Release 2.0.2 (#1851)

* bump version
* update release notes
* implement missing unit-test

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,48 @@
+---
+name: Bug report
+about: Create a report to help us improve
+title: ''
+labels: ''
+assignees: ''
+
+---
+
+**Describe the bug**
+A clear and concise description of what the bug is.
+
+**To Reproduce**
+Steps to reproduce the behavior:
+
+**Expected behavior**
+A clear and concise description of what you expected to happen.
+
+**Screenshots**
+If applicable, add screenshots to help explain your problem.
+
+**Environment**
+- Patroni version:
+- PostgreSQL version:
+- DCS (and its version):
+
+**Patroni configuration file**
+```
+Please copy&paste your Patroni configuration file here
+```
+
+**patronictl show-config**
+```
+Please copy&paste the output of "patronictl show-config" command here
+```
+
+**Have you checked Patroni logs?**
+Please provide a snippet of Patroni log files here
+
+**Have you checked PostgreSQL logs?**
+Please provide a snippet here
+
+**Have you tried to use GitHub issue search?**
+Maybe there is already a similar issue solved.
+
+
+**Additional context**
+Add any other context about the problem here.

.github/workflows/install_deps.py

@@ -0,0 +1,182 @@
+import inspect
+import os
+import shutil
+import subprocess
+import stat
+import sys
+import tarfile
+import time
+import zipfile
+
+
+def install_requirements(what):
+    old_path = sys.path[:]
+    w = os.path.join(os.getcwd(), os.path.dirname(inspect.getfile(inspect.currentframe())))
+    sys.path.insert(0, os.path.dirname(os.path.dirname(w)))
+    try:
+        from setup import EXTRAS_REQUIRE, read
+    finally:
+        sys.path = old_path
+    requirements = ['mock>=2.0.0', 'flake8', 'pytest', 'pytest-cov'] if what == 'all' else ['behave']
+    requirements += ['psycopg2-binary', 'coverage']
+    for r in read('requirements.txt').split('\n'):
+        r = r.strip()
+        if r != '':
+            extras = {e for e, v in EXTRAS_REQUIRE.items() if v and r.startswith(v[0])}
+            if not extras or what == 'all' or what in extras:
+                requirements.append(r)
+
+    subprocess.call([sys.executable, '-m', 'pip', 'install', '--upgrade', 'pip'])
+    r = subprocess.call([sys.executable, '-m', 'pip', 'install'] + requirements)
+    s = subprocess.call([sys.executable, '-m', 'pip', 'install', '--upgrade', 'setuptools'])
+    return s | r
+
+
+def install_packages(what):
+    packages = {
+        'zookeeper': ['zookeeper', 'zookeeper-bin', 'zookeeperd'],
+        'consul': ['consul'],
+    }
+    packages['exhibitor'] = packages['zookeeper']
+    packages = packages.get(what, [])
+    ver = str({'etcd': '9.6', 'etcd3': '9.6', 'consul': 10, 'exhibitor': 11, 'kubernetes': 12, 'raft': 13}.get(what))
+    subprocess.call(['sudo', 'apt-get', 'update', '-y'])
+    return subprocess.call(['sudo', 'apt-get', 'install', '-y', 'postgresql-' + ver, 'expect-dev', 'wget'] + packages)
+
+
+def get_file(url, name):
+    try:
+        from urllib.request import urlretrieve
+    except ImportError:
+        from urllib import urlretrieve
+
+    print('Downloading ' + url)
+    urlretrieve(url, name)
+
+
+def untar(archive, name):
+    with tarfile.open(archive) as tar:
+        f = tar.extractfile(name)
+        dest = os.path.basename(name)
+        with open(dest, 'wb') as d:
+            shutil.copyfileobj(f, d)
+            return dest
+
+
+def unzip(archive, name):
+    with zipfile.ZipFile(archive, 'r') as z:
+        name = z.extract(name)
+        dest = os.path.basename(name)
+        shutil.move(name, dest)
+        return dest
+
+
+def unzip_all(archive):
+    print('Extracting ' + archive)
+    with zipfile.ZipFile(archive, 'r') as z:
+        z.extractall()
+
+
+def chmod_755(name):
+    os.chmod(name, stat.S_IRUSR | stat.S_IWUSR | stat.S_IXUSR |
+             stat.S_IRGRP | stat.S_IXGRP | stat.S_IROTH | stat.S_IXOTH)
+
+
+def unpack(archive, name):
+    print('Extracting {0} from {1}'.format(name, archive))
+    func = unzip if archive.endswith('.zip') else untar
+    name = func(archive, name)
+    chmod_755(name)
+    return name
+
+
+def install_etcd():
+    version = os.environ.get('ETCDVERSION', '3.3.13')
+    platform = {'linux2': 'linux', 'win32': 'windows', 'cygwin': 'windows'}.get(sys.platform, sys.platform)
+    dirname = 'etcd-v{0}-{1}-amd64'.format(version, platform)
+    ext = 'tar.gz' if platform == 'linux' else 'zip'
+    name = '{0}.{1}'.format(dirname, ext)
+    url = 'https://github.com/etcd-io/etcd/releases/download/v{0}/{1}'.format(version, name)
+    get_file(url, name)
+    ext = '.exe' if platform == 'windows' else ''
+    return int(unpack(name, '{0}/etcd{1}'.format(dirname, ext)) is None)
+
+
+def install_postgres():
+    version = os.environ.get('PGVERSION', '12.1-1')
+    platform = {'darwin': 'osx', 'win32': 'windows-x64', 'cygwin': 'windows-x64'}[sys.platform]
+    name = 'postgresql-{0}-{1}-binaries.zip'.format(version, platform)
+    get_file('http://get.enterprisedb.com/postgresql/' + name, name)
+    unzip_all(name)
+    bin_dir = os.path.join('pgsql', 'bin')
+    for f in os.listdir(bin_dir):
+        chmod_755(os.path.join(bin_dir, f))
+    subprocess.call(['pgsql/bin/postgres', '-V'])
+    return 0
+
+
+def setup_kubernetes():
+    get_file('https://storage.googleapis.com/minikube/k8sReleases/v1.7.0/localkube-linux-amd64', 'localkube')
+    chmod_755('localkube')
+
+    devnull = open(os.devnull, 'w')
+    subprocess.Popen(['sudo', 'nohup', './localkube', '--logtostderr=true', '--enable-dns=false'],
+                     stdout=devnull, stderr=devnull)
+    for _ in range(0, 120):
+        if subprocess.call(['wget', '-qO', '-', 'http://127.0.0.1:8080/'], stdout=devnull, stderr=devnull) == 0:
+            break
+        time.sleep(1)
+    else:
+        print('localkube did not start')
+        return 1
+
+    subprocess.call('sudo chmod 644 /var/lib/localkube/certs/*', shell=True)
+    print('Set up .kube/config')
+    kube = os.path.join(os.path.expanduser('~'), '.kube')
+    os.makedirs(kube)
+    with open(os.path.join(kube, 'config'), 'w') as f:
+        f.write("""apiVersion: v1
+clusters:
+- cluster:
+    certificate-authority: /var/lib/localkube/certs/ca.crt
+    server: https://127.0.0.1:8443
+  name: local
+contexts:
+- context:
+    cluster: local
+    user: myself
+  name: local
+current-context: local
+kind: Config
+preferences: {}
+users:
+- name: myself
+  user:
+    client-certificate: /var/lib/localkube/certs/apiserver.crt
+    client-key: /var/lib/localkube/certs/apiserver.key
+""")
+    return 0
+
+
+def main():
+    what = os.environ.get('DCS', sys.argv[1] if len(sys.argv) > 1 else 'all')
+
+    if what != 'all':
+        if sys.platform.startswith('linux'):
+            r = install_packages(what)
+            if r == 0 and what == 'kubernetes':
+                r = setup_kubernetes()
+        else:
+            r = install_postgres()
+
+        if r == 0 and what.startswith('etcd'):
+            r = install_etcd()
+
+        if r != 0:
+            return r
+
+    return install_requirements(what)
+
+
+if __name__ == '__main__':
+    sys.exit(main())

.github/workflows/run_tests.py

@@ -0,0 +1,47 @@
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+
+
+def main():
+    what = os.environ.get('DCS', sys.argv[1] if len(sys.argv) > 1 else 'all')
+
+    if what == 'all':
+        flake8 = subprocess.call([sys.executable, 'setup.py', 'flake8'])
+        test = subprocess.call([sys.executable, 'setup.py', 'test'])
+        version = '.'.join(map(str, sys.version_info[:2]))
+        shutil.move('.coverage', os.path.join(tempfile.gettempdir(), '.coverage.' + version))
+        return flake8 | test
+    elif what == 'combine':
+        tmp = tempfile.gettempdir()
+        for name in os.listdir(tmp):
+            if name.startswith('.coverage.'):
+                shutil.move(os.path.join(tmp, name), name)
+        return subprocess.call([sys.executable, '-m', 'coverage', 'combine'])
+
+    env = os.environ.copy()
+    if sys.platform.startswith('linux'):
+        version = {'etcd': '9.6', 'etcd3': '9.6', 'consul': 10, 'exhibitor': 11, 'kubernetes': 12, 'raft': 13}.get(what)
+        path = '/usr/lib/postgresql/{0}/bin:.'.format(version)
+        unbuffer = ['timeout', '600', 'unbuffer']
+    else:
+        path = os.path.abspath(os.path.join('pgsql', 'bin'))
+        if sys.platform == 'darwin':
+            path += ':.'
+        unbuffer = []
+    env['PATH'] = path + os.pathsep + env['PATH']
+    env['DCS'] = what
+
+    ret = subprocess.call(unbuffer + [sys.executable, '-m', 'behave'], env=env)
+
+    if ret != 0:
+        if subprocess.call('grep . features/output/*_failed/*postgres?.*', shell=True) != 0:
+            subprocess.call('grep . features/output/*/*postgres?.*', shell=True)
+        return 1
+    return 0
+
+
+if __name__ == '__main__':
+    sys.exit(main())

.github/workflows/tests.yaml

@@ -0,0 +1,175 @@
+name: Tests
+
+on:
+  pull_request:
+  push:
+    branches:
+    - master
+    tags:
+    - v.*
+
+jobs:
+  unit:
+    runs-on: ${{ matrix.os }}-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu, windows, macos]
+
+    steps:
+    - uses: actions/checkout@v1
+    - name: Set up Python 2.7
+      uses: actions/setup-python@v2
+      with:
+        python-version: 2.7
+      if: matrix.os != 'windows'
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+      if: matrix.os != 'windows'
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+      if: matrix.os != 'windows'
+
+    - name: Set up Python 3.5
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.5
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+
+    - name: Set up Python 3.6
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.6
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+
+    - name: Set up Python 3.7
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.7
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+
+    - name: Set up Python 3.8
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.8
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+
+    - name: Set up Python 3.9
+      uses: actions/setup-python@v2
+      with:
+        python-version: 3.9
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run tests and flake8
+      run: python .github/workflows/run_tests.py
+
+    - name: Combine coverage
+      run: python .github/workflows/run_tests.py combine
+
+    - name: Install coveralls
+      run: python -m pip install coveralls
+
+    - name: Upload Coverage
+      env:
+        COVERALLS_FLAG_NAME: unit-${{ matrix.os }}
+        COVERALLS_PARALLEL: 'true'
+        GITHUB_TOKEN: ${{ secrets.github_token }}
+      run: python -m coveralls --service=github
+
+    - name: Run codacy-coverage-reporter
+      uses: codacy/codacy-coverage-reporter-action@master
+      env:
+        SECRETS_AVAILABLE: ${{ secrets.CODACY_PROJECT_TOKEN != '' }}
+      with:
+        project-token: ${{ secrets.CODACY_PROJECT_TOKEN }}
+        coverage-reports: coverage.xml
+      if: ${{ matrix.os == 'ubuntu' && env.SECRETS_AVAILABLE == 'true' }}
+
+  behave:
+    runs-on: ${{ matrix.os }}-latest
+    env:
+      DCS: ${{ matrix.dcs }}
+      ETCDVERSION: 3.3.13
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu]
+        python-version: [2.7, 3.5, 3.8]
+        dcs: [etcd, etcd3, consul, exhibitor, kubernetes, raft]
+
+    steps:
+    - uses: actions/checkout@v1
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run behave tests
+      run: python .github/workflows/run_tests.py
+    - uses: actions/setup-python@v2
+      with:
+        python-version: 3.9
+    - name: Install coveralls
+      run: python -m pip install coveralls
+    - name: Upload Coverage
+      env:
+        COVERALLS_FLAG_NAME: behave-${{ matrix.os }}-${{ matrix.dcs }}-${{ matrix.python-version }}
+        COVERALLS_PARALLEL: 'true'
+        GITHUB_TOKEN: ${{ secrets.github_token }}
+      run: python -m coveralls --service=github
+
+  behavem:
+    runs-on: ${{ matrix.os }}-latest
+    env:
+      DCS: ${{ matrix.dcs }}
+      ETCDVERSION: 3.3.13
+      PGVERSION: 12.1-1  # for windows and macos
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [macos]  #, windows]
+        python-version: [3.7]
+        dcs: [etcd, etcd3, raft]
+
+    steps:
+    - uses: actions/checkout@v1
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+    - name: Install dependencies
+      run: python .github/workflows/install_deps.py
+    - name: Run behave tests
+      run: python .github/workflows/run_tests.py
+    - name: Install coveralls
+      run: python -m pip install coveralls
+    - name: Upload Coverage
+      env:
+        COVERALLS_FLAG_NAME: behave-${{ matrix.os }}-${{ matrix.dcs }}-${{ matrix.python-version }}
+        COVERALLS_PARALLEL: 'true'
+        GITHUB_TOKEN: ${{ secrets.github_token }}
+      run: python -m coveralls --service=github
+
+  coveralls-finish:
+    name: Finalize coveralls.io
+    needs: [unit, behave, behavem]
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/setup-python@v2
+    - run: python -m pip install coveralls
+    - run: python -m coveralls --service=github --finish
+      env:
+        GITHUB_TOKEN: ${{ secrets.github_token }}

