6 +78,7 @@ postgresql:
   connect_address: 127.0.0.1:5432
   data_dir: data/postgresql0
 #  bin_dir:
+#  config_dir:
   pgpass: /tmp/pgpass0
   authentication:
     replication:
@@ -74,8 +87,22 @@ postgresql:
     superuser:
       username: postgres
       password: zalando
+    rewind:  # Has no effect on postgres 10 and lower
+      username: rewind_user
+      password: rewind_password
+  # Server side kerberos spn
+#  krbsrvname: postgres
   parameters:
+    # Fully qualified kerberos ticket file for the running user
+    # same as KRB5CCNAME used by the GSS
+#   krb_server_keyfile: /var/spool/keytabs/postgres
     unix_socket_directories: '.'
+
+#watchdog:
+#  mode: automatic # Allowed values: off, automatic, required
+#  device: /dev/watchdog
+#  safety_margin: 5
+
 tags:
     nofailover: false
     noloadbalance: false

postgres1.yml

@@ -11,6 +11,11 @@ restapi:
 #    username: username
 #    password: password
 
+# ctl:
+#   insecure: false # Allow connections to SSL sites without certs
+#   certfile: /etc/ssl/certs/ssl-cert-snakeoil.pem
+#   cacert: /etc/ssl/certs/ssl-cacert-snakeoil.pem
+
 etcd:
   host: 127.0.0.1:2379
 
@@ -29,8 +34,8 @@ bootstrap:
 #        wal_level: hot_standby
 #        hot_standby: "on"
 #        wal_keep_segments: 8
-#        max_wal_senders: 5
+#        max_wal_senders: 10
-#        max_replication_slots: 5
+#        max_replication_slots: 10
 #        wal_log_hints: "on"
 #        archive_mode: "on"
 #        archive_timeout: 1800s
@@ -44,6 +49,9 @@ bootstrap:
   - data-checksums
 
   pg_hba:  # Add following lines to pg_hba.conf after running 'initdb'
+  # For kerberos gss based connectivity (discard @.*$)
+  #- host replication replicator 127.0.0.1/32 gss include_realm=0
+  #- host all all 0.0.0.0/0 gss include_realm=0
   - host replication replicator 127.0.0.1/32 md5
   - host all all 0.0.0.0/0 md5
 #  - hostssl all all 0.0.0.0/0 md5
@@ -64,6 +72,7 @@ postgresql:
   connect_address: 127.0.0.1:5433
   data_dir: data/postgresql1
 #  bin_dir:
+#  config_dir:
   pgpass: /tmp/pgpass1
   authentication:
     replication:
@@ -72,8 +81,19 @@ postgresql:
     superuser:
       username: postgres
       password: zalando
+    rewind:  # Has no effect on postgres 10 and lower
+      username: rewind_user
+      password: rewind_password
+  # Server side kerberos spn
+#  krbsrvname: postgres
   parameters:
+    # Fully qualified kerberos ticket file for the running user
+    # same as KRB5CCNAME used by the GSS
+#    krb_server_keyfile: /var/spool/keytabs/postgres
     unix_socket_directories: '.'
+  basebackup:
+      - verbose
+      - max-rate: 100M
 tags:
     nofailover: false
     noloadbalance: false

postgres2.yml

@@ -11,6 +11,11 @@ restapi:
     username: username
     password: password
 
+# ctl:
+#   insecure: false # Allow connections to SSL sites without certs
+#   certfile: /etc/ssl/certs/ssl-cert-snakeoil.pem
+#   cacert: /etc/ssl/certs/ssl-cacert-snakeoil.pem
+
 etcd:
   host: 127.0.0.1:2379
 
@@ -29,8 +34,8 @@ bootstrap:
 #        wal_level: hot_standby
 #        hot_standby: "on"
 #        wal_keep_segments: 8
-#        max_wal_senders: 5
+#        max_wal_senders: 10
-#        max_replication_slots: 5
+#        max_replication_slots: 10
 #        wal_log_hints: "on"
 #        archive_mode: "on"
 #        archive_timeout: 1800s
@@ -44,6 +49,9 @@ bootstrap:
   - data-checksums
 
   pg_hba:  # Add following lines to pg_hba.conf after running 'initdb'
+  # For kerberos gss based connectivity (discard @.*$)
+  #- host replication replicator 127.0.0.1/32 gss include_realm=0
+  #- host all all 0.0.0.0/0 gss include_realm=0
   - host replication replicator 127.0.0.1/32 md5
   - host all all 0.0.0.0/0 md5
 #  - hostssl all all 0.0.0.0/0 md5
@@ -61,6 +69,7 @@ postgresql:
   connect_address: 127.0.0.1:5434
   data_dir: data/postgresql2
 #  bin_dir:
+#  config_dir:
   pgpass: /tmp/pgpass2
   authentication:
     replication:
@@ -69,7 +78,15 @@ postgresql:
     superuser:
       username: postgres
       password: zalando
+    rewind:  # Has no effect on postgres 10 and lower
+      username: rewind_user
+      password: rewind_password
+  # Server side kerberos spn
+#  krbsrvname: postgres
   parameters:
+    # Fully qualified kerberos ticket file for the running user
+    # same as KRB5CCNAME used by the GSS
+#    krb_server_keyfile: /var/spool/keytabs/postgres
     unix_socket_directories: '.'
 tags:
     nofailover: false

requirements.txt

@@ -1,13 +1,15 @@
-urllib3>=1.9
+urllib3>=1.19.1,!=1.21
 boto
-psycopg2>=2.6.1
 PyYAML
 requests
 six >= 1.7
-kazoo==2.2.1
+kazoo>=1.3.1
-python-etcd==0.4.3
+python-etcd>=0.4.3,<0.5
-python-consul==0.7.0
+python-consul>=0.7.0
 click>=4.1
 prettytable>=0.7
 tzlocal
 python-dateutil
+psutil>=2.0.0
+cdiff
+kubernetes>=2.0.0,<=7.0.0,!=4.0.*,!=5.0.*

setup.py

@@ -8,33 +8,29 @@ import inspect
 import os
 import sys
 
+from patroni import check_psycopg2, fatal
+from patroni.version import __version__ as VERSION
 from setuptools.command.test import test as TestCommand
 from setuptools import find_packages, setup
 
 if sys.version_info < (2, 7, 0):
-    sys.stderr.write('FATAL: patroni needs to be run with Python 2.7+\n')
+    fatal('patroni needs to be run with Python 2.7+')
-    sys.exit(1)
+check_psycopg2()
+del sys.modules['patroni']
+del sys.modules['patroni.version']
 
 __location__ = os.path.join(os.getcwd(), os.path.dirname(inspect.getfile(inspect.currentframe())))
 
-
-def read_version(package):
-    data = {}
-    with open(os.path.join(package, 'version.py'), 'r') as fd:
-        exec(fd.read(), data)
-    return data['__version__']
-
-
 NAME = 'patroni'
 MAIN_PACKAGE = NAME
 SCRIPTS = 'scripts'
-VERSION = read_version(MAIN_PACKAGE)
 DESCRIPTION = 'PostgreSQL High-Available orchestrator and CLI'
 LICENSE = 'The MIT License'
 URL = 'https://github.com/zalando/patroni'
-AUTHOR = 'Alexander Kukushkin, Oleksii Kliukin, Feike Steenbergen'
+AUTHOR = 'Alexander Kukushkin, Dmitrii Dolgov, Oleksii Kliukin'
-AUTHOR_EMAIL = '[email protected], [email protected], [email protected]'
+AUTHOR_EMAIL = '[email protected], [email protected], [email protected]'
-KEYWORDS = 'etcd governor patroni postgresql postgres ha haproxy confd zookeeper exhibitor consul streaming replication'
+KEYWORDS = 'etcd governor patroni postgresql postgres ha haproxy confd' +\
+    ' zookeeper exhibitor consul streaming replication kubernetes k8s'
 
 COVERAGE_XML = True
 COVERAGE_HTML = False
@@ -43,16 +39,22 @@ JUNIT_XML = True
 # Add here all kinds of additional classifiers as defined under
 # https://pypi.python.org/pypi?%3Aaction=list_classifiers
 CLASSIFIERS = [
-    'Development Status :: 4 - Beta',
+    'Development Status :: 5 - Production/Stable',
     'Environment :: Console',
     'Intended Audience :: Developers',
     'Intended Audience :: System Administrators',
     'License :: OSI Approved :: MIT License',
+    'Operating System :: MacOS',
     'Operating System :: POSIX :: Linux',
+    'Operating System :: POSIX :: BSD :: FreeBSD',
+    'Operating System :: Microsoft :: Windows',
     'Programming Language :: Python',
     'Programming Language :: Python :: 2.7',
+    'Programming Language :: Python :: 3',
     'Programming Language :: Python :: 3.4',
     'Programming Language :: Python :: 3.5',
+    'Programming Language :: Python :: 3.6',
+    'Programming Language :: Python :: 3.7',
     'Programming Language :: Python :: Implementation :: CPython',
 ]
 
@@ -87,7 +89,7 @@ class PyTest(TestCommand):
     def run_tests(self):
         try:
             import pytest
-        except:
+        except Exception:
             raise RuntimeError('py.test is not installed, run: pip install pytest')
         params = {'args': self.test_args}
         if self.cov:
@@ -100,28 +102,36 @@ class PyTest(TestCommand):
         silence = logging.WARNING
         logging.basicConfig(format='%(asctime)s %(levelname)s: %(message)s', level=os.getenv('LOGLEVEL', silence))
         params['args'] += ['-s' if logging.getLogger().getEffectiveLevel() < silence else '--capture=fd']
+        if not os.getenv('SYSTEMROOT'):
+            os.environ['SYSTEMROOT'] = '/'
         errno = pytest.main(**params)
         sys.exit(errno)
 
 
-def get_install_requirements(path):
-    content = open(os.path.join(__location__, path)).read()
-    return [req for req in content.split('\n') if req != '']
-
-
 def read(fname):
-    return open(os.path.join(__location__, fname)).read()
+    with open(os.path.join(__location__, fname)) as fd:
+        return fd.read()
 
 
 def setup_package():
     # Assemble additional setup commands
-    cmdclass = {}
+    cmdclass = {'test': PyTest}
-    cmdclass['test'] = PyTest
 
-    # Some helper variables
+    install_requires = []
-    version = os.getenv('GO_PIPELINE_LABEL', VERSION)
+    extras_require = {'aws': ['boto'], 'etcd': ['python-etcd'], 'consul': ['python-consul'],
+                      'exhibitor': ['kazoo'], 'zookeeper': ['kazoo'], 'kubernetes': ['kubernetes']}
 
-    install_reqs = get_install_requirements('requirements.txt')
+    for r in read('requirements.txt').split('\n'):
+        r = r.strip()
+        if r == '':
+            continue
+        extra = False
+        for e, v in extras_require.items():
+            if r.startswith(v[0]):
+                extras_require[e] = [r]
+                extra = True
+        if not extra:
+            install_requires.append(r)
 
     command_options = {'test': {'test_suite': ('setup.py', 'tests')}}
     if JUNIT_XML:
@@ -133,7 +143,7 @@ def setup_package():
 
     setup(
         name=NAME,
-        version=version,
+        version=VERSION,
         url=URL,
         author=AUTHOR,
         author_email=AUTHOR_EMAIL,
@@ -145,10 +155,10 @@ def setup_package():
         test_suite='tests',
         packages=find_packages(exclude=['tests', 'tests.*']),
         package_data={MAIN_PACKAGE: ["*.json"]},
-        install_requires=install_reqs,
+        install_requires=install_requires,
-        setup_requires=['flake8'],
+        extras_require=extras_require,
         cmdclass=cmdclass,
-        tests_require=['mock>=2.0.0', 'pytest-cov', 'pytest'],
+        tests_require=['flake8', 'mock>=2.0.0', 'pytest-cov', 'pytest'],
         command_options=command_options,
         entry_points={'console_scripts': CONSOLE_SCRIPTS},
     )