.gitignore

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

.travis.yml

@@ -1,153 +0,0 @@
-sudo: true
-dist: trusty
-language: python
-env:
-  global:
-  - ETCDVERSION=3.0.17 ZKVERSION=3.4.11 CONSULVERSION=0.7.4
-  - PYVERSIONS="2.7 3.5 3.6"
-  - EXCLUDE_BEHAVE="3.5"
-  - BOTO_CONFIG=/doesnotexist
-matrix:
-  include:
-  - 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
-before_cache:
-- |
-  rm -fr $HOME/mycache/python*
-  for pv in $PYVERSIONS; do
-      if [[ $TEST_SUITE != "behave" || $pv != $EXCLUDE_BEHAVE ]]; then
-          fpv=$(basename $(readlink $HOME/virtualenv/python${pv}))
-          mv $HOME/virtualenv/${fpv} $HOME/mycache/${fpv}
-      fi
-  done
-install:
-  - |
-    set -e
-
-    if [[ $TEST_SUITE == "behave" ]]; then
-        function get_consul() {
-            CC=~/mycache/consul_${CONSULVERSION}
-            if [[ ! -x $CC ]]; then
-                curl -L https://releases.hashicorp.com/consul/${CONSULVERSION}/consul_${CONSULVERSION}_linux_amd64.zip \
-                    | gunzip > $CC
-                [[ ${PIPESTATUS[0]} == 0 ]] || return 1
-                chmod +x $CC
-            fi
-            ln -s $CC consul
-        }
-
-        function get_etcd() {
-            EC=~/mycache/etcd_${ETCDVERSION}
-            if [[ ! -x $EC ]]; then
-                curl -L https://github.com/coreos/etcd/releases/download/v${ETCDVERSION}/etcd-v${ETCDVERSION}-linux-amd64.tar.gz \
-                    | tar xz -C . --strip=1 --wildcards --no-anchored etcd
-                [[ ${PIPESTATUS[0]} == 0 ]] || return 1
-                mv etcd $EC
-            fi
-            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 $ZC ]]; then
-                curl -L http://www.apache.org/dist/zookeeper/zookeeper-${ZKVERSION}/zookeeper-${ZKVERSION}.tar.gz | tar xz
-                [[ ${PIPESTATUS[0]} == 0 ]] || return 1
-                mv zookeeper-${ZKVERSION}/conf/zoo_sample.cfg zookeeper-${ZKVERSION}/conf/zoo.cfg
-                mv zookeeper-${ZKVERSION} $ZC
-            fi
-            $ZC/bin/zkServer.sh start
-            # following lines are 'emulating' exhibitor REST API
-            while true; do
-                echo -e 'HTTP/1.0 200 OK\nContent-Type: application/json\n\n{"servers":["127.0.0.1"],"port":2181}' \
-                    | nc -l 8181 &> /dev/null
-            done&
-        }
-
-        attempt_num=1
-        until get_${DCS}; do
-            [[ $attempt_num -ge 3 ]] && exit 1
-            echo "Attempt $attempt_num failed! Trying again in $attempt_num seconds..."
-            sleep $(( attempt_num++ ))
-        done
-    fi
-
-    for pv in $PYVERSIONS; do
-        if [[ $TEST_SUITE != "behave" || $pv != $EXCLUDE_BEHAVE ]]; then
-            fpv=$(basename $(readlink $HOME/virtualenv/python$pv))
-            if [[ -d ~/mycache/${fpv} ]]; then
-                mv ~/virtualenv/${fpv} ~/virtualenv/${fpv}.bckp
-                mv ~/mycache/${fpv} ~/virtualenv/${fpv}
-            fi
-            source ~/virtualenv/python${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
-                pip install $p --upgrade
-            done
-        fi
-    done
-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
-            $TEST_SUITE flake8
-        elif [[ $pv != $EXCLUDE_BEHAVE ]]; then
-            echo Running acceptance tests using python${pv}
-            if ! PATH=.:/usr/lib/postgresql/9.6/bin:$PATH $TEST_SUITE; then
-                # output all log files when tests are failing
-                grep . features/output/*/*postgres?.*
-                exit 1
-            fi
-        fi
-    done
-
-    set +e
-after_success:
-  # before_cache is executed earlier than after_success, so we need to restore one of virtualenv directories
-  - fpv=$(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

@@ -1,45 +1,158 @@
 ## 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 postgres:10
-MAINTAINER Alexander Kukushkin <[email protected]>
+ARG PG_MAJOR=10
+ARG COMPRESS=false
+ARG PGHOME=/home/postgres
+ARG PGDATA=$PGHOME/data
+ARG LC_ALL=C.UTF-8
+ARG LANG=C.UTF-8
 
-RUN export DEBIAN_FRONTEND=noninteractive \
+FROM postgres:$PG_MAJOR as builder
+
+ARG PGHOME
+ARG PGDATA
+ARG LC_ALL
+ARG LANG
+
+ENV ETCDVERSION=3.3.13 CONFDVERSION=0.16.0
+
+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 upgrade -y \
-    # postgres:10 is based on debian, which has patroni package. We will install all required dependencies
-    && apt-get install -s patroni | sed -n -e '/^Inst patroni /d' -e 's/^Inst \([^ ]\+\) .*$/\1/p' \
-            | xargs apt-get install -y curl jq haproxy locales python3-etcd python3-kazoo \
-
-    ## Make sure we have a en_US.UTF-8 locale available
+    # 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' \
+            | grep -Ev '^python3-(sphinx|etcd|consul|kazoo|kubernetes)' \
+            | xargs apt-get install -y vim curl less jq locales haproxy sudo \
+                            python3-etcd python3-kazoo python3-pip busybox \
+                            net-tools iputils-ping --fix-missing \
+    && pip3 install dumb-init \
+\
+    # 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 \
-
-    && mkdir -p /home/postgres \
-    && chown postgres:postgres /home/postgres \
-
-    # Clean up
-    && apt-get purge -y libpython2.7-stdlib libpython2.7-minimal \
+\
+    # 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 3.2.23
-RUN curl -L 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
+# perform compression if it is necessary
+ARG COMPRESS
+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 0.16.0
-RUN curl -L https://github.com/kelseyhightower/confd/releases/download/v${CONFDVERSION}/confd-${CONFDVERSION}-linux-amd64 > /usr/local/bin/confd \
-    && chmod +x /usr/local/bin/confd
+FROM scratch
+COPY --from=builder / /
 
-ADD patronictl.py patroni.py docker/entrypoint.sh /
-ADD patroni /patroni/
-ADD extras/confd /etc/confd
+LABEL maintainer="Alexander Kukushkin <[email protected]>"
 
-RUN sed -i 's/env python/&3/' patroni*.py && ln -s /patronictl.py /usr/local/bin/patronictl && mkdir /data/ /run/haproxy \
-    && touch /pgpass /patroni.yml && chown postgres:postgres -R /patroni/ /data/ /pgpass /patroni.yml /etc/haproxy /var/run/ /var/lib/ /var/log/
+ARG PG_MAJOR
+ARG COMPRESS
+ARG PGHOME
+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 \
+    && chmod +s /bin/ping \
+    && chown -R postgres:postgres $PGHOME /run /etc/haproxy
 
-ENV LC_ALL=en_US.UTF-8 LANG=en_US.UTF-8
-ENTRYPOINT ["/bin/bash", "/entrypoint.sh"]
 USER postgres
+
+ENTRYPOINT ["/bin/sh", "/entrypoint.sh"]

README.rst

@@ -12,6 +12,8 @@ Patroni is a template for you to create your own customized, high-availability s
 
 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.
 
+Currently supported PostgreSQL versions: 9.3 to 13.
+
 **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::
@@ -59,6 +61,33 @@ To install requirements on a Mac, run the following:
 
     brew install postgresql etcd haproxy libyaml python
 
+**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:
@@ -69,7 +98,7 @@ Patroni can be installed with pip:
 
 where dependencies can be either empty, or consist of one or more of the following:
 
-etcd
+etcd or etcd3
     `python-etcd` module in order to use Etcd as DCS
 consul
     `python-consul` module in order to use Consul as DCS
@@ -79,6 +108,8 @@ 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
+raft
+    `pysyncobj` module in order to use python Raft implementation as DCS
 aws
     `boto` in order to use AWS callbacks
 
@@ -97,7 +128,7 @@ Running and Configuring
 To get started, do the following from different terminals:
 ::
 
-    > etcd --data-dir=data/etcd
+    > etcd --data-dir=data/etcd --enable-v2=true
     > ./patroni.py postgres0.yml
     > ./patroni.py postgres1.yml
 

docker-compose.yml

@@ -1,58 +1,77 @@
 # 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:
-    container_name: patroni_etcd
-    image: patroni
-    command: --etcd
+networks:
+    demo:
 
-dbnode1:
-    image: patroni
-    hostname: dbnode1
-    links:
-        - patroni_etcd:patroni_etcd
-    volumes:
-        - ./patroni:/patroni
-    env_file: docker/patroni-secrets.env
-    environment:
-        PATRONI_ETCD_URL: http://patroni_etcd:2379
-        PATRONI_NAME: dbnode1
-        PATRONI_SCOPE: testcluster
+services:
+    etcd1: &etcd
+        image: patroni
+        networks: [ demo ]
+        environment:
+            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
+        container_name: demo-etcd1
+        hostname: etcd1
+        command: etcd -name etcd1 -initial-advertise-peer-urls http://etcd1:2380
 
-dbnode2:
-    image: patroni
-    hostname: dbnode2
-    links:
-        - patroni_etcd:patroni_etcd
-    volumes:
-        - ./patroni:/patroni
-    env_file: docker/patroni-secrets.env
-    environment:
-        PATRONI_ETCD_URL: http://patroni_etcd:2379
-        PATRONI_NAME: dbnode2
-        PATRONI_SCOPE: testcluster
+    etcd2:
+        <<: *etcd
+        container_name: demo-etcd2
+        hostname: etcd2
+        command: etcd -name etcd2 -initial-advertise-peer-urls http://etcd2:2380
 
-dbnode3:
-    image: patroni
-    hostname: dbnode3
-    links:
-        - patroni_etcd:patroni_etcd
-    volumes:
-        - ./patroni:/patroni
-    env_file: docker/patroni-secrets.env
-    environment:
-        PATRONI_ETCD_URL: http://patroni_etcd:2379
-        PATRONI_NAME: dbnode3
-        PATRONI_SCOPE: testcluster
+    etcd3:
+        <<: *etcd
+        container_name: demo-etcd3
+        hostname: etcd3
+        command: etcd -name etcd3 -initial-advertise-peer-urls http://etcd3:2380
 
-haproxy:
-    image: patroni
-    links:
-        - patroni_etcd:patroni_etcd
-    ports:
-        - "5000:5000"
-        - "5001:5001"
-    environment:
-        PATRONI_ETCD_URL: http://patroni_etcd:2379
-        PATRONI_SCOPE: testcluster
-    command: --confd
+    haproxy:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: haproxy
+        container_name: demo-haproxy
+        ports:
+            - "5000:5000"
+            - "5001:5001"
+        command: haproxy
+        environment: &haproxy_env
+            ETCDCTL_ENDPOINTS: http://etcd1:2379,http://etcd2:2379,http://etcd3:2379
+            PATRONI_ETCD3_HOSTS: "'etcd1:2379','etcd2:2379','etcd3:2379'"
+            PATRONI_SCOPE: demo
+
+    patroni1:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: patroni1
+        container_name: demo-patroni1
+        environment:
+            <<: *haproxy_env
+            PATRONI_NAME: patroni1
+
+    patroni2:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: patroni2
+        container_name: demo-patroni2
+        environment:
+            <<: *haproxy_env
+            PATRONI_NAME: patroni2
+
+    patroni3:
+        image: patroni
+        networks: [ demo ]
+        env_file: docker/patroni.env
+        hostname: patroni3
+        container_name: demo-patroni3
+        environment:
+            <<: *haproxy_env
+            PATRONI_NAME: patroni3

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
-
-    https://registry.opensource.zalan.do/v1/repositories/acid/patroni/tags
+You can run Patroni in a docker container using this Dockerfile
 
 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
-
-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
+## Three-node Patroni cluster with three-node etcd cluster and one haproxy container using docker-compose
 
 Example session:
 
-    $ ./dev_patroni_cluster.sh --image registry.opensource.zalan.do/acid/patroni:1.0-SNAPSHOT --members=2 --name=bravo
-    The etcd container is 6be871a11cb373406ca5ea1c6b39e1.0-SNAPSHOTfdde9fb1d6177212d6ad0c0d1bd9b563, ip=172.17.1.24
-    Started Patroni container 67e611f2eca7c40f9e6e0e24a4a8f2cba7e3e56d22a420e15ab9240a37a9d7a4, ip=172.17.1.25
-    Started Patroni container 47dd12ae635ab83b039f5889e250048b606ed5e48e3650b69e365e7e1d4acbcf, ip=172.17.1.26
+    $ docker-compose up -d
+    Creating demo-haproxy ...
+    Creating demo-patroni2 ...
+    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
-    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
-    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
-    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
+    CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS                              NAMES
+    5b7a90b4cfbf        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 27 seconds                                          demo-etcd2
+    e30eea5222f2        patroni             "/bin/sh /entrypoint…"   29 seconds ago      Up 27 seconds                                          demo-etcd1
+    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_URL=http://${ETCD_CONTAINER}:2379"
-PATRONI_ENV=$(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()
-{
-    cat <<__EOF__
-Usage: $0
+if [ -f /a.tar.xz ]; then
+    echo "decompressing image..."
+    sudo tar xpJf /a.tar.xz -C / > /dev/null 2>&1
+    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
-    --confd     Do not run Patroni, run a standalone confd
-    --zookeeper Do not run Patroni, run a standalone zookeeper
-
-Examples:
-
-    $0 --etcd
-    $0 --confd
-    $0 --zookeeper
-    $0
-__EOF__
-}
-
-DOCKER_IP=$(hostname --ip-address)
-PATRONI_SCOPE=${PATRONI_SCOPE:-batman}
-ETCD_ARGS="--data-dir /tmp/etcd.data -advertise-client-urls=http://${DOCKER_IP}:2379 -listen-client-urls=http://0.0.0.0:2379"
-
-optspec=":vh-:"
-while getopts "$optspec" optchar; do
-    case "${optchar}" in
-        -)
-            case "${OPTARG}" in
-                confd)
-                    haproxy -f /etc/haproxy/haproxy.cfg -p /var/run/haproxy.pid -D
-                    CONFD="confd -prefix=${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_URL}/v2/members | jq -r '.members[0].clientURLs[0]' | grep -q http; do
-                            sleep 1
-                        done
-                        exec $CONFD etcd -node $PATRONI_ETCD_URL
-                    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
+case "$1" in
+    haproxy)
+        haproxy -f /etc/haproxy/haproxy.cfg -p /var/run/haproxy.pid -D
+        CONFD="confd -prefix=$PATRONI_NAMESPACE/$PATRONI_SCOPE -interval=10 -backend"
+        if [ ! -z "$PATRONI_ZOOKEEPER_HOSTS" ]; then
+            while ! /usr/share/zookeeper/bin/zkCli.sh -server $PATRONI_ZOOKEEPER_HOSTS ls /; do
+                sleep 1
+            done
+            exec dumb-init $CONFD zookeeper -node $PATRONI_ZOOKEEPER_HOSTS
+        else
+            while ! etcdctl cluster-health 2> /dev/null; do
+                sleep 1
+            done
+            exec dumb-init $CONFD etcdv3 -node $(echo $ETCDCTL_ENDPOINTS | sed 's/,/ -node /g')
+        fi
+        ;;
+    etcd)
+        exec "$@" -advertise-client-urls http://$DOCKER_IP:2379
+        ;;
+    zookeeper)
+        exec /usr/share/zookeeper/bin/zkServer.sh start-foreground
+        ;;
+esac
 
 ## We start an etcd
-if [[ -z ${PATRONI_ETCD_URL} && -z ${PATRONI_ZOOKEEPER_HOSTS} ]]; then
-    etcd $ETCD_ARGS > /var/log/etcd.log 2> /var/log/etcd.err &
+if [ -z "$PATRONI_ETCD3_HOSTS" ] && [ -z "$PATRONI_ZOOKEEPER_HOSTS" ]; then
     export PATRONI_ETCD_URL="http://127.0.0.1:2379"
+    etcd --data-dir /tmp/etcd.data -advertise-client-urls=$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_RESTAPI_CONNECT_ADDRESS="${DOCKER_IP}:8008"
+export PATRONI_NAMESPACE
+export PATRONI_NAME="${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__
-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 python3 /patroni.py /patroni.yml
-
-while true; do
-    sleep 60
-done
+exec python3 /patroni.py postgres0.yml

docs/CONTRIBUTING.rst

@@ -10,6 +10,39 @@ 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/>`__.
 