tests/__init__.py

@@ -0,0 +1,216 @@
+import datetime
+import json
+import os
+import shutil
+import unittest
+
+from mock import Mock, patch
+from tempfile import gettempdir
+
+import psycopg2
+import requests
+
+from patroni.dcs import Leader, Member
+from patroni.postgresql import Postgresql
+from patroni.postgresql.config import ConfigHandler
+from patroni.utils import RetryFailedError
+
+
+class SleepException(Exception):
+    pass
+
+
+class MockResponse(object):
+
+    def __init__(self, status_code=200):
+        self.status_code = status_code
+        self.content = '{}'
+        self.ok = True
+
+    def json(self):
+        return json.loads(self.content)
+
+    @property
+    def data(self):
+        return self.content.encode('utf-8')
+
+    @property
+    def text(self):
+        return self.content
+
+    @property
+    def status(self):
+        return self.status_code
+
+    @staticmethod
+    def getheader(*args):
+        return ''
+
+
+def requests_get(url, **kwargs):
+    members = '[{"id":14855829450254237642,"peerURLs":["http://localhost:2380","http://localhost:7001"],' +\
+              '"name":"default","clientURLs":["http://localhost:2379","http://localhost:4001"]}]'
+    response = MockResponse()
+    if url.startswith('http://local'):
+        raise requests.exceptions.RequestException()
+    elif ':8011/patroni' in url:
+        response.content = '{"role": "replica", "xlog": {"received_location": 0}, "tags": {}}'
+    elif url.endswith('/members'):
+        response.content = '[{}]' if url.startswith('http://error') else members
+    elif url.startswith('http://exhibitor'):
+        response.content = '{"servers":["127.0.0.1","127.0.0.2","127.0.0.3"],"port":2181}'
+    elif url.endswith(':8011/reinitialize'):
+        data = kwargs.get('data', '')
+        if ' false}' in data:
+            response.status_code = 503
+            response.ok = False
+            response.content = 'restarting after failure already in progress'
+    else:
+        response.status_code = 404
+        response.ok = False
+    return response
+
+
+class MockPostmaster(object):
+    def __init__(self, is_running=True, is_single_master=False):
+        self.is_running = Mock(return_value=is_running)
+        self.is_single_master = Mock(return_value=is_single_master)
+        self.wait_for_user_backends_to_close = Mock()
+        self.signal_stop = Mock(return_value=None)
+        self.wait = Mock()
+
+
+class MockCursor(object):
+
+    def __init__(self, connection):
+        self.connection = connection
+        self.closed = False
+        self.rowcount = 0
+        self.results = []
+
+    def execute(self, sql, *params):
+        if sql.startswith('blabla'):
+            raise psycopg2.ProgrammingError()
+        elif sql == 'CHECKPOINT' or sql.startswith('SELECT pg_catalog.pg_create_'):
+            raise psycopg2.OperationalError()
+        elif sql.startswith('RetryFailedError'):
+            raise RetryFailedError('retry')
+        elif sql.startswith('SELECT slot_name'):
+            self.results = [('blabla', 'physical'), ('foobar', 'physical'), ('ls', 'logical', 'a', 'b')]
+        elif sql.startswith('SELECT CASE WHEN pg_catalog.pg_is_in_recovery()'):
+            self.results = [(1, 2)]
+        elif sql.startswith('SELECT pg_catalog.pg_is_in_recovery()'):
+            self.results = [(False, 2)]
+        elif sql.startswith('WITH replication_info AS ('):
+            replication_info = '[{"application_name":"walreceiver","client_addr":"1.2.3.4",' +\
+                               '"state":"streaming","sync_state":"async","sync_priority":0}]'
+            self.results = [('', 0, '', '', '', '', False, replication_info)]
+        elif sql.startswith('SELECT name, setting'):
+            self.results = [('wal_segment_size', '2048', '8kB', 'integer', 'internal'),
+                            ('search_path', 'public', None, 'string', 'user'),
+                            ('port', '5433', None, 'integer', 'postmaster'),
+                            ('listen_addresses', '*', None, 'string', 'postmaster'),
+                            ('autovacuum', 'on', None, 'bool', 'sighup'),
+                            ('unix_socket_directories', '/tmp', None, 'string', 'postmaster')]
+        elif sql.startswith('IDENTIFY_SYSTEM'):
+            self.results = [('1', 2, '0/402EEC0', '')]
+        elif sql.startswith('SELECT isdir, modification'):
+            self.results = [(False, datetime.datetime.now())]
+        elif sql.startswith('SELECT pg_catalog.pg_read_file'):
+            self.results = [('1\t0/40159C0\tno recovery target specified\n\n'
+                             '2\t1/40159C0\tno recovery target specified\n',)]
+        elif sql.startswith('TIMELINE_HISTORY '):
+            self.results = [('', b'x\t0/40159C0\tno recovery target specified\n\n'
+                                 b'1\t0/40159C0\tno recovery target specified\n\n'
+                                 b'2\t0/402DD98\tno recovery target specified\n\n'
+                                 b'3\t0/403DD98\tno recovery target specified\n')]
+        else:
+            self.results = [(None, None, None, None, None, None, None, None, None, None)]
+
+    def fetchone(self):
+        return self.results[0]
+
+    def fetchall(self):
+        return self.results
+
+    def __iter__(self):
+        for i in self.results:
+            yield i
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *args):
+        pass
+
+
+class MockConnect(object):
+
+    server_version = 99999
+    autocommit = False
+    closed = 0
+
+    def cursor(self):
+        return MockCursor(self)
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, *args):
+        pass
+
+    @staticmethod
+    def close():
+        pass
+
+
+def psycopg2_connect(*args, **kwargs):
+    return MockConnect()
+
+
+class PostgresInit(unittest.TestCase):
+    _PARAMETERS = {'wal_level': 'hot_standby', 'max_replication_slots': 5, 'f.oo': 'bar',
+                   'search_path': 'public', 'hot_standby': 'on', 'max_wal_senders': 5,
+                   'wal_keep_segments': 8, 'wal_log_hints': 'on', 'max_locks_per_transaction': 64,
+                   'max_worker_processes': 8, 'max_connections': 100, 'max_prepared_transactions': 0,
+                   'track_commit_timestamp': 'off', 'unix_socket_directories': '/tmp', 'trigger_file': 'bla'}
+
+    @patch('psycopg2.connect', psycopg2_connect)
+    @patch.object(ConfigHandler, 'write_postgresql_conf', Mock())
+    @patch.object(ConfigHandler, 'replace_pg_hba', Mock())
+    @patch.object(ConfigHandler, 'replace_pg_ident', Mock())
+    def setUp(self):
+        data_dir = 'data/test0'
+        self.p = Postgresql({'name': 'postgresql0', 'scope': 'batman', 'data_dir': data_dir,
+                             'config_dir': data_dir, 'retry_timeout': 10,
+                             'krbsrvname': 'postgres', 'pgpass': os.path.join(gettempdir(), 'pgpass0'),
+                             'listen': '127.0.0.2, 127.0.0.3:5432', 'connect_address': '127.0.0.2:5432',
+                             'authentication': {'superuser': {'username': 'foo', 'password': 'test'},
+                                                'replication': {'username': '', 'password': 'rep-pass'}},
+                             'remove_data_directory_on_rewind_failure': True,
+                             'use_pg_rewind': True, 'pg_ctl_timeout': 'bla',
+                             'parameters': self._PARAMETERS,
+                             'recovery_conf': {'foo': 'bar'},
+                             'pg_hba': ['host all all 0.0.0.0/0 md5'],
+                             'pg_ident': ['krb realm postgres'],
+                             'callbacks': {'on_start': 'true', 'on_stop': 'true', 'on_reload': 'true',
+                                           'on_restart': 'true', 'on_role_change': 'true'}})
+
+
+class BaseTestPostgresql(PostgresInit):
+
+    def setUp(self):
+        super(BaseTestPostgresql, self).setUp()
+
+        if not os.path.exists(self.p.data_dir):
+            os.makedirs(self.p.data_dir)
+
+        self.leadermem = Member(0, 'leader', 28, {'conn_url': 'postgres://replicator:[email protected]:5435/postgres'})
+        self.leader = Leader(-1, 28, self.leadermem)
+        self.other = Member(0, 'test-1', 28, {'conn_url': 'postgres://replicator:[email protected]:5433/postgres',
+                                              'tags': {'replicatefrom': 'leader'}})
+        self.me = Member(0, 'test0', 28, {'conn_url': 'postgres://replicator:[email protected]:5434/postgres'})
+
+    def tearDown(self):
+        if os.path.exists(self.p.data_dir):
+            shutil.rmtree(self.p.data_dir)