+Running tests
+-------------
+
+Requirements for running behave tests:
+
+1. PostgreSQL packages need to be installed.
+2. PostgreSQL binaries must be available in your `PATH`. You may need to add them to the path with something like `PATH=/usr/lib/postgresql/11/bin:$PATH python -m behave`.
+3. If you'd like to test with external DCSs (e.g., Etcd, Consul, and Zookeeper) you'll need the packages installed and respective services running and accepting unencrypted/unprotected connections on localhost and default port. In the case of Etcd or Consul, the behave test suite could start them up if binaries are available in the `PATH`.
+
+ Install dependencies:
+
+.. code-block:: bash
+
+    # You may want to use Virtualenv or specify pip3.
+    pip install -r requirements.txt
+    pip install -r requirements.dev.txt
+
+After you have all dependencies installed, you can run the various test suites:
+
+.. code-block:: bash
+
+    # You may want to use Virtualenv or specify python3.
+
+    # Run flake8 to check syntax and formatting:
+    python setup.py flake8
+
+    # Run the pytest suite in tests/:
+    python setup.py test
+
+    # Run the behave (https://behave.readthedocs.io/en/latest/) test suite in features/;
+    # modify DCS as desired (raft has no dependencies so is the easiest to start with):
+    DCS=raft python -m behave
+
 Reporting issues
 ----------------
 

docs/ENVIRONMENT.rst

@@ -11,8 +11,18 @@ 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\_LOGLEVEL**: sets the general logging level (see `the docs for Python logging <https://docs.python.org/3.6/library/logging.html#levels>`_)
--  **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>`_)
+
+Log
+---
+-  **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\_TRACEBACK\_LEVEL**: sets the level where tracebacks will be visible. Default value is **ERROR**. Set it to **DEBUG** if you want to see tracebacks only if you enable **PATRONI\_LOG\_LEVEL=DEBUG**.
+-  **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}"``
 
 Bootstrap configuration
 -----------------------
@@ -25,8 +35,8 @@ 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\_HOST**: the host:port for the Consul local agent.
+-  **PATRONI\_CONSUL\_URL**: url for the Consul local agent, 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
@@ -35,19 +45,49 @@ Consul
 -  **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\_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\_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. By default an empty list is used.
+-  **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\_HOSTS**: list of etcd endpoints in format host1:port1,host2:port2,etc...
--  **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\_USERNAME**: username for etcd authentication.
+-  **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\_CERT**: File with the client certificate.
 -  **PATRONI\_ETCD\_KEY**: File with the client key. Can be empty if the key is part of certificate.
 
+Etcdv3
+------
+Environment names for Etcdv3 are similar as for Etcd, you just need to use ``ETCD3`` instead of ``ETCD`` in the variable name. Example: ``PATRONI_ETCD3_HOST``, ``PATRONI_ETCD3_CACERT``, and so on.
+
+.. warning::
+    Keys created with protocol version 2 are not visible with protocol version 3 and the other way around, therefore it is not possible to switch from Etcd to Etcdv3 just by updating Patroni configuration.
+
+
+ZooKeeper
+---------
+-  **PATRONI\_ZOOKEEPER\_HOSTS**: Comma separated list of ZooKeeper cluster members: "'host1:port1','host2:port2','etc...'". It is important to quote every single entity!
+-  **PATRONI\_ZOOKEEPER\_USE\_SSL**: (optional) Whether SSL is used or not. Defaults to ``false``. If set to ``false``, all SSL specific parameters are ignored.
+-  **PATRONI\_ZOOKEEPER\_CACERT**: (optional) The CA certificate. If present it will enable validation.
+-  **PATRONI\_ZOOKEEPER\_CERT**: (optional) File with the client certificate.
+-  **PATRONI\_ZOOKEEPER\_KEY**: (optional) File with the client key.
+-  **PATRONI\_ZOOKEEPER\_KEY\_PASSWORD**: (optional) The client key password.
+-  **PATRONI\_ZOOKEEPER\_VERIFY**: (optional) Whether to verify certificate or not. Defaults to ``true``.
+
+.. note::
+    It is required to install ``kazoo>=2.6.0`` to support SSL.
+
+
 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.
@@ -57,13 +97,23 @@ Exhibitor
 
 Kubernetes
 ----------
+-  **PATRONI\_KUBERNETES\_BYPASS\_API\_SERVICE**: (optional) When communicating with the Kubernetes API, Patroni is usually relying on the `kubernetes` service, the address of which is exposed in the pods via the `KUBERNETES_SERVICE_HOST` environment variable. If `PATRONI_KUBERNETES_BYPASS_API_SERVICE` is set to ``true``, Patroni will resolve the list of API nodes behind the service and connect directly to them.
 -  **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.
+-  **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.
+
+Raft
+----
+
+-  **PATRONI\_RAFT\_SELF\_ADDR**: ``ip:port`` to listen on for Raft connections. The ``self_addr`` must be accessible from other nodes of the cluster. If not set, the node will not participate in consensus.
+-  **PATRONI\_RAFT\_BIND\_ADDR**: (optional) ``ip:port`` to listen on for Raft connections. If not specified the ``self_addr`` will be used.
+-  **PATRONI\_RAFT\_PARTNER\_ADDRS**: list of other Patroni nodes in the cluster in format ``"'ip1:port1','ip2:port2'"``. It is important to quote every single entity!
+-  **PATRONI\_RAFT\_DATA\_DIR**: directory where to store Raft log and snapshot. If not specified the current working directory is used.
+-  **PATRONI\_RAFT\_PASSWORD**: (optional) Encrypt Raft traffic with a specified password, requires ``cryptography`` python module.
 
 PostgreSQL
 ----------
@@ -75,18 +125,54 @@ PostgreSQL
 -  **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\_REPLICATION\_SSLMODE**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+-  **PATRONI\_REPLICATION\_SSLKEY**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+-  **PATRONI\_REPLICATION\_SSLPASSWORD**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``PATRONI_REPLICATION_SSLKEY``.
+-  **PATRONI\_REPLICATION\_SSLCERT**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+-  **PATRONI\_REPLICATION\_SSLROOTCERT**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+-  **PATRONI\_REPLICATION\_SSLCRL**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+-  **PATRONI\_REPLICATION\_GSSENCMODE**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+-  **PATRONI\_REPLICATION\_CHANNEL\_BINDING**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
 -  **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\_SUPERUSER\_SSLMODE**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+-  **PATRONI\_SUPERUSER\_SSLKEY**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+-  **PATRONI\_SUPERUSER\_SSLPASSWORD**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``PATRONI_SUPERUSER_SSLKEY``.
+-  **PATRONI\_SUPERUSER\_SSLCERT**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+-  **PATRONI\_SUPERUSER\_SSLROOTCERT**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+-  **PATRONI\_SUPERUSER\_SSLCRL**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+-  **PATRONI\_SUPERUSER\_GSSENCMODE**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+-  **PATRONI\_SUPERUSER\_CHANNEL\_BINDING**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
+-  **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.
+-  **PATRONI\_REWIND\_SSLMODE**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+-  **PATRONI\_REWIND\_SSLKEY**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+-  **PATRONI\_REWIND\_SSLPASSWORD**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``PATRONI_REWIND_SSLKEY``.
+-  **PATRONI\_REWIND\_SSLCERT**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+-  **PATRONI\_REWIND\_SSLROOTCERT**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+-  **PATRONI\_REWIND\_SSLCRL**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+-  **PATRONI\_REWIND\_GSSENCMODE**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+-  **PATRONI\_REWIND\_CHANNEL\_BINDING**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
 
 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.
 -  **PATRONI\_RESTAPI\_PASSWORD**: Basic-auth password to protect unsafe REST API endpoints.
 -  **PATRONI\_RESTAPI\_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.
 -  **PATRONI\_RESTAPI\_KEYFILE**: Specifies the file with the secret key in the PEM format.
+-  **PATRONI\_RESTAPI\_KEYFILE\_PASSWORD**: Specifies a password for decrypting the keyfile.
+-  **PATRONI\_RESTAPI\_CAFILE**: Specifies the file with the CA_BUNDLE with certificates of trusted CAs to use while verifying client certs.
+-  **PATRONI\_RESTAPI\_CIPHERS**: (optional) Specifies the permitted cipher suites (e.g. "ECDHE-RSA-AES256-GCM-SHA384:DHE-RSA-AES256-GCM-SHA384:ECDHE-RSA-AES128-GCM-SHA256:DHE-RSA-AES128-GCM-SHA256:!SSLv1:!SSLv2:!SSLv3:!TLSv1:!TLSv1.1")
+-  **PATRONI\_RESTAPI\_VERIFY\_CLIENT**: ``none`` (default), ``optional`` or ``required``. When ``none`` REST API will not check client certificates. When ``required`` client certificates are required for all REST API calls. When ``optional`` client certificates are required for all unsafe REST API endpoints. When ``required`` is used, then client authentication succeeds, if the certificate signature verification succeeds. For ``optional`` the client cert will only be checked for ``PUT``, ``POST``, ``PATCH``, and ``DELETE`` requests.
+-  **PATRONI\_RESTAPI\_HTTP\_EXTRA\_HEADERS**: (optional) HTTP headers let the REST API server pass additional information with an HTTP response.
+-  **PATRONI\_RESTAPI\_HTTPS\_EXTRA\_HEADERS**: (optional) HTTPS headers let the REST API server pass additional information with an HTTP response when TLS is enabled. This will also pass additional information set in ``http_extra_headers``.
 
-ZooKeeper
----------
--  **PATRONI\_ZOOKEEPER\_HOSTS**: comma separated list of ZooKeeper cluster members: "'host1:port1','host2:port2','etc...'". It is important to quote every single entity!
+CTL
+---
+-  **PATRONICTL\_CONFIG\_FILE**: location of the configuration file.
+-  **PATRONI\_CTL\_INSECURE**: Allow connections to REST API without verifying SSL certs.
+-  **PATRONI\_CTL\_CACERT**: Specifies the file with the CA_BUNDLE file or directory with certificates of trusted CAs to use while verifying REST API SSL certs. If not provided patronictl will use the value provided for REST API "cafile" parameter.
+-  **PATRONI\_CTL\_CERTFILE**: Specifies the file with the client certificate in the PEM format. If not provided patronictl will use the value provided for REST API "certfile" parameter.
+-  **PATRONI\_CTL\_KEYFILE**: Specifies the file with the client secret key in the PEM format. If not provided patronictl will use the value provided for REST API "keyfile" parameter.

docs/README.rst

@@ -33,6 +33,35 @@ To install requirements on a Mac, run the following:
 
     brew install postgresql etcd haproxy libyaml python
 
+.. _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:
@@ -43,7 +72,7 @@ Patroni can be installed with pip:
 
 where dependencies can be either empty, or consist of one or more of the following:
 
-etcd
+etcd or etcd3
     `python-etcd` module in order to use Etcd as DCS
 consul
     `python-consul` module in order to use Consul as DCS
@@ -53,6 +82,8 @@ 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
+raft
+    `pysyncobj` module in order to use python Raft implementation as DCS
 aws
     `boto` in order to use AWS callbacks
 
@@ -66,6 +97,15 @@ Note that external tools to call in the replica creation or custom bootstrap scr
 independently of Patroni.
 
 
+.. _running_configuring:
+
+Planning the Number of PostgreSQL Nodes
+---------------------------------------
+
+Patroni/PostgreSQL nodes are decoupled from DCS nodes (except when Patroni implements RAFT on its own) and therefore
+there is no requirement on the minimal number of nodes. Running a cluster consisting of one master and one standby is 
+perfectly fine. You can add more standby nodes later.
+
 Running and Configuring
 -----------------------
 
@@ -76,7 +116,7 @@ obtain those files from the git repository and replace `./patroni.py` below with
 To get started, do the following from different terminals:
 ::
 
-    > etcd --data-dir=data/etcd
+    > etcd --data-dir=data/etcd --enable-v2=true
     > ./patroni.py postgres0.yml
     > ./patroni.py postgres1.yml
 
@@ -123,3 +163,20 @@ When connecting from an application, always use a non-superuser. Patroni require
    :target: https://travis-ci.org/zalando/patroni
 .. |Coverage Status| image:: https://coveralls.io/repos/zalando/patroni/badge.svg?branch=master
    :target: https://coveralls.io/r/zalando/patroni?branch=master
+
+Testing Your HA Solution
+--------------------------------------
+Testing an HA solution is a time consuming process, with many variables. This is particularly true considering a cross-platform application. You need a trained system administrator or a consultant to do this work. It is not something we can cover in depth in the documentaiton.
+
+That said, here are some pieces of your infrastructure you should be sure to test:
+
+* Network (the network in front of your system as well as the NICs [physical or virtual] themselves)
+* Disk IO 
+* file limits (nofile in Linux)
+* RAM. Even if you have oomkiller turned off as suggested, the unavailability of RAM could cause issues.
+* CPU
+* Virtualization Contention (overcommitting the hypervisor)
+* Any cgroup limitation (likely to be related to the above)
+* ``kill -9`` of any postgres process (except postmaster!). This is a decent simulation of a segfault.
+
+One thing that you should not do is run ``kill -9`` on a postmaster process. This is because doing so does not mimic any real life scenario. If you are concerned your infrastructure is insecure and an attacker could run ``kill -9``, no amount of HA process is going to fix that. The attacker will simply kill the process again, or cause chaos in another way.

docs/SETTINGS.rst

@@ -4,45 +4,112 @@
 YAML Configuration Settings
 ===========================
 
+.. _dynamic_configuration_settings:
+
+Dynamic configuration settings
+------------------------------
+
+Dynamic configuration is stored in the DCS (Distributed Configuration Store) and applied on all cluster nodes. Some parameters, like **loop_wait**, **ttl**, **postgresql.parameters.max_connections**, **postgresql.parameters.max_worker_processes** and so on could be set only in the dynamic configuration. Some other parameters like **postgresql.listen**, **postgresql.data_dir** could be set only locally, i.e. in the Patroni config file or via :ref:`configuration <environment>` variable. In most cases the local configuration will override the dynamic configuration. In order to change the dynamic configuration you can use either ``patronictl edit-config`` tool or Patroni :ref:`REST API <rest_api>`.
+
+-  **loop\_wait**: the number of seconds the loop will sleep. Default value: 10
+-  **ttl**: the TTL to acquire the leader lock (in seconds). 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 (in seconds). 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.
+-  **maximum\_lag\_on\_syncnode**: the maximum bytes a synchronous follower may lag before it is considered as an unhealthy candidate and swapped by healthy asynchronous follower. Patroni utilize the max replica lsn if there is more than one follower, otherwise it will use leader's current wal lsn. Default is -1, Patroni will not take action to swap synchronous unhealthy follower when the value is set to 0 or below. Please set the value high enough so Patroni won't swap synchrounous follower fequently during high transaction volume.
+-  **max\_timelines\_history**: maximum number of timeline history items kept in DCS.  Default value: 0. When set to 0, it keeps the full history in DCS.
+-  **master\_start\_timeout**: the amount of time a master is allowed to recover from failures before failover is triggered (in seconds). 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. 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.
+- **master\_stop\_timeout**: The number of seconds Patroni is allowed to wait when stopping Postgres and effective only when synchronous_mode is enabled. When set to > 0 and the synchronous_mode is enabled, Patroni sends SIGKILL to the postmaster if the stop operation is running for more than the value set by master_stop_timeout. Set the value according to your durability/availability tradeoff. If the parameter is not set or set <= 0, master_stop_timeout does not apply.
+-  **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. Defaults to `false`.
+    -  **use\_slots**: whether or not to use replication slots. Defaults to `true` on PostgreSQL 9.4+.
+    -  **recovery\_conf**: additional configuration settings written to recovery.conf when configuring follower. There is no recovery.conf anymore in PostgreSQL 12, but you may continue using this section, because Patroni handles it transparently.
+    -  **parameters**: list of configuration settings for Postgres.
+-  **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. If the permanent slot name matches with the name of the current primary it will not be created. Everything else 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.
+-  **ignore_slots**: list of sets of replication slot properties for which Patroni should ignore matching slots. This configuration/feature/etc. is useful when some replication slots are managed outside of Patroni. Any subset of matching properties will cause a slot to be ignored.
+    -  **name**: the name of the replication slot.
+    -  **type**: slot type. Can be ``physical`` or ``logical``. If the slot is logical, you may additionally define ``database`` and/or ``plugin``.
+    -  **database**: the database name (when matching a ``logical`` slot).
+    -  **plugin**: the logical decoding plugin (when matching a ``logical`` slot).
+
+Note: **slots** is a hashmap while **ignore_slots** is an array. For example:
+
+.. code:: YAML
+
+        slots:
+          permanent_logical_slot_name:
+            type: logical
+            database: my_db
+            plugin: test_decoding
+          permanent_physical_slot_name:
+            type: physical
+          ...
+        ignore_slots:
+          - name: ignored_logical_slot_name
+            type: logical
+            database: my_db
+            plugin: test_decoding
+          - name: ignored_physical_slot_name
+            type: physical
+          ...
+
 Global/Universal
 ----------------
 -  **name**: the name of the host. Must be unique for the cluster.
 -  **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>`_)
+-  **traceback\_level**: sets the level where tracebacks will be visible. Default value is **ERROR**. Set it to **DEBUG** if you want to see tracebacks only if you enable **log.level=DEBUG**.
+-  **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 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\_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. 
-        -  **parameters**: list of configuration settings for Postgres. Many of these are required for replication to work.
--  **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.
-        -  **- locale: UTF8**: default locale for new databases.
--  **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 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\_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.
+-  **bootstrap**:
+    -  **dcs**: This section will be written into `/<namespace>/<scope>/config` of the given configuration store after initializing of new cluster. The global dynamic configuration for the cluster. Under the ``bootstrap.dcs`` you can put any of the parameters described in the :ref:`Dynamic Configuration settings <dynamic_configuration_settings>` and after Patroni initialized (bootstrapped) the new cluster, it will write this section into `/<namespace>/<scope>/config` of the configuration store. All later changes of ``bootstrap.dcs`` will not take any effect! If you want to change them please use either ``patronictl edit-config`` or Patroni :ref:`REST API <rest_api>`.
+    -  **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.
+            -  **- locale: UTF8**: default locale for new databases.
+    -  **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 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\_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:
 
@@ -50,17 +117,35 @@ Consul
 ------
 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
+-  **host**: the host:port for the Consul local agent.
+-  **url**: url for the Consul local agent, in format: http(s)://host:port.
+-  **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
+-  **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.
--  **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
+-  **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. By default an empty list is used.
+-  **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\_tags**: (optional) additional static tags to add to the Consul service apart from the role (``master``/``replica``/``standby-leader``).  By default an empty list is used.
+-  **service\_check\_interval**: (optional) how often to perform health check against registered url.
+
+The ``token`` needs to have the following ACL permissions:
+
+::
+
+    service_prefix "${scope}" {
+        policy = "write"
+    }
+    key_prefix "${namespace}/${scope}" {
+        policy = "write"
+    }
+    session_prefix "" {
+        policy = "write"
+    }
 
 Etcd
 ----
@@ -68,90 +153,222 @@ Most of the parameters are optional, but you have to specify one of the **host**
 
 -  **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.
--  **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**
+-  **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 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**.
 
+Etcdv3
+------
+If you want that Patroni works with Etcd cluster via protocol version 3, you need to use the ``etcd3`` section in the Patroni configuration file. All configuration parameters are the same as for ``etcd``.
+
+.. warning::
+    Keys created with protocol version 2 are not visible with protocol version 3 and the other way around, therefore it is not possible to switch from ``etcd`` to ``etcd3`` just by updating Patroni config file.
+
+
+ZooKeeper
+----------
+-  **hosts**: List of ZooKeeper cluster members in format: ['host1:port1', 'host2:port2', 'etc...'].
+-  **use_ssl**: (optional) Whether SSL is used or not. Defaults to ``false``. If set to ``false``, all SSL specific parameters are ignored.
+-  **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.
+-  **key_password**: (optional) The client key password.
+-  **verify**: (optional) Whether to verify certificate or not. Defaults to ``true``.
+
+.. note::
+    It is required to install ``kazoo>=2.6.0`` to support SSL.
+
+
 Exhibitor
 ---------
 -  **hosts**: initial list of Exhibitor (ZooKeeper) nodes in format: 'host1,host2,etc...'. This list updates automatically whenever the Exhibitor (ZooKeeper) cluster topology changes.
--  **poll\_interval**: how often the list of ZooKeeper and Exhibitor nodes should be updated from Exhibitor
+-  **poll\_interval**: how often the list of ZooKeeper and Exhibitor nodes should be updated from Exhibitor.
 -  **port**: Exhibitor port.
 
 .. _kubernetes_settings:
 
 Kubernetes
 ----------
+-  **bypass\_api\_service**: (optional) When communicating with the Kubernetes API, Patroni is usually relying on the `kubernetes` service, the address of which is exposed in the pods via the `KUBERNETES_SERVICE_HOST` environment variable. If `bypass_api_service` is set to ``true``, Patroni will resolve the list of API nodes behind the service and connect directly to them.
 -  **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.
+-  **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.
+
+
+.. _raft_settings:
+
+Raft
+----
+-  **self\_addr**: ``ip:port`` to listen on for Raft connections. The ``self_addr`` must be accessible from other nodes of the cluster. If not set, the node will not participate in consensus.
+-  **bind\_addr**: (optional) ``ip:port`` to listen on for Raft connections. If not specified the ``self_addr`` will be used.
+-  **partner\_addrs**: list of other Patroni nodes in the cluster in format: ['ip1:port', 'ip2:port', 'etc...']
+-  **data\_dir**: directory where to store Raft log and snapshot. If not specified the current working directory is used.
+-  **password**: (optional) Encrypt Raft traffic with a specified password, requires ``cryptography`` python module.
+
+  Short FAQ about Raft implementation
+
+  - Q: How to list all the nodes providing consensus?
+
+    A: ``syncobj_admin -conn host:port`` -status where the host:port is the address of one of the cluster nodes
+
+  - Q: Node that was a part of consensus and has gone and I can't reuse the same IP for other node. How to remove this node from the consensus?
+
+    A: ``syncobj_admin -conn host:port -remove host2:port2`` where the ``host2:port2`` is the address of the node you want to remove from consensus.
+
+  - Q: Where to get the ``syncobj_admin`` utility?
+
+    A: It is installed together with ``pysyncobj`` module (python RAFT implementation), which is Patroni dependancy.
+
+  - Q: it is possible to run Patroni node without adding in to the consensus?
+
+    A: Yes, just comment out or remove ``raft.self_addr`` from Patroni configuration.
+
+  - Q: It is possible to run Patroni and PostgreSQL only on two nodes?
+
+    A: Yes, on the third node you can run ``patroni_raft_controller`` (without Patroni and PostgreSQL). In such a setup, one can temporarily lose one node without affecting the primary.
+
 
 .. _postgresql_settings:
 
 PostgreSQL
 ----------
--  **authentication**:
-    -  **superuser**:
-        -  **username**: name for the superuser, set during initialization (initdb) and later used by Patroni to connect to the postgres.
-        -  **password**: password for the superuser, set during initialization (initdb).
-    -  **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.
--  **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\_role\_change**: run this script when the cluster is being promoted or demoted.
-        -  **on\_start**: run this script when the cluster starts.
-        -  **on\_stop**: run this script when the cluster 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. See :ref:`custom replica creation methods documentation <custom_replica_creation>` for further explanation.
--  **data\_dir**: The location of the Postgres data directory, either existing 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 definded, Patroni will use 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 default value should be used and omit ``host`` from 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 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\_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**.
--  **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".
+-  **postgresql**:
+    -  **authentication**:
+        -  **superuser**:
+            -  **username**: name for the superuser, set during initialization (initdb) and later used by Patroni to connect to the postgres.
+            -  **password**: password for the superuser, set during initialization (initdb).
+            -  **sslmode**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+            -  **sslkey**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+            -  **sslpassword**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``sslkey``.
+            -  **sslcert**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+            -  **sslrootcert**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+            -  **sslcrl**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+            -  **gssencmode**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+            -  **channel_binding**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
+        -  **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.
+            -  **sslmode**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+            -  **sslkey**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+            -  **sslpassword**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``sslkey``.
+            -  **sslcert**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+            -  **sslrootcert**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+            -  **sslcrl**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+            -  **gssencmode**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+            -  **channel_binding**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
+        -  **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.
+            -  **sslmode**: (optional) maps to the `sslmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLMODE>`__ connection parameter, which allows a client to specify the type of TLS negotiation mode with the server. For more information on how each mode works, please visit the `PostgreSQL documentation <https://www.postgresql.org/docs/current/libpq-ssl.html#LIBPQ-SSL-SSLMODE-STATEMENTS>`__. The default mode is ``prefer``.
+            -  **sslkey**: (optional) maps to the `sslkey <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLKEY>`__ connection parameter, which specifies the location of the secret key used with the client's certificate.
+            -  **sslpassword**: (optional) maps to the `sslpassword <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLPASSWORD>`__ connection parameter, which specifies the password for the secret key specified in ``sslkey``.
+            -  **sslcert**: (optional) maps to the `sslcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCERT>`__ connection parameter, which specifies the location of the client certificate.
+            -  **sslrootcert**: (optional) maps to the `sslrootcert <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLROOTCERT>`__ connection parameter, which specifies the location of a file containing one ore more certificate authorities (CA) certificates that the client will use to verify a server's certificate.
+            -  **sslcrl**: (optional) maps to the `sslcrl <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-SSLCRL>`__ connection parameter, which specifies the location of a file containing a certificate revocation list. A client will reject connecting to any server that has a certificate present in this list.
+            -  **gssencmode**: (optional) maps to the `gssencmode <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-GSSENCMODE>`__ connection parameter, which determines whether or with what priority a secure GSS TCP/IP connection will be negotiated with the server
+            -  **channel_binding**: (optional) maps to the `channel_binding <https://www.postgresql.org/docs/current/libpq-connect.html#LIBPQ-CONNECT-CHANNEL-BINDING>`__ connection parameter, which controls the client's use of channel binding.
+    -  **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 postgres restarts (without changing role).
+            -  **on\_role\_change**: run this script when the postgres is being promoted or demoted.
+            -  **on\_start**: run this script when the postgres starts.
+            -  **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. 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 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 the PostgreSQL data directory and recreate the replica. Otherwise it will try to follow the new leader. Default value is **false**.
+    -  **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".
+    -  **pre\_promote**: a fencing script that executes during a failover after acquiring the leader lock but before promoting the replica. If the script exits with a non-zero code, Patroni does not promote the replica and removes the leader key from DCS.
 
 REST API
--------- 
--  **connect\_address**: IP address (or hostname) and port, to access the Patroni's REST API. 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 (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**:
+--------
+-  **restapi**:
+        -  **connect\_address**: IP address (or hostname) and port, to access the Patroni's :ref:`REST API <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 address (ie: "localhost" or "127.0.0.1"). It can serve as an 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 (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).
+
+        -  **authentication**: (optional)
             -  **username**: Basic-auth username to protect unsafe REST API endpoints.
             -  **password**: Basic-auth password to protect unsafe REST API endpoints.
+        -  **certfile**: (optional): 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**: (optional): Specifies the file with the secret key in the PEM format.
+        -  **keyfile_password**: (optional): Specifies a password for decrypting the keyfile.
+        -  **cafile**: (optional): Specifies the file with the CA_BUNDLE with certificates of trusted CAs to use while verifying client certs.
+        -  **ciphers**: (optional): Specifies the permitted cipher suites (e.g. "ECDHE-RSA-AES256-GCM-SHA384:DHE-RSA-AES256-GCM-SHA384:ECDHE-RSA-AES128-GCM-SHA256:DHE-RSA-AES128-GCM-SHA256:!SSLv1:!SSLv2:!SSLv3:!TLSv1:!TLSv1.1")
+        -  **verify\_client**: (optional): ``none`` (default), ``optional`` or ``required``. When ``none`` REST API will not check client certificates. When ``required`` client certificates are required for all REST API calls. When ``optional`` client certificates are required for all unsafe REST API endpoints. When ``required`` is used, then client authentication succeeds, if the certificate signature verification succeeds.  For ``optional`` the client cert will only be checked for ``PUT``, ``POST``, ``PATCH``, and ``DELETE`` requests.
+        -  **http\_extra\_headers**: (optional): HTTP headers let the REST API server pass additional information with an HTTP response.
+        -  **https\_extra\_headers**: (optional): HTTPS headers let the REST API server pass additional information with an HTTP response when TLS is enabled. This will also pass additional information set in ``http_extra_headers``.
 
-        -  **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.
+Here is an example of both **http_extra_headers** and **https_extra_headers**:
 
-ZooKeeper
-----------
--  **hosts**: list of ZooKeeper cluster members in format: ['host1:port1', 'host2:port2', 'etc...'].
+.. code:: YAML
+
+        restapi:
+          listen: <listen>
+          connect_address: <connect_address>
+          authentication:
+            username: <username>
+            password: <password>
+          http_extra_headers:
+            'X-Frame-Options': 'SAMEORIGIN'
+            'X-XSS-Protection': '1; mode=block'
+            'X-Content-Type-Options': 'nosniff'
+          cafile: <ca file>
+          certfile: <cert>
+          keyfile: <key>
+          https_extra_headers:
+            'Strict-Transport-Security': 'max-age=31536000; includeSubDomains'
+
+.. _patronictl_settings:
+
+CTL
+---
+- **ctl**: (optional)
+    -  **insecure**: Allow connections to REST API without verifying SSL certs.
+    -  **cacert**: Specifies the file with the CA_BUNDLE file or directory with certificates of trusted CAs to use while verifying REST API SSL certs. If not provided patronictl will use the value provided for REST API "cafile" parameter.
+    -  **certfile**: Specifies the file with the client certificate in the PEM format. If not provided patronictl will use the value provided for REST API "certfile" parameter.
+    -  **keyfile**: Specifies the file with the client secret key in the PEM format. If not provided patronictl will use the value provided for REST API "keyfile" parameter.
 
 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.
+
+Tags
+----
+- **nofailover**: ``true`` or ``false``, controls whether this node is allowed to participate in the leader race and become a leader. Defaults to ``false``
+- **clonefrom**: ``true`` or ``false``. If set to ``true`` other nodes might prefer to use this node for bootstrap (take ``pg_basebackup`` from). If there are several nodes with ``clonefrom`` tag set to ``true`` the node to bootstrap from will be chosen randomly. The default value is ``false``.
+- **noloadbalance**: ``true`` or ``false``. If set to ``true`` the node will return HTTP Status Code 503 for the ``GET /replica`` REST API health-check and therefore will be excluded from the load-balancing. Defaults to ``false``.
+- **replicatefrom**: The IP address/hostname of another replica. Used to support cascading replication.
+- **nosync**: ``true`` or ``false``. If set to ``true`` the node will never be selected as a synchronous replica.

docs/dynamic_configuration.rst

@@ -14,12 +14,14 @@ Patroni configuration is stored in the DCS (Distributed Configuration Store). Th
 
 - Local :ref:`configuration <settings>` (patroni.yml).
 	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.
+	patroni.yml could be changed and reloaded in runtime (without restart of Patroni) by sending SIGHUP to the Patroni process, performing ``POST /reload`` REST-API request or executing ``patronictl reload``.
 
 - 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 your external IP address when you are running inside ``docker``).
 