tests/test_api.py

@@ -2,14 +2,16 @@ import datetime
 import json
 import psycopg2
 import unittest
+import socket
 
-from mock import Mock, patch
+from mock import Mock, PropertyMock, patch
 from patroni.api import RestApiHandler, RestApiServer
 from patroni.dcs import ClusterConfig, Member
+from patroni.ha import _MemberStatus
 from patroni.utils import tzutc
 from six import BytesIO as IO
 from six.moves import BaseHTTPServer
-from test_postgresql import psycopg2_connect, MockCursor
+from . import psycopg2_connect, MockCursor
 
 
 future_restart_time = datetime.datetime.now(tzutc) + datetime.timedelta(days=5)
@@ -25,6 +27,8 @@ class MockPostgresql(object):
     sysid = 'dummysysid'
     scope = 'dummy'
     pending_restart = True
+    wal_name = 'wal'
+    lsn_name = 'lsn'
 
     @staticmethod
     def connection():
@@ -34,17 +38,30 @@ class MockPostgresql(object):
     def postmaster_start_time():
         return str(postmaster_start_time)
 
+    @staticmethod
+    def replica_cached_timeline(_):
+        return 2
+
+
+class MockWatchdog(object):
+    is_healthy = False
+
 
 class MockHa(object):
 
     state_handler = MockPostgresql()
+    watchdog = MockWatchdog()
 
     @staticmethod
-    def reinitialize():
+    def is_leader():
+        return False
+
+    @staticmethod
+    def reinitialize(_):
         return 'reinitialize'
 
     @staticmethod
-    def restart():
+    def restart(*args, **kwargs):
         return (True, '')
 
     @staticmethod
@@ -57,14 +74,14 @@ class MockHa(object):
 
     @staticmethod
     def fetch_nodes_statuses(members):
-        return [[None, True, None, None, {}]]
+        return [_MemberStatus(None, True, None, 0, None, {}, False)]
 
     @staticmethod
     def schedule_future_restart(data):
         return True
 
     @staticmethod
-    def is_lagging(xlog):
+    def is_lagging(wal):
         return False
 
     @staticmethod
@@ -75,6 +92,21 @@ class MockHa(object):
     def wakeup():
         pass
 
+    @staticmethod
+    def is_paused():
+        return True
+
+    @staticmethod
+    def is_standby_cluster():
+        return False
+
+
+class MockLogger(object):
+
+    NORMAL_LOG_QUEUE_SIZE = 2
+    queue_size = 3
+    records_lost = 1
+
 
 class MockPatroni(object):
 
@@ -82,9 +114,10 @@ class MockPatroni(object):
     config = Mock()
     postgresql = ha.state_handler
     dcs = Mock()
+    logger = MockLogger()
     tags = {}
     version = '0.00'
-    noloadbalance = Mock(return_value=False)
+    noloadbalance = PropertyMock(return_value=False)
     scheduled_restart = {'schedule': future_restart_time,
                          'postmaster_start_time': postgresql.postmaster_start_time()}
 
@@ -101,23 +134,25 @@ class MockRequest(object):
     def makefile(self, *args, **kwargs):
         return IO(self.request)
 
+    def sendall(self, *args, **kwargs):
+        pass
+
 
 class MockRestApiServer(RestApiServer):
 
-    def __init__(self, Handler, request):
+    def __init__(self, Handler, request, config=None):
         self.socket = 0
         self.serve_forever = Mock()
-        BaseHTTPServer.HTTPServer.__init__ = Mock()
         MockRestApiServer._BaseServer__is_shut_down = Mock()
         MockRestApiServer._BaseServer__shutdown_request = True
-        config = {'listen': '127.0.0.1:8008', 'auth': 'test:test'}
+        config = config or {'listen': '127.0.0.1:8008', 'auth': 'test:test', 'certfile': 'dumb'}
         super(MockRestApiServer, self).__init__(MockPatroni(), config)
-        config['certfile'] = 'dumb'
-        self.reload_config(config)
         Handler(MockRequest(request), ('0.0.0.0', 8080), self)
 