+The local configuration can be either a single YAML file or a directory. When it is a directory, all YAML files in that directory are loaded one by one in sorted order. In case a key is defined in multiple files, the occurrence in the last file takes precedence.
+
 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:
 
 - max_connections: 100
@@ -35,6 +37,7 @@ For the parameters below, PostgreSQL does not require equal values among the mas
 - max_wal_senders: 5
 - max_replication_slots: 5
 - wal_keep_segments: 8
+- wal_keep_size: 128MB
 
 These parameters are validated to ensure they are sane, or meet a minimum value.
 
@@ -76,153 +79,11 @@ Also, the following Patroni configuration options can be changed only dynamicall
 - loop_wait: 10
 - retry_timeouts: 10
 - maximum_lag_on_failover: 1048576
+- max_timelines_history: 0
+- 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
 run-time values.
 
 Patroni nodes are dumping the state of the DCS options to disk upon for every change of the configuration into the file ``patroni.dynamic.json`` located in the Postgres data directory. Only the master is allowed to restore these options from the on-disk dump if these are completely absent from the DCS or if they are invalid.
-
-REST API
-========
-
-We provide a REST API endpoint for working with dynamic configuration.
-
-GET /config
------------
-Get current version of dynamic configuration.
-
-.. code-block:: bash
-
-	$ curl -s localhost:8008/config | jq .
-	{
-	  "ttl": 30,
-	  "loop_wait": 10,
-	  "retry_timeout": 10,
-	  "maximum_lag_on_failover": 1048576,
-	  "postgresql": {
-	    "use_slots": true,
-	    "use_pg_rewind": true,
-	    "parameters": {
-	      "hot_standby": "on",
-	      "wal_log_hints": "on",
-	      "wal_keep_segments": 8,
-	      "wal_level": "hot_standby",
-	      "max_wal_senders": 5,
-	      "max_replication_slots": 5,
-	      "max_connections": "100"
-	    }
-	  }
-	}
-
-PATCH /config
--------------
-Change existing configuration.
-
-.. code-block:: bash
-
-	$ curl -s -XPATCH -d \
-		'{"loop_wait":5,"ttl":20,"postgresql":{"parameters":{"max_connections":"101"}}}' \
-		http://localhost:8008/config | jq .
-	{
-	  "ttl": 20,
-	  "loop_wait": 5,
-	  "maximum_lag_on_failover": 1048576,
-	  "retry_timeout": 10,
-	  "postgresql": {
-	    "use_slots": true,
-	    "use_pg_rewind": true,
-	    "parameters": {
-	      "hot_standby": "on",
-	      "wal_log_hints": "on",
-	      "wal_keep_segments": 8,
-	      "wal_level": "hot_standby",
-	      "max_wal_senders": 5,
-	      "max_replication_slots": 5,
-	      "max_connections": "101"
-	    }
-	  }
-	}
-
-The above REST API call patches the existing configuration and returns the new configuration.
-
-Let's check that the node processed this configuration. First of all it should start printing log lines every 5 seconds (loop_wait=5). The change of "max_connections" requires a restart, so the "restart_pending" flag should be exposed:
-
-.. code-block:: bash
-
-	$ curl -s http://localhost:8008/patroni | jq .
-	{
-	  "pending_restart": true,
-	  "database_system_identifier": "6287881213849985952",
-	  "postmaster_start_time": "2016-06-13 13:13:05.211 CEST",
-	  "xlog": {
-	    "location": 2197818976
-	  },
-	  "patroni": {
-	    "scope": "batman",
-	    "version": "1.0"
-	  },
-	  "state": "running",
-	  "role": "master",
-	  "server_version": 90503
-	}
-
-Removing parameters:
-
-If you want to remove (reset) some setting just patch it with ``null``:
-
-.. code-block:: bash
-
-	$ curl -s -XPATCH -d \
-		'{"postgresql":{"parameters":{"max_connections":null}}}' \
-		http://localhost:8008/config | jq .
-	{
-	  "ttl": 20,
-	  "loop_wait": 5,
-	  "retry_timeout": 10,
-	  "maximum_lag_on_failover": 1048576,
-	  "postgresql": {
-	    "use_slots": true,
-	    "use_pg_rewind": true,
-	    "parameters": {
-	      "hot_standby": "on",
-	      "unix_socket_directories": ".",
-	      "wal_keep_segments": 8,
-	      "wal_level": "hot_standby",
-	      "wal_log_hints": "on",
-	      "max_wal_senders": 5,
-	      "max_replication_slots": 5
-	    }
-	  }
-	}
-
-Above call removes ``postgresql.parameters.max_connections`` from the dynamic configuration.
-
-PUT /config
------------
-
-It's also possible to perform the full rewrite of an existing dynamic configuration unconditionally:
-
-.. code-block:: bash
-
-	$ curl -s -XPUT -d \
-		'{"maximum_lag_on_failover":1048576,"retry_timeout":10,"postgresql":{"use_slots":true,"use_pg_rewind":true,"parameters":{"hot_standby":"on","wal_log_hints":"on","wal_keep_segments":8,"wal_level":"hot_standby","unix_socket_directories":".","max_wal_senders":5}},"loop_wait":3,"ttl":20}' \
-		http://localhost:8008/config | jq .
-	{
-	  "ttl": 20,
-	  "maximum_lag_on_failover": 1048576,
-	  "retry_timeout": 10,
-	  "postgresql": {
-	    "use_slots": true,
-	    "parameters": {
-	      "hot_standby": "on",
-	      "unix_socket_directories": ".",
-	      "wal_keep_segments": 8,
-	      "wal_level": "hot_standby",
-	      "wal_log_hints": "on",
-	      "max_wal_senders": 5
-	    },
-	    "use_pg_rewind": true
-	  },
-	  "loop_wait": 3
-	}

docs/existing_data.rst

@@ -0,0 +1,51 @@
+.. _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.
+
+Major Upgrade of PostgreSQL Version
+===================================
+
+The only possible way to do a major upgrade currently is:
+
+1. Stop Patroni
+2. Upgrade PostgreSQL binaries and perform `pg_upgrade <https://www.postgresql.org/docs/current/pgupgrade.html>`_ on the master node
+3. Update patroni.yml
+4. Remove the initialize key from DCS or wipe complete cluster state from DCS. The second one could be achieved by running ``patronictl remove <cluster-name>``. It is necessary because pg_upgrade runs initdb which actually creates a new database with a new PostgreSQL system identifier.
+5. If you wiped the cluster state in the previous step, you may wish to copy patroni.dynamic.json from old data dir to the new one.  It will help you to retain some PostgreSQL parameters you had set before.
+6. Start Patroni on the master node.
+7. Upgrade PostgreSQL binaries, update patroni.yml and wipe the data_dir on standby nodes.
+8. Start Patroni on the standby nodes and wait for the replication to complete.
+
+Running pg_upgrade on standby nodes is not supported by PostgreSQL. If you know what you are doing, you can try the rsync procedure described in https://www.postgresql.org/docs/current/pgupgrade.html instead of wiping data_dir on standby nodes. The safest way is however to let Patroni replicate the data for you.
+
+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

@@ -10,6 +10,8 @@ Patroni is a template for you to create your own customized, high-availability s
 
 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>`__.
 
+Currently supported PostgreSQL versions: 9.3 to 13.
+
 **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.
 
 
@@ -19,8 +21,11 @@ We call Patroni a "template" because it is far from being a one-size-fits-all or
 
    README
    dynamic_configuration
+   rest_api
+   existing_data
    ENVIRONMENT
    SETTINGS
+   security
    replica_bootstrap
    replication_modes
    pause

docs/replica_bootstrap.rst

@@ -24,6 +24,7 @@ arguments to them, i.e. the name of the cluster and the path to the data directo
         <custom_bootstrap_method_name>:
             command: <path_to_custom_bootstrap_script> [param1 [, ...]]
             keep_existing_recovery_conf: False
+            no_params: False
             recovery_conf:
                 recovery_target_action: promote
                 recovery_target_timeline: latest
@@ -40,6 +41,8 @@ in the configuration files, Patroni supplies two cluster-specific ones:
 --datadir
     Path to the data directory of the cluster instance to be bootstrapped
 
+Passing these two additional flags can be disabled by setting a special ``no_params`` parameter to ``True``.
+
 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.
@@ -65,6 +68,19 @@ for outdated backup files. Some people prefer other backup solutions, such as ``
 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:
@@ -79,6 +95,21 @@ scripts to clone a new replica. Those are configured in the ``postgresql`` confi
         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
@@ -99,6 +130,10 @@ A special ``no_master`` parameter, if defined, allows Patroni to call the replic
 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 instruct 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
@@ -108,6 +143,10 @@ possible to specify a ``basebackup`` configuration section. Same rules as with t
 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.
+Also note that in case symlinks are used for the WAL folder it is up to the user to specify the correct ``--waldir``
+path as an option, so that after replica buildup or re-initialization the symlink would persist. This option is supported
+only since v10 though.
+
 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:
@@ -127,5 +166,50 @@ and
         basebackup:
             - verbose
             - max-rate: '100M'
+            - waldir: /pg-wal-mount/external-waldir
 
 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 permanent 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

@@ -13,6 +13,8 @@ In asynchronous mode the cluster is allowed to lose some committed transactions
 
 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
 ----------------------------------
 
@@ -44,7 +46,7 @@ When ``synchronous_mode`` is on and a standby crashes, commits will block until
 
 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_mode``. 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``),
@@ -55,11 +57,16 @@ You can ensure that a standby never becomes the synchronous standby by setting `
 
 Synchronous mode can be switched on and off via Patroni REST interface. See :ref:`dynamic configuration <dynamic_configuration>` for instructions.
 
+Note: Because of the way synchronous replication is implemented in PostgreSQL it is still possible to lose transactions even when using ``synchronous_mode_strict``. If the PostgreSQL backend is cancelled while waiting to acknowledge replication (as a result of packet cancellation due to client timeout or backend failure) transaction changes become visible for other backends. Such changes are not yet replicated and may be lost in case of standby promotion.
+
+Synchronous Replication Factor
+------------------------------
+The parameter ``synchronous_node_count`` is used by Patroni to manage number of synchronous standby databases. It is set to 1 by default. It has no effect when ``synchronous_mode`` is set to off. When enabled, Patroni manages precise number of synchronous standby databases based on parameter ``synchronous_node_count`` and adjusts the state in DCS & synchronous_standby_names as members join and leave.
 
 Synchronous mode implementation
 -------------------------------
 
-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:
+When in synchronous mode Patroni maintains synchronization state in the DCS, containing the latest primary and current synchronous standby databases. 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.
 
@@ -67,9 +74,9 @@ When in synchronous mode Patroni maintains synchronization state in the DCS, con
 
 - A node that is not the leader or current synchronous standby is not allowed to promote itself automatically.
 
-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.
+Patroni will only assign one or more synchronous standby nodes based on ``synchronous_node_count`` parameter to ``synchronous_standby_names``.
 
-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.
+On each HA loop iteration Patroni re-evaluates synchronous standby nodes choice. If the current list of synchronous standby nodes are 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 primary with the cluster.

docs/rest_api.rst

@@ -0,0 +1,379 @@
+.. _rest_api:
+
+Patroni REST API
+================
+
+Patroni has a rich REST API, which is used by Patroni itself during the leader race, by the ``patronictl`` tool in order to perform failovers/switchovers/reinitialize/restarts/reloads, by HAProxy or any other kind of load balancer to perform HTTP health checks, and of course could also be used for monitoring. Below you will find the list of Patroni REST API endpoints.
+
+Health check endpoints
+----------------------
+For all health check ``GET`` requests Patroni returns a JSON document with the status of the node, along with the HTTP status code. If you don't want or don't need the JSON document, you might consider using the ``OPTIONS`` method instead of ``GET``.
+
+- The following requests to Patroni REST API will return HTTP status code **200** only when the Patroni node is running as the leader:
+
+  - ``GET /``
+  - ``GET /master``
+  - ``GET /leader``
+  - ``GET /primary``
+  - ``GET /read-write``
+
+- ``GET /replica``: replica health check endpoint. It returns HTTP status code **200** only when the Patroni node is in the state ``running``, the role is ``replica`` and ``noloadbalance`` tag is not set.
+
+- ``GET /replica?lag=<max-lag>``: replica check endpoint. In addition to checks from ``replica``, it also checks replication latency and returns status code **200** only when it is below specified value. The key cluster.last_leader_operation from DCS is used for Leader wal position and compute latency on replica for performance reasons. max-lag can be specified in bytes (integer) or in human readable values, for e.g. 16kB, 64MB, 1GB.
+
+  - ``GET /replica?lag=1048576``
+  - ``GET /replica?lag=1024kB``
+  - ``GET /replica?lag=10MB``
+  - ``GET /replica?lag=1GB``
+
+- ``GET /read-only``: like the above endpoint, but also includes the primary.
+
+- ``GET /standby-leader``: returns HTTP status code **200** only when the Patroni node is running as the leader in a :ref:`standby cluster <standby_cluster>`.
+
+- ``GET /synchronous`` or ``GET /sync``: returns HTTP status code **200** only when the Patroni node is running as a synchronous standby.
+
+- ``GET /asynchronous`` or ``GET /async``: returns HTTP status code **200** only when the Patroni node is running as an asynchronous standby.
+
+- ``GET /asynchronous?lag=<max-lag>`` or ``GET /async?lag=<max-lag>``: asynchronous standby check endpoint. In addition to checks from ``asynchronous`` or ``async``, it also checks replication latency and returns status code **200** only when it is below specified value. The key cluster.last_leader_operation from DCS is used for Leader wal position and compute latency on replica for performance reasons. max-lag can be specified in bytes (integer) or in human readable values, for e.g. 16kB, 64MB, 1GB.
+
+  - ``GET /async?lag=1048576``
+  - ``GET /async?lag=1024kB``
+  - ``GET /async?lag=10MB``
+  - ``GET /async?lag=1GB``
+
+- ``GET /health``: returns HTTP status code **200** only when PostgreSQL is up and running.
+
+- ``GET /liveness``: always returns HTTP status code **200** what only indicates that Patroni is running. Could be used for ``livenessProbe``.
+
+- ``GET /readiness``: returns HTTP status code **200** when the Patroni node is running as the leader or when PostgreSQL is up and running. The endpoint could be used for ``readinessProbe`` when it is not possible to use Kubenetes endpoints for leader elections (OpenShift).
+
+Both, ``readiness`` and ``liveness`` endpoints are very light-weight and not executing any SQL. Probes should be configured in such a way that they start failing about time when the leader key is expiring. With the default value of ``ttl``, which is ``30s`` example probes would look like:
+
+.. code-block:: yaml
+
+    readinessProbe:
+      httpGet:
+        scheme: HTTP
+        path: /readiness
+        port: 8008
+      initialDelaySeconds: 3
+      periodSeconds: 10
+      timeoutSeconds: 5
+      successThreshold: 1
+      failureThreshold: 3
+    livenessProbe:
+      httpGet:
+        scheme: HTTP
+        path: /liveness
+        port: 8008
+      initialDelaySeconds: 3
+      periodSeconds: 10
+      timeoutSeconds: 5
+      successThreshold: 1
+      failureThreshold: 3
+
+
+Monitoring endpoint
+-------------------
+
+The ``GET /patroni`` is used by Patroni during the leader race. It also could be used by your monitoring system. The JSON document produced by this endpoint has the same structure as the JSON produced by the health check endpoints.
+
+.. code-block:: bash
+
+    $ curl -s http://localhost:8008/patroni | jq .
+    {
+      "state": "running",
+      "postmaster_start_time": "2019-09-24 09:22:32.555 CEST",
+      "role": "master",
+      "server_version": 110005,
+      "cluster_unlocked": false,
+      "xlog": {
+        "location": 25624640
+      },
+      "timeline": 3,
+      "database_system_identifier": "6739877027151648096",
+      "patroni": {
+        "version": "1.6.0",
+        "scope": "batman"
+      }
+    }
+
+Cluster status endpoints
+------------------------
+
+- The ``GET /cluster`` endpoint generates a JSON document describing the current cluster topology and state:
+
+.. code-block:: bash
+
+    $ curl -s http://localhost:8008/cluster | jq .
+    {
+      "members": [
+        {
+          "name": "postgresql0",
+          "host": "127.0.0.1",
+          "port": 5432,
+          "role": "leader",
+          "state": "running",
+          "api_url": "http://127.0.0.1:8008/patroni",
+          "timeline": 5,
+          "tags": {
+            "clonefrom": true
+          }
+        },
+        {
+          "name": "postgresql1",
+          "host": "127.0.0.1",
+          "port": 5433,
+          "role": "replica",
+          "state": "running",
+          "api_url": "http://127.0.0.1:8009/patroni",
+          "timeline": 5,
+          "tags": {
+            "clonefrom": true
+          },
+          "lag": 0
+        }
+      ],
+      "scheduled_switchover": {
+        "at": "2019-09-24T10:36:00+02:00",
+        "from": "postgresql0"
+      }
+    }
+
+
+- The ``GET /history`` endpoint provides a view on the history of cluster switchovers/failovers. The format is very similar to the content of history files in the ``pg_wal`` directory. The only difference is the timestamp field showing when the new timeline was created.
+
+.. code-block:: bash
+
+    $ curl -s http://localhost:8008/history | jq .
+    [
+      [
+        1,
+        25623960,
+        "no recovery target specified",
+        "2019-09-23T16:57:57+02:00"
+      ],
+      [
+        2,
+        25624344,
+        "no recovery target specified",
+        "2019-09-24T09:22:33+02:00"
+      ],
+      [
+        3,
+        25624752,
+        "no recovery target specified",
+        "2019-09-24T09:26:15+02:00"
+      ],
+      [
+        4,
+        50331856,
+        "no recovery target specified",
+        "2019-09-24T09:35:52+02:00"
+      ]
+    ]
+
+
+Config endpoint
+---------------
+
+``GET /config``: Get the current version of the dynamic configuration:
+
+.. code-block:: bash
+
+	$ curl -s localhost:8008/config | jq .
+	{
+	  "ttl": 30,
+	  "loop_wait": 10,
+	  "retry_timeout": 10,
+	  "maximum_lag_on_failover": 1048576,
+	  "postgresql": {
+	    "use_slots": true,
+	    "use_pg_rewind": true,
+	    "parameters": {
+	      "hot_standby": "on",
+	      "wal_log_hints": "on",
+	      "wal_level": "hot_standby",
+	      "max_wal_senders": 5,
+	      "max_replication_slots": 5,
+	      "max_connections": "100"
+	    }
+	  }
+	}
+
+
+``PATCH /config``: Change the existing configuration.
+
+.. code-block:: bash
+
+	$ curl -s -XPATCH -d \
+		'{"loop_wait":5,"ttl":20,"postgresql":{"parameters":{"max_connections":"101"}}}' \
+		http://localhost:8008/config | jq .
+	{
+	  "ttl": 20,
+	  "loop_wait": 5,
+	  "maximum_lag_on_failover": 1048576,
+	  "retry_timeout": 10,
+	  "postgresql": {
+	    "use_slots": true,
+	    "use_pg_rewind": true,
+	    "parameters": {
+	      "hot_standby": "on",
+	      "wal_log_hints": "on",
+	      "wal_level": "hot_standby",
+	      "max_wal_senders": 5,
+	      "max_replication_slots": 5,
+	      "max_connections": "101"
+	    }
+	  }
+	}
+
+The above REST API call patches the existing configuration and returns the new configuration.
+
+Let's check that the node processed this configuration. First of all it should start printing log lines every 5 seconds (loop_wait=5). The change of "max_connections" requires a restart, so the "pending_restart" flag should be exposed:
+
+.. code-block:: bash
+
+	$ curl -s http://localhost:8008/patroni | jq .
+	{
+	  "pending_restart": true,
+	  "database_system_identifier": "6287881213849985952",
+	  "postmaster_start_time": "2016-06-13 13:13:05.211 CEST",
+	  "xlog": {
+	    "location": 2197818976
+	  },
+	  "patroni": {
+	    "scope": "batman",
+	    "version": "1.0"
+	  },
+	  "state": "running",
+	  "role": "master",
+	  "server_version": 90503
+	}
+
+Removing parameters:
+
+If you want to remove (reset) some setting just patch it with ``null``:
+
+.. code-block:: bash
+
+	$ curl -s -XPATCH -d \
+		'{"postgresql":{"parameters":{"max_connections":null}}}' \
+		http://localhost:8008/config | jq .
+	{
+	  "ttl": 20,
+	  "loop_wait": 5,
+	  "retry_timeout": 10,
+	  "maximum_lag_on_failover": 1048576,
+	  "postgresql": {
+	    "use_slots": true,
+	    "use_pg_rewind": true,
+	    "parameters": {
+	      "hot_standby": "on",
+	      "unix_socket_directories": ".",
+	      "wal_level": "hot_standby",
+	      "wal_log_hints": "on",
+	      "max_wal_senders": 5,
+	      "max_replication_slots": 5
+	    }
+	  }
+	}
+
+The above call removes ``postgresql.parameters.max_connections`` from the dynamic configuration.
+
+``PUT /config``: It's also possible to perform the full rewrite of an existing dynamic configuration unconditionally:
+
+.. code-block:: bash
+
+	$ curl -s -XPUT -d \
+		'{"maximum_lag_on_failover":1048576,"retry_timeout":10,"postgresql":{"use_slots":true,"use_pg_rewind":true,"parameters":{"hot_standby":"on","wal_log_hints":"on","wal_level":"hot_standby","unix_socket_directories":".","max_wal_senders":5}},"loop_wait":3,"ttl":20}' \
+		http://localhost:8008/config | jq .
+	{
+	  "ttl": 20,
+	  "maximum_lag_on_failover": 1048576,
+	  "retry_timeout": 10,
+	  "postgresql": {
+	    "use_slots": true,
+	    "parameters": {
+	      "hot_standby": "on",
+	      "unix_socket_directories": ".",
+	      "wal_level": "hot_standby",
+	      "wal_log_hints": "on",
+	      "max_wal_senders": 5
+	    },
+	    "use_pg_rewind": true
+	  },
+	  "loop_wait": 3
+	}
+
+
+Switchover and failover endpoints
+---------------------------------
+
+``POST /switchover`` or ``POST /failover``. These endpoints are very similar to each other. There are a couple of minor differences though:
+
+1. The failover endpoint allows to perform a manual failover when there are no healthy nodes, but at the same time it will not allow you to schedule a switchover.
+
+2. The switchover endpoint is the opposite. It works only when the cluster is healthy (there is a leader) and allows to schedule a switchover at a given time.
+
+
+In the JSON body of the ``POST`` request you must specify at least the ``leader`` or ``candidate`` fields and optionally the ``scheduled_at`` field if you want to schedule a switchover at a specific time.
+
+
+Example: perform a failover to the specific node:
+
+.. code-block:: bash
+
+    $ curl -s http://localhost:8009/failover -XPOST -d '{"candidate":"postgresql1"}'
+    Successfully failed over to "postgresql1"
+
+
+Example: schedule a switchover from the leader to any other healthy replica in the cluster at a specific time:
+
+.. code-block:: bash
+
+    $ curl -s http://localhost:8008/switchover -XPOST -d \
+	    '{"leader":"postgresql0","scheduled_at":"2019-09-24T12:00+00"}'
+    Switchover scheduled
+
+
+Depending on the situation the request might finish with a different HTTP status code and body. The status code **200** is returned when the switchover or failover successfully completed. If the switchover was successfully scheduled, Patroni will return HTTP status code **202**. In case something went wrong, the error status code (one of **400**, **412** or **503**) will be returned with some details in the response body. For more information please check the source code of ``patroni/api.py:do_POST_failover()`` method.
+
+- ``DELETE /switchover``: delete the scheduled switchover
+
+The ``POST /switchover`` and ``POST failover`` endpoints are used by ``patronictl switchover`` and ``patronictl failover``, respectively.
+The ``DELETE /switchover`` is used by ``patronictl flush <cluster-name> switchover``.
+
+
+Restart endpoint
+----------------
+
+- ``POST /restart``: You can restart Postgres on the specific node by performing the ``POST /restart`` call. In the JSON body of ``POST`` request it is possible to optionally specify some restart conditions:
+
+  - **restart_pending**: boolean, if set to ``true`` Patroni will restart PostgreSQL only when restart is pending in order to apply some changes in the PostgreSQL config.
+  - **role**: perform restart only if the current role of the node matches with the role from the POST request.
+  - **postgres_version**: perform restart only if the current version of postgres is smaller than specified in the POST request.
+  - **timeout**: how long we should wait before PostgreSQL starts accepting connections. Overrides ``master_start_timeout``.
+  - **schedule**: timestamp with time zone, schedule the restart somewhere in the future.
+
+- ``DELETE /restart``: delete the scheduled restart
+
+``POST /restart`` and ``DELETE /restart`` endpoints are used by ``patronictl restart`` and ``patronictl flush <cluster-name> restart`` respectively.
+
+
+Reload endpoint
+---------------
+
+The ``POST /reload`` call will order Patroni to re-read and apply the configuration file. This is the equivalent of sending the ``SIGHUP`` signal to the Patroni process. In case you changed some of the Postgres parameters which require a restart (like **shared_buffers**), you still have to explicitly do the restart of Postgres by either calling the ``POST /restart`` endpoint or with the help of ``patronictl restart``.
+
+The reload endpoint is used by ``patronictl reload``.
+
+
+Reinitialize endpoint
+---------------------
+
+``POST /reinitialize``: reinitialize the PostgreSQL data directory on the specified node. It is allowed to be executed only on replicas. Once called, it will remove the data directory and start ``pg_basebackup`` or some alternative :ref:`replica creation method <custom_replica_creation>`.
+
+The call might fail if Patroni is in a loop trying to recover (restart) a failed Postgres. In order to overcome this problem one can specify ``{"force":true}`` in the request body.
+
+The reinitialize endpoint is used by ``patronictl reinit``.

docs/security.rst

@@ -0,0 +1,37 @@
+.. _security:
+
+=======================
+Security Considerations
+=======================
+
+A Patroni cluster has two interfaces to be protected from unauthorized access: the distributed configuration storage (DCS) and the Patroni REST API.
+
+Protecting DCS
+==============
+
+Patroni and patronictl both store and retrieve data to/from the DCS. 
+
+Despite DCS doesn't contain any sensitive information, it allows changing some of Patroni/Postgres configuration. Therefore the very first thing that should be protected is DCS itself.
+
+The details of protection depend on the type of DCS used. The authentication and encryption parameters (tokens/basic-auth/client certificates) for the supported types of DCS are covered in :ref:`SETTINGS <bootstrap_settings>` 
+
+The general recommendation is to enable TLS for all DCS communication.
+
+Protecting the REST API
+=======================
+
+Protecting the REST API is a more complicated task.
+
+The Patroni REST API is used by Patroni itself during the leader race, by the ``patronictl`` tool in order to perform failovers/switchovers/reinitialize/restarts/reloads, by HAProxy or any other kind of load balancer to perform HTTP health checks, and of course could also be used for monitoring. 
+
+From the point of view of security, REST API contains safe (``GET`` requests, only retrieve information) and unsafe (``PUT``, ``POST``, ``PATCH`` and ``DELETE`` requests, change the state of nodes) endpoints.
+
+The unsafe endpoints can be protected with HTTP basic-auth by setting the ``restapi.authentication.username`` and ``restapi.authentication.password`` parameters. There is no way to protect the safe endpoints without enabling TLS.
+
+When TLS for the REST API is enabled and a PKI is established, mutual authentication of the API server and API client is possible for all endpoints.
+
+The ``restapi`` section parameters enable TLS client authentication to the server. Depending on the value of the ``verify_client`` parameter, the API server requires a successful client certificate verification for both safe and unsafe API calls (``verify_client: required``), or only for unsafe API calls (``verify_client: optional``), or for no API calls (``verify_client: none``).
+
+The ``ctl`` section parameters enable TLS server authentication to the client (the ``patronictl`` tool which uses the same config as patroni). Set ``insecure: true`` to disable the server certificate verification by the client. See :ref:`SETTINGS <patronictl_settings>` for a detailed description of the TLS client parameters.
+
+Protecting the PostgreSQL database proper from unauthorized access is beyond the scope of this document and is covered in https://www.postgresql.org/docs/current/client-authentication.html

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_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/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/patroni.service

@@ -14,12 +14,19 @@ Group=postgres
 # Read in configuration file if it exists, otherwise proceed
 EnvironmentFile=-/etc/patroni_env.conf
 
-WorkingDirectory=~
+# the default is the user's home directory, and if you want to change it, you must provide an absolute path.
+# WorkingDirectory=/home/sameuser
 
 # 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

features/archive-restore.py

@@ -0,0 +1,21 @@
+#!/usr/bin/env python
+import os
+import argparse
+import shutil
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--dirname", required=True)
+    parser.add_argument("--pathname", required=True)
+    parser.add_argument("--filename", required=True)
+    parser.add_argument("--mode", required=True, choices=("archive", "restore"))
+    args, _ = parser.parse_known_args()
+
+    full_filename = os.path.join(args.dirname, args.filename)
+    if args.mode == "archive":
+        if not os.path.isdir(args.dirname):
+            os.makedirs(args.dirname)
+        if not os.path.exists(full_filename):
+            shutil.copy(args.pathname, full_filename)
+    else:
+        shutil.copy(full_filename, args.pathname)

features/backup_create.py

@@ -0,0 +1,14 @@
+#!/usr/bin/env python
+import argparse
+import subprocess
+import sys
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--datadir", required=True)
+    parser.add_argument("--dbname", required=True)
+    parser.add_argument("--walmethod", required=True, choices=("fetch", "stream", "none"))
+    args, _ = parser.parse_known_args()
+
+    walmethod = ["-X", args.walmethod] if args.walmethod != "none" else []
+    sys.exit(subprocess.call(["pg_basebackup", "-D", args.datadir, "-c", "fast", "-d", args.dbname] + walmethod))

features/backup_create.sh

@@ -1,22 +0,0 @@
-#!/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.py

@@ -0,0 +1,11 @@
+#!/usr/bin/env python
+import argparse
+import shutil
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--datadir", required=True)
+    parser.add_argument("--sourcedir", required=True)
+    args, _ = parser.parse_known_args()
+
+    shutil.copytree(args.sourcedir, args.datadir)

features/backup_restore.sh

@@ -1,21 +0,0 @@
-#!/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

@@ -4,6 +4,7 @@ Feature: basic replication
   Scenario: check replication of a single table
     Given I start postgres0
     Then postgres0 is a leader after 10 seconds
+    And there is a non empty initialize key in DCS after 15 seconds
     When I issue a PATCH request to http://127.0.0.1:8008/config with {"ttl": 20, "loop_wait": 2, "synchronous_mode": true}
     Then I receive a response code 200
     When I start postgres1
@@ -14,29 +15,73 @@ 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
-    And "sync" key in DCS has sync_standby=postgres1 after 2 seconds
-    And I run patronictl.py restart batman postgres1 --force
-    Then I receive a response returncode 0
-    And "sync" key in DCS has sync_standby=postgres2 after 10 seconds
+    Given I shut down postgres2
+    Then "sync" key in DCS has sync_standby=postgres1 after 5 seconds
+    When I start postgres2
+    And I shut down postgres1
+    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 stuck sync replica
+    Given I issue a PATCH request to http://127.0.0.1:8008/config with {"maximum_lag_on_syncnode": 15000000, "postgresql": {"parameters": {"synchronous_commit": "remote_apply"}}}
+    Then I receive a response code 200
+    And I create table on postgres0
+    And table mytest is present on postgres1 after 2 seconds
+    And table mytest is present on postgres2 after 2 seconds
+    When I pause wal replay on postgres2
+    And I load data on postgres0
+    Then "sync" key in DCS has sync_standby=postgres1 after 15 seconds
+    And I resume wal replay on postgres2
+    And I sleep for 2 seconds
+    And I issue a GET request to http://127.0.0.1:8009/sync
+    Then I receive a response code 200
+    When I issue a GET request to http://127.0.0.1:8010/async
+    Then I receive a response code 200
+    When I issue a PATCH request to http://127.0.0.1:8008/config with {"maximum_lag_on_syncnode": -1, "postgresql": {"parameters": {"synchronous_commit": "on"}}}
+    Then I receive a response code 200
+    And I drop table on postgres0
+
+  Scenario: check multi sync replication
+    Given I issue a PATCH request to http://127.0.0.1:8008/config with {"synchronous_node_count": 2}
+    Then I receive a response code 200
+    And I sleep for 10 seconds
+    Then "sync" key in DCS has sync_standby=postgres1,postgres2 after 5 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/sync
+    Then I receive a response code 200
+    When I issue a PATCH request to http://127.0.0.1:8008/config with {"synchronous_node_count": 1}
+    Then I receive a response code 200
+    And I shut down postgres1
+    And I sleep for 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
+    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
     Given I run patronictl.py pause batman
     Then I receive a response returncode 0
     When I sleep for 2 seconds
     And I shut down postgres0
-    And I run patronictl.py resume batman 
+    And I run patronictl.py resume batman
     Then I receive a response returncode 0
     And postgres2 role is the primary after 24 seconds
+    And Response on GET http://127.0.0.1:8010/history contains recovery after 10 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
+    And Response on GET http://127.0.0.1:8010/config contains master_start_timeout after 10 seconds
 