 
-@patch('ssl.wrap_socket', Mock(return_value=0))
+@patch('ssl.SSLContext.load_cert_chain', Mock())
+@patch('ssl.SSLContext.wrap_socket', Mock(return_value=0))
+@patch.object(BaseHTTPServer.HTTPServer, '__init__', Mock())
 class TestRestApiHandler(unittest.TestCase):
 
     _authorization = '\nAuthorization: Basic dGVzdDp0ZXN0'
@@ -128,8 +163,17 @@ class TestRestApiHandler(unittest.TestCase):
             MockRestApiServer(RestApiHandler, 'GET /replica')
         with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'master'})):
             MockRestApiServer(RestApiHandler, 'GET /replica')
+        with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'state': 'running'})):
+            MockRestApiServer(RestApiHandler, 'GET /health')
         MockRestApiServer(RestApiHandler, 'GET /master')
-        MockPatroni.dcs.cluster.leader.name = MockPostgresql.name
+        MockPatroni.dcs.cluster.sync.sync_standby = MockPostgresql.name
+        MockPatroni.dcs.cluster.is_synchronous_mode = Mock(return_value=True)
+        with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'replica'})):
+            MockRestApiServer(RestApiHandler, 'GET /synchronous')
+        with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'replica'})):
+            MockPatroni.dcs.cluster.sync.sync_standby = ''
+            MockRestApiServer(RestApiHandler, 'GET /asynchronous')
+        MockPatroni.ha.is_leader = Mock(return_value=True)
         MockRestApiServer(RestApiHandler, 'GET /replica')
         MockPatroni.dcs.cluster = None
         with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'master'})):
@@ -137,10 +181,15 @@ class TestRestApiHandler(unittest.TestCase):
         with patch.object(MockHa, 'restart_scheduled', Mock(return_value=True)):
             MockRestApiServer(RestApiHandler, 'GET /master')
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'GET /master'))
+        with patch.object(RestApiServer, 'query', Mock(return_value=[('', 1, '', '', '', '', False, '')])):
+            self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'GET /patroni'))
+        with patch.object(MockHa, 'is_standby_cluster', Mock(return_value=True)):
+            MockRestApiServer(RestApiHandler, 'GET /standby_leader')
 
     def test_do_OPTIONS(self):
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'OPTIONS / HTTP/1.0'))
 
+    @patch.object(MockPostgresql, 'state', PropertyMock(return_value='stopped'))
     def test_do_GET_patroni(self):
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'GET /patroni'))
 
@@ -252,7 +301,7 @@ class TestRestApiHandler(unittest.TestCase):
     def test_do_POST_reinitialize(self, mock_dcs):
         cluster = mock_dcs.get_cluster.return_value
         cluster.is_paused.return_value = False
-        request = 'POST /reinitialize HTTP/1.0' + self._authorization
+        request = 'POST /reinitialize HTTP/1.0' + self._authorization + '\nContent-Length: 15\n\n{"force": true}'
         MockRestApiServer(RestApiHandler, request)
         with patch.object(MockHa, 'reinitialize', Mock(return_value=None)):
             MockRestApiServer(RestApiHandler, request)
@@ -266,11 +315,13 @@ class TestRestApiHandler(unittest.TestCase):
 
     @patch('time.sleep', Mock())
     @patch.object(MockPatroni, 'dcs')
-    def test_do_POST_failover(self, dcs):
+    def test_do_POST_switchover(self, dcs):
         dcs.loop_wait = 10
         cluster = dcs.get_cluster.return_value
+        cluster.is_synchronous_mode.return_value = False
+        cluster.is_paused.return_value = False
 
-        post = 'POST /failover HTTP/1.0' + self._authorization + '\nContent-Length: '
+        post = 'POST /switchover HTTP/1.0' + self._authorization + '\nContent-Length: '
 
         MockRestApiServer(RestApiHandler, post + '7\n\n{"1":2}')
 
@@ -280,14 +331,22 @@ class TestRestApiHandler(unittest.TestCase):
         cluster.leader.name = 'postgresql1'
         MockRestApiServer(RestApiHandler, request)
 
-        MockRestApiServer(RestApiHandler, post + '25\n\n{"leader": "postgresql1"}')
+        request = post + '25\n\n{"leader": "postgresql1"}'
+
+        cluster.is_paused.return_value = True
+        MockRestApiServer(RestApiHandler, request)
+
+        cluster.is_paused.return_value = False
+        for cluster.is_synchronous_mode.return_value in (True, False):
+            MockRestApiServer(RestApiHandler, request)
 
         cluster.leader.name = 'postgresql2'
         request = post + '53\n\n{"leader": "postgresql1", "candidate": "postgresql2"}'
         MockRestApiServer(RestApiHandler, request)
 
         cluster.leader.name = 'postgresql1'
-        MockRestApiServer(RestApiHandler, request)
+        for cluster.is_synchronous_mode.return_value in (True, False):
+            MockRestApiServer(RestApiHandler, request)
 
         cluster.members = [Member(0, 'postgresql0', 30, {'api_url': 'http'}),
                            Member(0, 'postgresql2', 30, {'api_url': 'http'})]
@@ -301,6 +360,7 @@ class TestRestApiHandler(unittest.TestCase):
 
         cluster2 = cluster.copy()
         cluster2.leader.name = 'postgresql0'
+        cluster2.is_unlocked.return_value = False
         dcs.get_cluster.side_effect = [cluster, cluster2]
         MockRestApiServer(RestApiHandler, request)
 
@@ -335,3 +395,30 @@ class TestRestApiHandler(unittest.TestCase):
 
         # Invalid date
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, request + '2010-02-29T18:13:30.568224+01:00"}'))
+
+    @patch.object(MockPatroni, 'dcs', Mock())
+    def test_do_POST_failover(self):
+        post = 'POST /failover HTTP/1.0' + self._authorization + '\nContent-Length: '
+        MockRestApiServer(RestApiHandler, post + '14\n\n{"leader":"1"}')
+        MockRestApiServer(RestApiHandler, post + '37\n\n{"candidate":"2","scheduled_at": "1"}')
+
+
+@patch('ssl.SSLContext.load_cert_chain', Mock())
+@patch('ssl.SSLContext.wrap_socket', Mock(return_value=0))
+@patch.object(BaseHTTPServer.HTTPServer, '__init__', Mock())
+class TestRestApiServer(unittest.TestCase):
+
+    def test_reload_config(self):
+        bad_config = {'listen': 'foo'}
+        self.assertRaises(ValueError, MockRestApiServer, None, '', bad_config)
+        srv = MockRestApiServer(lambda a1, a2, a3: None, '')
+        self.assertRaises(ValueError, srv.reload_config, bad_config)
+        self.assertRaises(ValueError, srv.reload_config, {})
+        with patch.object(socket.socket, 'setsockopt', Mock(side_effect=socket.error)):
+            srv.reload_config({'listen': ':8008'})
+
+    def test_handle_error(self):
+        try:
+            raise Exception()
+        except Exception:
+            self.assertIsNone(MockRestApiServer.handle_error(None, ('127.0.0.1', 55555)))

tests/test_async_executor.py

@@ -1,14 +1,14 @@
 import unittest
 
 from mock import Mock, patch
-from patroni.async_executor import AsyncExecutor
+from patroni.async_executor import AsyncExecutor, CriticalTask
 from threading import Thread
 
 
 class TestAsyncExecutor(unittest.TestCase):
 
     def setUp(self):
-        self.a = AsyncExecutor(Mock())
+        self.a = AsyncExecutor(Mock(), Mock())
 
     @patch.object(Thread, 'start', Mock())
     def test_run_async(self):
@@ -16,3 +16,17 @@ class TestAsyncExecutor(unittest.TestCase):
 
     def test_run(self):
         self.a.run(Mock(side_effect=Exception()))
+
+    def test_cancel(self):
+        self.a.cancel()
+        self.a.schedule('foo')
+        self.a.cancel()
+        self.a.run(Mock())
+
+
+class TestCriticalTask(unittest.TestCase):
+
+    def test_completed_task(self):
+        ct = CriticalTask()
+        ct.complete(1)
+        self.assertFalse(ct.cancel())

tests/test_aws.py

@@ -1,5 +1,4 @@
 import boto.ec2
-import requests
 import sys
 import unittest
 
@@ -11,82 +10,60 @@ from requests.exceptions import RequestException
 
 class MockEc2Connection(object):
 
-    def __init__(self, error=False):
+    @staticmethod
-        self.error = error
+    def get_all_volumes(*args, **kwargs):
-
-    def get_all_volumes(self, filters):
-        if self.error:
-            raise Exception("get_all_volumes")
         oid = namedtuple('Volume', 'id')
         return [oid(id='a'), oid(id='b')]
 
-    def create_tags(self, objects, tags):
+    @staticmethod
-        if self.error or len(objects) == 0:
+    def create_tags(objects, *args, **kwargs):
-            raise Exception("create_tags")
+        if len(objects) == 0:
+            raise boto.exception.BotoServerError(503, 'Service Unavailable', 'Request limit exceeded')
         return True
 
 
 class MockResponse(object):
+    ok = True
 
     def __init__(self, content):
         self.content = content
-        self.ok = True
 
     def json(self):
         return self.content
 
 
+def requests_get(url, **kwargs):
+    if url.split('/')[-1] == 'document':
+        result = {"instanceId": "012345", "region": "eu-west-1"}
+    else:
+        result = 'foo'
+    return MockResponse(result)
+
+
+@patch('boto.ec2.connect_to_region', Mock(return_value=MockEc2Connection()))
 class TestAWSConnection(unittest.TestCase):
 
-    def boto_ec2_connect_to_region(self, region):
+    @patch('requests.get', requests_get)
-        return MockEc2Connection(self.error)
-
-    def requests_get(self, url, **kwargs):
-        if self.error:
-            raise RequestException("foo")
-        result = namedtuple('Request', 'ok content')
-        result.ok = True
-        if url.split('/')[-1] == 'document' and not self.json_error:
-            result = {"instanceId": "012345", "region": "eu-west-1"}
-        else:
-            result = 'foo'
-        return MockResponse(result)
-
     def setUp(self):
-        self.error = False
-        self.json_error = False
-        requests.get = self.requests_get
-        boto.ec2.connect_to_region = self.boto_ec2_connect_to_region
         self.conn = AWSConnection('test')
 
-    def test_aws_available(self):
-        self.assertTrue(self.conn.aws_available())
-
     def test_on_role_change(self):
-        self.assertTrue(self.conn._tag_ebs('master'))
-        self.assertTrue(self.conn._tag_ec2('master'))
         self.assertTrue(self.conn.on_role_change('master'))
+        with patch.object(MockEc2Connection, 'get_all_volumes', Mock(return_value=[])):
+            self.conn._retry.max_tries = 1
+            self.assertFalse(self.conn.on_role_change('master'))
 
+    @patch('requests.get', Mock(side_effect=RequestException('foo')))
     def test_non_aws(self):
-        self.error = True
         conn = AWSConnection('test')
-        self.assertFalse(conn.aws_available())
+        self.assertFalse(conn.on_role_change("master"))
-        self.assertFalse(conn._tag_ebs('master'))
-        self.assertFalse(conn._tag_ec2('master'))
 
+    @patch('requests.get', Mock(return_value=MockResponse('foo')))
     def test_aws_bizare_response(self):
-        self.json_error = True
         conn = AWSConnection('test')
         self.assertFalse(conn.aws_available())
 