   Scenario: check immediate failover when master_start_timeout=0
     Given I kill postmaster on postgres2

features/callback.py

@@ -0,0 +1,17 @@
+#!/usr/bin/env python
+import os
+import psycopg2
+import sys
+
+
+if __name__ == '__main__':
+    if not (len(sys.argv) >= 3 and sys.argv[3] == "master"):
+        sys.exit(1)
+
+    os.environ['PGPASSWORD'] = 'zalando'
+    connection = psycopg2.connect(host='127.0.0.1', port=sys.argv[1], user='postgres')
+    cursor = connection.cursor()
+    cursor.execute("SELECT slot_name FROM pg_replication_slots WHERE slot_type = 'logical'")
+
+    with open("data/postgres0/label", "w") as label:
+        label.write(next(iter(cursor.fetchone()), ""))

features/callback2.py

@@ -0,0 +1,5 @@
+#!/usr/bin/env python
+
+import sys
+with open("data/{0}/{0}_cb.log".format(sys.argv[1]), "a+") as log:
+    log.write(" ".join(sys.argv[-3:]) + "\n")

features/custom_bootstrap.feature

@@ -13,5 +13,5 @@ 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
+    Then postgres2 is a leader of batman2 after 30 seconds
     And table bar is present on postgres2 after 10 seconds

features/environment.py

@@ -1,22 +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
 
+from six.moves.BaseHTTPServer import BaseHTTPRequestHandler, HTTPServer
+
 
 @six.add_metaclass(abc.ABCMeta)
 class AbstractController(object):
@@ -84,7 +82,7 @@ class AbstractController(object):
 
 
 class PatroniController(AbstractController):
-    __PORT = 5440
+    __PORT = 5360
     PATRONI_CONFIG = '{}.yml'
     """ starts and stops individual patronis"""
 
@@ -110,19 +108,31 @@ 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):
         if self.watchdog:
@@ -130,7 +140,8 @@ class PatroniController(AbstractController):
         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(['coverage', 'run', '--source=patroni', '-p', 'patroni.py', self._config],
+        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):
@@ -156,13 +167,20 @@ class PatroniController(AbstractController):
             config = yaml.safe_load(f)
             config.pop('etcd', None)
 
+        raft_port = os.environ.get('RAFT_PORT')
+        if raft_port:
+            os.environ['RAFT_PORT'] = str(int(raft_port) + 1)
+            config['raft'] = {'data_dir': self._output_dir, 'self_addr': 'localhost:' + os.environ['RAFT_PORT']}
+
         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']['basebackup'] = [{'checkpoint': 'fast'}]
+        config['postgresql']['use_unix_socket'] = os.name != 'nt'  # windows doesn't yet support unix-domain sockets
+        config['postgresql']['pgpass'] = os.path.join(tempfile.gettempdir(), 'pgpass_' + name)
         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',
@@ -174,13 +192,12 @@ class PatroniController(AbstractController):
                 config['bootstrap']['initdb'].extend([{'auth': 'md5'}, {'auth-host': 'md5'}])
 
         if custom_config is not None:
-            def recursive_update(dst, src):
-                for k, v in src.items():
-                    if k in dst and isinstance(dst[k], dict):
-                        recursive_update(dst[k], v)
-                    else:
-                        dst[k] = v
-            recursive_update(config, custom_config)
+            self.recursive_update(config, custom_config)
+
+        self.recursive_update(config, {
+            'bootstrap': {'dcs': {'postgresql': {'parameters': {'wal_keep_segments': 100}}}}})
+        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)
@@ -238,59 +255,24 @@ class PatroniController(AbstractController):
         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='basebackup'):
-        subprocess.call([PatroniPoolController.BACKUP_SCRIPT, '--walmethod=none',
-                         '--datadir=' + os.path.join(self._output_dir, dest),
-                         '--dbname=' + self.backup_source])
+    def backup(self, dest=os.path.join('data', 'basebackup')):
+        subprocess.call(PatroniPoolController.BACKUP_SCRIPT + ['--walmethod=none',
+                        '--datadir=' + os.path.join(self._work_directory, dest),
+                        '--dbname=' + self.backup_source])
 
 
 class ProcessHang(object):
@@ -363,9 +345,12 @@ class ConsulController(AbstractDcsController):
     def __init__(self, context):
         super(ConsulController, self).__init__(context)
         os.environ['PATRONI_CONSUL_HOST'] = 'localhost:8500'
-        self._client = consul.Consul()
+        os.environ['PATRONI_CONSUL_REGISTER_SERVICE'] = 'on'
         self._config_file = None
 
+        import consul
+        self._client = consul.Consul()
+
     def _start(self):
         self._config_file = self._work_directory + '.json'
         with open(self._config_file, 'wb') as f:
@@ -398,26 +383,45 @@ class ConsulController(AbstractDcsController):
         super(ConsulController, self).start(max_wait_limit)
 
 
-class EtcdController(AbstractDcsController):
+class AbstractEtcdController(AbstractDcsController):
 
     """ handles all etcd related tasks, used for the tests setup and cleanup """
 
-    def __init__(self, context):
-        super(EtcdController, self).__init__(context)
-        os.environ['PATRONI_ETCD_HOST'] = 'localhost:2379'
-        self._client = etcd.Client(port=2379)
+    def __init__(self, context, client_cls):
+        super(AbstractEtcdController, self).__init__(context)
+        self._client_cls = client_cls
 
     def _start(self):
         return subprocess.Popen(["etcd", "--debug", "--data-dir", self._work_directory],
                                 stdout=self._log, stderr=subprocess.STDOUT)
 
+    def _is_running(self):
+        from patroni.dcs.etcd import DnsCachingResolver
+        # if etcd is running, but we didn't start it
+        try:
+            self._client = self._client_cls({'host': 'localhost', 'port': 2379, 'retry_timeout': 30,
+                                             'patronictl': 1}, DnsCachingResolver())
+            return True
+        except Exception:
+            return False
+
+
+class EtcdController(AbstractEtcdController):
+
+    def __init__(self, context):
+        from patroni.dcs.etcd import EtcdClient
+        super(EtcdController, self).__init__(context, EtcdClient)
+        os.environ['PATRONI_ETCD_HOST'] = 'localhost:2379'
+
     def query(self, key, scope='batman'):
+        import etcd
         try:
             return self._client.get(self.path(key, scope)).value
         except etcd.EtcdKeyNotFound:
             return None
 
     def cleanup_service_tree(self):
+        import etcd
         try:
             self._client.delete(self.path(scope=''), recursive=True)
         except (etcd.EtcdKeyNotFound, etcd.EtcdConnectionFailed):
@@ -425,12 +429,25 @@ class EtcdController(AbstractDcsController):
         except Exception as e:
             assert False, "exception when cleaning up etcd contents: {0}".format(e)
 
-    def _is_running(self):
-        # if etcd is running, but we didn't start it
+
+class Etcd3Controller(AbstractEtcdController):
+
+    def __init__(self, context):
+        from patroni.dcs.etcd3 import Etcd3Client
+        super(Etcd3Controller, self).__init__(context, Etcd3Client)
+        os.environ['PATRONI_ETCD3_HOST'] = 'localhost:2379'
+
+    def query(self, key, scope='batman'):
+        import base64
+        response = self._client.range(self.path(key, scope))
+        for k in response.get('kvs', []):
+            return base64.b64decode(k['value']).decode('utf-8') if 'value' in k else None
+
+    def cleanup_service_tree(self):
         try:
-            return bool(self._client.machines)
-        except Exception:
-            return False
+            self._client.deleteprefix(self.path(scope=''))
+        except Exception as e:
+            assert False, "exception when cleaning up etcd contents: {0}".format(e)
 
 
 class KubernetesController(AbstractDcsController):
@@ -442,8 +459,9 @@ class KubernetesController(AbstractDcsController):
         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'
+        os.environ['PATRONI_KUBERNETES_BYPASS_API_SERVICE'] = 'true'
 
-        from kubernetes import client as k8s_client, config as k8s_config
+        from patroni.dcs.kubernetes import k8s_client, k8s_config
         k8s_config.load_kube_config(context='local')
         self._client = k8s_client
         self._api = self._client.CoreV1Api()
@@ -461,13 +479,13 @@ class KubernetesController(AbstractDcsController):
 
     def delete_pod(self, name):
         try:
-            self._api.delete_namespaced_pod(name, self._namespace, self._client.V1DeleteOptions())
-        except:
+            self._api.delete_namespaced_pod(name, self._namespace, body=self._client.V1DeleteOptions())
+        except Exception:
             pass
         while True:
             try:
                 self._api.read_namespaced_pod(name, self._namespace)
-            except:
+            except Exception:
                 break
 
     def query(self, key, scope='batman'):
@@ -476,22 +494,23 @@ class KubernetesController(AbstractDcsController):
             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':
+                ep = scope + {'leader': '', 'history': '-config', 'initialize': '-config'}.get(key, '-' + key)
+                e = self._api.read_namespaced_endpoints(ep, self._namespace)
+                if key != 'sync':
                     return e.metadata.annotations[key]
                 else:
                     return json.dumps(e.metadata.annotations)
-            except:
+            except Exception:
                 return None
 
     def cleanup_service_tree(self):
         try:
             self._api.delete_collection_namespaced_pod(self._namespace, label_selector=self._label_selector)
-        except:
+        except Exception:
             pass
         try:
             self._api.delete_collection_namespaced_endpoints(self._namespace, label_selector=self._label_selector)
-        except:
+        except Exception:
             pass
 
         while True:
@@ -511,18 +530,22 @@ class ZooKeeperController(AbstractDcsController):
         super(ZooKeeperController, self).__init__(context, False)
         if export_env:
             os.environ['PATRONI_ZOOKEEPER_HOSTS'] = "'localhost:2181'"
+
+        import kazoo.client
         self._client = kazoo.client.KazooClient()
 
     def _start(self):
         pass  # TODO: implement later
 
     def query(self, key, scope='batman'):
+        import kazoo.exceptions
         try:
             return self._client.get(self.path(key, scope))[0].decode('utf-8')
         except kazoo.exceptions.NoNodeError:
             return None
 
     def cleanup_service_tree(self):
+        import kazoo.exceptions
         try:
             self._client.delete(self.path(scope=''), recursive=True)
         except (kazoo.exceptions.NoNodeError):
@@ -540,16 +563,76 @@ class ZooKeeperController(AbstractDcsController):
             return False
 
 
+class MockExhibitor(BaseHTTPRequestHandler):
+
+    def do_GET(self):
+        self.send_response(200)
+        self.end_headers()
+        self.wfile.write(b'{"servers":["127.0.0.1"],"port":2181}')
+
+    def log_message(self, fmt, *args):
+        pass
+
+
 class ExhibitorController(ZooKeeperController):
 
     def __init__(self, context):
         super(ExhibitorController, self).__init__(context, False)
-        os.environ.update({'PATRONI_EXHIBITOR_HOSTS': 'localhost', 'PATRONI_EXHIBITOR_PORT': '8181'})
+        port = 8181
+        exhibitor = HTTPServer(('', port), MockExhibitor)
+        exhibitor.daemon_thread = True
+        exhibitor_thread = threading.Thread(target=exhibitor.serve_forever)
+        exhibitor_thread.daemon = True
+        exhibitor_thread.start()
+        os.environ.update({'PATRONI_EXHIBITOR_HOSTS': 'localhost', 'PATRONI_EXHIBITOR_PORT': str(port)})
+
+
+class RaftController(AbstractDcsController):
+
+    CONTROLLER_ADDR = 'localhost:1234'
+
+    def __init__(self, context):
+        super(RaftController, self).__init__(context)
+        os.environ.update(PATRONI_RAFT_PARTNER_ADDRS="'" + self.CONTROLLER_ADDR + "'", RAFT_PORT='1234')
+        self._raft = None
+
+    def _start(self):
+        env = os.environ.copy()
+        del env['PATRONI_RAFT_PARTNER_ADDRS']
+        env['PATRONI_RAFT_SELF_ADDR'] = self.CONTROLLER_ADDR
+        env['PATRONI_RAFT_DATA_DIR'] = self._work_directory
+        return subprocess.Popen([sys.executable, '-m', 'coverage', 'run',
+                                '--source=patroni', '-p', 'patroni_raft_controller.py'],
+                                stdout=self._log, stderr=subprocess.STDOUT, env=env)
+
+    def query(self, key, scope='batman'):
+        ret = self._raft.get(self.path(key, scope))
+        return ret and ret['value']
+
+    def set(self, key, value):
+        self._raft.set(self.path(key), value)
+
+    def cleanup_service_tree(self):
+        from patroni.dcs.raft import KVStoreTTL
+        from pysyncobj import SyncObjConf
+
+        if self._raft:
+            self._raft.destroy()
+            self._raft._SyncObj__thread.join()
+            self.stop()
+            os.makedirs(self._work_directory)
+            self.start()
+
+        ready_event = threading.Event()
+        conf = SyncObjConf(appendEntriesUseBatch=False, dynamicMembershipChange=True, onReady=ready_event.set)
+        self._raft = KVStoreTTL(None, [self.CONTROLLER_ADDR], conf)
+        ready_event.wait()
 
 
 class PatroniPoolController(object):
 