-    def test_aws_tag_ebs_error(self):
+    @patch('requests.get', requests_get)
-        self.error = True
-        self.assertFalse(self.conn._tag_ebs("master"))
-
-    def test_aws_tag_ec2_error(self):
-        self.error = True
-        self.assertFalse(self.conn._tag_ec2("master"))
-
     @patch('sys.exit', Mock())
     def test_main(self):
         self.assertIsNone(_main())

tests/test_bootstrap.py

@@ -0,0 +1,220 @@
+import os
+
+from mock import Mock, PropertyMock, patch
+
+from patroni.async_executor import CriticalTask
+from patroni.postgresql import Postgresql
+from patroni.postgresql.bootstrap import Bootstrap
+from patroni.postgresql.cancellable import CancellableSubprocess
+from patroni.postgresql.config import ConfigHandler
+
+from . import psycopg2_connect, BaseTestPostgresql
+
+
+@patch('subprocess.call', Mock(return_value=0))
+@patch('psycopg2.connect', psycopg2_connect)
+@patch('os.rename', Mock())
+class TestBootstrap(BaseTestPostgresql):
+
+    def setUp(self):
+        super(TestBootstrap, self).setUp()
+        self.b = self.p.bootstrap
+
+    @patch('time.sleep', Mock())
+    @patch.object(CancellableSubprocess, 'call')
+    @patch.object(Postgresql, 'remove_data_directory', Mock(return_value=True))
+    @patch.object(Postgresql, 'data_directory_empty', Mock(return_value=False))
+    @patch.object(Bootstrap, '_post_restore', Mock(side_effect=OSError))
+    def test_create_replica(self, mock_cancellable_subprocess_call):
+        self.p.config._config['create_replica_methods'] = ['pgBackRest']
+        self.p.config._config['pgBackRest'] = {'command': 'pgBackRest', 'keep_data': True, 'no_params': True}
+        mock_cancellable_subprocess_call.return_value = 0
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        self.p.config._config['create_replica_methods'] = ['basebackup']
+        self.p.config._config['basebackup'] = [{'max_rate': '100M'}, 'no-sync']
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        self.p.config._config['basebackup'] = [{'max_rate': '100M', 'compress': '9'}]
+        with patch('patroni.postgresql.bootstrap.logger.error', new_callable=Mock()) as mock_logger:
+            self.b.create_replica(self.leader)
+            mock_logger.assert_called_once()
+            self.assertTrue("only one key-value is allowed and value should be a string" in mock_logger.call_args[0][0],
+                            "not matching {0}".format(mock_logger.call_args[0][0]))
+
+        self.p.config._config['basebackup'] = [42]
+        with patch('patroni.postgresql.bootstrap.logger.error', new_callable=Mock()) as mock_logger:
+            self.b.create_replica(self.leader)
+            mock_logger.assert_called_once()
+            self.assertTrue("value should be string value or a single key-value pair" in mock_logger.call_args[0][0],
+                            "not matching {0}".format(mock_logger.call_args[0][0]))
+
+        self.p.config._config['basebackup'] = {"foo": "bar"}
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        self.p.config._config['create_replica_methods'] = ['wale', 'basebackup']
+        del self.p.config._config['basebackup']
+        mock_cancellable_subprocess_call.return_value = 1
+        self.assertEqual(self.b.create_replica(self.leader), 1)
+
+        mock_cancellable_subprocess_call.side_effect = Exception('foo')
+        self.assertEqual(self.b.create_replica(self.leader), 1)
+
+        mock_cancellable_subprocess_call.side_effect = [1, 0]
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        mock_cancellable_subprocess_call.side_effect = [Exception(), 0]
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        self.p.cancellable.cancel()
+        self.assertEqual(self.b.create_replica(self.leader), 1)
+
+    @patch('time.sleep', Mock())
+    @patch.object(CancellableSubprocess, 'call')
+    @patch.object(Postgresql, 'remove_data_directory', Mock(return_value=True))
+    @patch.object(Bootstrap, '_post_restore', Mock(side_effect=OSError))
+    def test_create_replica_old_format(self, mock_cancellable_subprocess_call):
+        """ The same test as before but with old 'create_replica_method'
+            to test backward compatibility
+        """
+        self.p.config._config['create_replica_method'] = ['wale', 'basebackup']
+        self.p.config._config['wale'] = {'command': 'foo'}
+        mock_cancellable_subprocess_call.return_value = 0
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+        del self.p.config._config['wale']
+        self.assertEqual(self.b.create_replica(self.leader), 0)
+
+        self.p.config._config['create_replica_method'] = ['wale']
+        mock_cancellable_subprocess_call.return_value = 1
+        self.assertEqual(self.b.create_replica(self.leader), 1)
+
+    def test_basebackup(self):
+        self.p.cancellable.cancel()
+        self.b.basebackup(None, None, {'foo': 'bar'})
+
+    def test__initdb(self):
+        self.assertRaises(Exception, self.b.bootstrap, {'initdb': [{'pgdata': 'bar'}]})
+        self.assertRaises(Exception, self.b.bootstrap, {'initdb': [{'foo': 'bar', 1: 2}]})
+        self.assertRaises(Exception, self.b.bootstrap, {'initdb': [1]})
+        self.assertRaises(Exception, self.b.bootstrap, {'initdb': 1})
+
+    @patch.object(CancellableSubprocess, 'call', Mock())
+    @patch.object(Postgresql, 'is_running', Mock(return_value=True))
+    @patch.object(Postgresql, 'data_directory_empty', Mock(return_value=False))
+    def test_bootstrap(self):
+        with patch('subprocess.call', Mock(return_value=1)):
+            self.assertFalse(self.b.bootstrap({}))
+
+        config = {'users': {'replicator': {'password': 'rep-pass', 'options': ['replication']}}}
+
+        with patch.object(Postgresql, 'is_running', Mock(return_value=False)),\
+                patch('multiprocessing.Process', Mock(side_effect=Exception)):
+            self.assertRaises(Exception, self.b.bootstrap, config)
+        with open(os.path.join(self.p.data_dir, 'pg_hba.conf')) as f:
+            lines = f.readlines()
+            self.assertTrue('host all all 0.0.0.0/0 md5\n' in lines)
+
+        self.p.config._config.pop('pg_hba')
+        config.update({'post_init': '/bin/false',
+                       'pg_hba': ['host replication replicator 127.0.0.1/32 md5',
+                                  'hostssl all all 0.0.0.0/0 md5',
+                                  'host all all 0.0.0.0/0 md5']})
+        self.b.bootstrap(config)
+        with open(os.path.join(self.p.data_dir, 'pg_hba.conf')) as f:
+            lines = f.readlines()
+            self.assertTrue('host replication replicator 127.0.0.1/32 md5\n' in lines)
+
+    @patch.object(CancellableSubprocess, 'call')
+    @patch.object(Postgresql, 'get_major_version', Mock(return_value=90600))
+    def test_custom_bootstrap(self, mock_cancellable_subprocess_call):
+        self.p.config._config.pop('pg_hba')
+        config = {'method': 'foo', 'foo': {'command': 'bar'}}
+
+        mock_cancellable_subprocess_call.return_value = 1
+        self.assertFalse(self.b.bootstrap(config))
+
+        mock_cancellable_subprocess_call.return_value = 0
+        with patch('multiprocessing.Process', Mock(side_effect=Exception("42"))),\
+                patch('os.path.isfile', Mock(return_value=True)),\
+                patch('os.unlink', Mock()),\
+                patch.object(ConfigHandler, 'save_configuration_files', Mock()),\
+                patch.object(ConfigHandler, 'restore_configuration_files', Mock()),\
+                patch.object(ConfigHandler, 'write_recovery_conf', Mock()):
+            with self.assertRaises(Exception) as e:
+                self.b.bootstrap(config)
+            self.assertEqual(str(e.exception), '42')
+
+            config['foo']['recovery_conf'] = {'foo': 'bar'}
+
+            with self.assertRaises(Exception) as e:
+                self.b.bootstrap(config)
+            self.assertEqual(str(e.exception), '42')
+
+        mock_cancellable_subprocess_call.side_effect = Exception
+        self.assertFalse(self.b.bootstrap(config))
+
+    @patch('time.sleep', Mock())
+    @patch('os.unlink', Mock())
+    @patch('shutil.copy', Mock())
+    @patch('os.path.isfile', Mock(return_value=True))
+    @patch.object(Bootstrap, 'call_post_bootstrap', Mock(return_value=True))
+    @patch.object(Bootstrap, '_custom_bootstrap', Mock(return_value=True))
+    @patch.object(Postgresql, 'start', Mock(return_value=True))
+    @patch.object(Postgresql, 'get_major_version', Mock(return_value=110000))
+    def test_post_bootstrap(self):
+        config = {'method': 'foo', 'foo': {'command': 'bar'}}
+        self.b.bootstrap(config)
+
+        task = CriticalTask()
+        with patch.object(Bootstrap, 'create_or_update_role', Mock(side_effect=Exception)):
+            self.b.post_bootstrap({}, task)
+            self.assertFalse(task.result)
+
+        self.p.config._config.pop('pg_hba')
+        self.b.post_bootstrap({}, task)
+        self.assertTrue(task.result)
+
+        self.b.bootstrap(config)
+        with patch.object(Postgresql, 'pending_restart', PropertyMock(return_value=True)), \
+                patch.object(Postgresql, 'restart', Mock()) as mock_restart:
+            self.b.post_bootstrap({}, task)
+            mock_restart.assert_called_once()
+
+        self.b.bootstrap(config)
+        self.p.set_state('stopped')
+        self.p.reload_config({'authentication': {'superuser': {'username': 'p', 'password': 'p'},
+                                                 'replication': {'username': 'r', 'password': 'r'},
+                                                 'rewind': {'username': 'rw', 'password': 'rw'}},
+                              'listen': '*', 'retry_timeout': 10, 'parameters': {'wal_level': '', 'hba_file': 'foo'}})
+        with patch.object(Postgresql, 'major_version', PropertyMock(return_value=110000)), \
+                patch.object(Postgresql, 'restart', Mock()) as mock_restart:
+            self.b.post_bootstrap({}, task)
+            mock_restart.assert_called_once()
+
+    @patch.object(CancellableSubprocess, 'call')
+    def test_call_post_bootstrap(self, mock_cancellable_subprocess_call):
+        mock_cancellable_subprocess_call.return_value = 1
+        self.assertFalse(self.b.call_post_bootstrap({'post_init': '/bin/false'}))
+
+        mock_cancellable_subprocess_call.return_value = 0
+        self.p.config.superuser.pop('username')
+        self.assertTrue(self.b.call_post_bootstrap({'post_init': '/bin/false'}))
+        mock_cancellable_subprocess_call.assert_called()
+        args, kwargs = mock_cancellable_subprocess_call.call_args
+        self.assertTrue('PGPASSFILE' in kwargs['env'])
+        self.assertEqual(args[0], ['/bin/false', 'postgres://127.0.0.2:5432/postgres'])
+
+        mock_cancellable_subprocess_call.reset_mock()
+        self.p.config._local_address.pop('host')
+        self.assertTrue(self.b.call_post_bootstrap({'post_init': '/bin/false'}))
+        mock_cancellable_subprocess_call.assert_called()
+        self.assertEqual(mock_cancellable_subprocess_call.call_args[0][0], ['/bin/false', 'postgres://:5432/postgres'])
+
+        mock_cancellable_subprocess_call.side_effect = OSError
+        self.assertFalse(self.b.call_post_bootstrap({'post_init': '/bin/false'}))
+
+    @patch('os.path.exists', Mock(return_value=True))
+    @patch('os.unlink', Mock())
+    @patch.object(Bootstrap, 'create_replica', Mock(return_value=0))
+    def test_clone(self):
+        self.b.clone(self.leader)

tests/test_callback_executor.py

@@ -1,7 +1,7 @@
 import unittest
 
 from mock import Mock, patch
-from patroni.callback_executor import CallbackExecutor
+from patroni.postgresql.callback_executor import CallbackExecutor
 
 
 class TestCallbackExecutor(unittest.TestCase):
@@ -17,6 +17,9 @@ class TestCallbackExecutor(unittest.TestCase):
 
         self.assertIsNone(ce.call([]))
 
+        mock_popen.return_value.kill.side_effect = OSError
+        self.assertIsNone(ce.call([]))
+
         mock_popen.side_effect = Exception
         ce = CallbackExecutor()
         ce._callback_event.wait = Mock(side_effect=[None, Exception])