-    BACKUP_SCRIPT = 'features/backup_create.sh'
+    BACKUP_SCRIPT = [sys.executable, 'features/backup_create.py']
+    ARCHIVE_RESTORE_SCRIPT = ' '.join((sys.executable, os.path.abspath('features/archive-restore.py')))
 
     def __init__(self, context):
         self._context = context
@@ -582,9 +665,8 @@ class PatroniPoolController(object):
         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',
-                        'get_watchdog', 'database_is_running', 'checkpoint_hang', 'patroni_hang',
-                        'terminate_backends', 'backup']:
+        if func not in ['stop', 'query', 'write_label', 'read_label', 'check_role_has_changed_to',
+                        'add_tag_to_config', 'get_watchdog', 'patroni_hang', 'backup']:
             raise AttributeError("PatroniPoolController instance has no attribute '{0}'".format(func))
 
         def wrapper(name, *args, **kwargs):
@@ -598,7 +680,7 @@ class PatroniPoolController(object):
         self._processes.clear()
 
     def create_and_set_output_directory(self, feature_name):
-        feature_dir = os.path.join(self.patroni_path, 'features/output', feature_name.replace(' ', '_'))
+        feature_dir = os.path.join(self.patroni_path, 'features', 'output', feature_name.replace(' ', '_'))
         if os.path.exists(feature_dir):
             shutil.rmtree(feature_dir)
         os.makedirs(feature_dir)
@@ -611,7 +693,7 @@ class PatroniPoolController(object):
             'bootstrap': {
                 'method': 'pg_basebackup',
                 'pg_basebackup': {
-                    'command': self.BACKUP_SCRIPT + ' --walmethod=stream --dbname=' + f.backup_source
+                    'command': " ".join(self.BACKUP_SCRIPT) + ' --walmethod=stream --dbname=' + f.backup_source
                 },
                 'dcs': {
                     'postgresql': {
@@ -624,8 +706,9 @@ class PatroniPoolController(object):
             'postgresql': {
                 'parameters': {
                     'archive_mode': 'on',
-                    'archive_command': 'mkdir -p {0} && test ! -f {0}/%f && cp %p {0}/%f'.format(
-                            os.path.join(self._output_dir, 'wal_archive'))
+                    'archive_command': (self.ARCHIVE_RESTORE_SCRIPT + ' --mode archive ' +
+                                        '--dirname {} --filename %f --pathname %p').format(
+                                        os.path.join(self.patroni_path, 'data', 'wal_archive'))
                 },
                 'authentication': {
                     'superuser': {'password': 'zalando1'},
@@ -641,11 +724,14 @@ class PatroniPoolController(object):
             'bootstrap': {
                 'method': 'backup_restore',
                 'backup_restore': {
-                    'command': 'features/backup_restore.sh --sourcedir=' + os.path.join(self._output_dir, 'basebackup'),
+                    'command': (sys.executable + ' features/backup_restore.py --sourcedir=' +
+                                os.path.join(self.patroni_path, 'data', 'basebackup')),
                     'recovery_conf': {
                         'recovery_target_action': 'promote',
                         'recovery_target_timeline': 'latest',
-                        'restore_command': 'cp {0}/wal_archive/%f %p'.format(self._output_dir)
+                        'restore_command': (self.ARCHIVE_RESTORE_SCRIPT + ' --mode restore ' +
+                                            '--dirname {} --filename %f --pathname %p').format(
+                                            os.path.join(self.patroni_path, 'data', 'wal_archive'))
                     }
                 }
             },
@@ -785,8 +871,8 @@ class WatchdogMonitor(object):
 # 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.ci = any(a in os.environ for a in ('TRAVIS_BUILD_NUMBER', 'BUILD_NUMBER', 'GITHUB_ACTIONS'))
+    context.timeout_multiplier = 5 if context.ci else 1  # MacOS sometimes is VERY slow
     context.pctl = PatroniPoolController(context)
     context.dcs_ctl = context.pctl.known_dcs[context.pctl.dcs](context)
     context.dcs_ctl.start()
@@ -799,8 +885,8 @@ def before_all(context):
 
 def after_all(context):
     context.dcs_ctl.stop()
-    subprocess.call(['coverage', 'combine'])
-    subprocess.call(['coverage', 'report'])
+    subprocess.call([sys.executable, '-m', 'coverage', 'combine'])
+    subprocess.call([sys.executable, '-m', 'coverage', 'report'])
 
 
 def before_feature(context, feature):
@@ -813,3 +899,5 @@ 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/ignored_slots.feature

@@ -0,0 +1,61 @@
+Feature: ignored slots
+  Scenario: check ignored slots aren't removed on failover/switchover
+    Given I start postgres1
+    Then postgres1 is a leader after 10 seconds
+    And there is a non empty initialize key in DCS after 15 seconds
+    When I issue a PATCH request to http://127.0.0.1:8009/config with {"loop_wait": 2, "ignore_slots": [{"name": "unmanaged_slot_0", "database": "postgres", "plugin": "test_decoding", "type": "logical"}, {"name": "unmanaged_slot_1", "database": "postgres", "plugin": "test_decoding"}, {"name": "unmanaged_slot_2", "database": "postgres"}, {"name": "unmanaged_slot_3"}], "postgresql": {"parameters": {"wal_level": "logical"}}}
+    Then I receive a response code 200
+    And Response on GET http://127.0.0.1:8009/config contains ignore_slots after 10 seconds
+    # Make sure the wal_level has been changed.
+    When I shut down postgres1
+    And I start postgres1
+    Then postgres1 is a leader after 10 seconds
+    And "members/postgres1" key in DCS has role=master after 3 seconds
+    # Make sure Patroni has finished telling Postgres it should be accepting writes.
+    And postgres1 role is the primary after 20 seconds
+    # 1. Create our test logical replication slot.
+    # Test that ny subset of attributes in the ignore slots matcher is enough to match a slot
+    # by using 3 different slots.
+    When I create a logical replication slot unmanaged_slot_0 on postgres1 with the test_decoding plugin
+    And I create a logical replication slot unmanaged_slot_1 on postgres1 with the test_decoding plugin
+    And I create a logical replication slot unmanaged_slot_2 on postgres1 with the test_decoding plugin
+    And I create a logical replication slot unmanaged_slot_3 on postgres1 with the test_decoding plugin
+    And I create a logical replication slot dummy_slot on postgres1 with the test_decoding plugin
+    # It seems like it'd be obvious that these slots exist since we just created them,
+    # but Patroni can actually end up dropping them almost immediately, so it's helpful
+    # to verify they exist before we begin testing whether they persist through failover
+    # cycles.
+    Then postgres1 has a logical replication slot named unmanaged_slot_0 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_1 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_2 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_3 with the test_decoding plugin
+
+    When I start postgres0
+    Then "members/postgres0" key in DCS has role=replica after 3 seconds
+    And postgres0 role is the secondary after 20 seconds
+    # Verify that the replica has advanced beyond the point in the WAL
+    # where we created the replication slot so that on the next failover
+    # cycle we don't accidentally rewind to before the slot creation.
+    And replication works from postgres1 to postgres0 after 20 seconds
+    When I shut down postgres1
+    Then "members/postgres0" key in DCS has role=master after 3 seconds
+
+    # 2. After a failover the server (now a replica) still has the slot.
+    When I start postgres1
+    Then postgres1 role is the secondary after 20 seconds
+    And "members/postgres1" key in DCS has role=replica after 3 seconds
+    # give Patroni time to sync replication slots
+    And I sleep for 2 seconds
+    And postgres1 has a logical replication slot named unmanaged_slot_0 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_1 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_2 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_3 with the test_decoding plugin
+    And postgres1 does not have a logical replication slot named dummy_slot
+
+    # 3. After a failover the server (now a master) still has the slot.
+    When I shut down postgres0
+    Then "members/postgres1" key in DCS has role=master after 3 seconds
+    And postgres1 has a logical replication slot named unmanaged_slot_0 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_1 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_2 with the test_decoding plugin
+    And postgres1 has a logical replication slot named unmanaged_slot_3 with the test_decoding plugin

features/patroni_api.feature

@@ -8,6 +8,10 @@ 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/standby_leader
+	Then I receive a response code 503
+	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
@@ -26,10 +30,7 @@ Scenario: check API requests on a stand-alone server
 	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
-	Then I receive a response code 200
-	And I receive a response text nothing changed
-	When I add tag new_tag new_value to postgres0 config
+	Given I add tag new_tag new_value to postgres0 config
 	And I issue an empty POST request to http://127.0.0.1:8008/reload
 	Then I receive a response code 202
 
@@ -43,12 +44,32 @@ Scenario: check dynamic configuration change via DCS
 	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'}
+	And I sleep for 4 seconds
+
+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"}}}
+	Then I receive a response code 200
+		And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 5 seconds
+	Given I issue a scheduled restart at http://127.0.0.1:8008 in 3 seconds with {"role": "replica"}
+	Then I receive a response code 202
+		And I sleep for 4 seconds
+		And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 10 seconds
+	Given I issue a scheduled restart at http://127.0.0.1:8008 in 3 seconds with {"restart_pending": "True"}
+	Then I receive a response code 202
+		And Response on GET http://127.0.0.1:8008/patroni does not contain pending_restart after 10 seconds
+		And postgres0 role is the primary after 10 seconds
 
 Scenario: check API requests for the primary-replica pair in the pause mode
-	Given I run patronictl.py pause batman
+	Given I start postgres1
+	Then replication works from postgres0 to postgres1 after 20 seconds
+	When I run patronictl.py pause batman
+	Then I receive a response returncode 0
+	When I kill postmaster on postgres1
+	And I issue a GET request to http://127.0.0.1:8009/replica
+	Then I receive a response code 503
+	When I run patronictl.py restart batman postgres1 --force
 	Then I receive a response returncode 0
-	When I start postgres1
 	Then replication works from postgres0 to postgres1 after 20 seconds
 	When I issue a GET request to http://127.0.0.1:8009/replica
 	Then I receive a response code 200
@@ -71,6 +92,15 @@ Scenario: check the switchover via the API in the pause mode
 	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
+	And "members/postgres0" key in DCS has state=running after 10 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 switchover
 	Given I issue a scheduled switchover from postgres1 to postgres0 in 3 seconds
@@ -84,16 +114,12 @@ Scenario: check the scheduled switchover
 	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
-
-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"}}}
+	And "members/postgres1" key in DCS has state=running after 10 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
-		And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 5 seconds
-	Given I issue a scheduled restart at http://127.0.0.1:8008 in 3 seconds with {"role": "replica"}
-	Then I receive a response code 202
-		And I sleep for 4 seconds
-		And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 10 seconds
-	Given I issue a scheduled restart at http://127.0.0.1:8008 in 3 seconds with {"restart_pending": "True"}
-	Then I receive a response code 202
-		And Response on GET http://127.0.0.1:8008/patroni does not contain pending_restart after 10 seconds
-

features/standby_cluster.feature

@@ -0,0 +1,50 @@
+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 there is a non empty initialize key in DCS after 15 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
+    And I sleep for 3 seconds
+    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/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_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 503
+    And I sleep for 3 seconds
+    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

@@ -33,6 +33,37 @@ def add_table(context, table_name, pg_name):
         assert False, "Error creating table {0} on {1}: {2}".format(table_name, pg_name, e)
 
 
+@step('I {action:w} wal replay on {pg_name:w}')
+def toggle_wal_replay(context, action, pg_name):
+    # pause or resume the wal replay process
+    try:
+       version = context.pctl.query(pg_name, "select pg_catalog.pg_read_file('PG_VERSION', 0, 2)").fetchone()
+       wal = version and version[0] and int(version[0].split('.')[0]) < 10 and "xlog" or "wal"
+       context.pctl.query(pg_name, "SELECT pg_{0}_replay_{1}()".format(wal, action))
+    except pg.Error as e:
+        assert False, "Error during {0} wal recovery on {1}: {2}".format(action, pg_name, e)
+
+
+@step('I {action:w} table on {pg_name:w}')
+def crdr_mytest(context, action, pg_name):
+    try:
+       if (action == "create"):
+           context.pctl.query(pg_name, "create table if not exists mytest(id Numeric)")
+       else:
+           context.pctl.query(pg_name, "drop table if exists mytest")
+    except pg.Error as e:
+        assert False, "Error {0} table mytest on {1}: {2}".format(action, pg_name, e)
+
+
+@step('I load data on {pg_name:w}')
+def initiate_load(context, pg_name):
+    # perform dummy load
+    try:
+       context.pctl.query(pg_name, "begin; insert into mytest select r::numeric from generate_series(1, 350000) r; commit;")
+    except pg.Error as e:
+        assert False, "Error loading test data on {0}: {1}".format(pg_name, e)
+
+
 @then('Table {table_name:w} is present on {pg_name:w} after {max_replication_delay:d} seconds')
 def table_is_present_on(context, table_name, pg_name, max_replication_delay):
     max_replication_delay *= context.timeout_multiplier

features/steps/cascading_replication.py

@@ -9,10 +9,13 @@ def start_patroni_with_a_name_value_tag(context, name, 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')
-def check_label(context, content, name):
-    label = context.pctl.read_label(name)
-    assert label == content, "{0} is not equal to {1}".format(label, content)
+@then('There is a {label} with "{content}" in {name:w} data directory')
+def check_label(context, label, content, name):
+    label = context.pctl.read_label(name, label)
+    if label is None:
+        label = ""
+    label = label.replace('\n', '\\n')
+    assert content in label, "\"{0}\" doesn't contain {1}".format(label, content)
 
 
 @step('I create label with "{content:w}" in {name:w} data directory')
@@ -20,16 +23,33 @@ def write_label(context, content, name):
     context.pctl.write_label(name, content)
 
 
-@step('"{name}" key in DCS has {key:w}={value:w} after {time_limit:d} seconds')
+@step('"{name}" key in DCS has {key:w}={value} after {time_limit:d} seconds')
 def check_member(context, name, key, value, time_limit):
     time_limit *= context.timeout_multiplier
     max_time = time.time() + int(time_limit)
+    dcs_value = None
     while time.time() < max_time:
         try:
             response = json.loads(context.dcs_ctl.query(name))
-            if response.get(key) == value:
+            dcs_value = response.get(key)
+            if dcs_value == value:
                 return
         except Exception:
             pass
         time.sleep(1)
-    assert False, "{0} does not have {1}={2} in dcs after {3} seconds".format(name, key, value, time_limit)
+    assert False, "{0} does not have {1}={2} (found {3}) in dcs after {4} seconds".format(name, key, value,
+                                                                                          dcs_value, time_limit)
+
+
+@step('there is a non empty {key:w} key in DCS after {time_limit:d} seconds')
+def check_initialize(context, key, time_limit):
+    time_limit *= context.timeout_multiplier
+    max_time = time.time() + int(time_limit)
+    while time.time() < max_time:
+        try:
+            if context.dcs_ctl.query(key):
+                return
+        except Exception:
+            pass
+        time.sleep(1)
+    assert False, "There is no {0} in dcs after {1} seconds".format(key, time_limit)

features/steps/patroni_api.py

@@ -1,18 +1,19 @@
-import base64
 import json
 import os
 import parse
-import requests
 import shlex
 import subprocess
+import sys
 import time
 import yaml
 
 from behave import register_type, step, then
 from dateutil import tz
 from datetime import datetime, timedelta
+from patroni.request import PatroniRequest
 
 tzutc = tz.tzutc()
+request_executor = PatroniRequest({'ctl': {'auth': 'username:password'}})
 
 
 @parse.with_pattern(r'https?://(?:\w|\.|:|/)+')
@@ -44,9 +45,9 @@ def sleep_for_n_seconds(context, value):
 
 
 def _set_response(context, response):
-    context.status_code = response.status_code
-    data = response.content.decode('utf-8')
-    ct = response.headers.get('content-type', '')
+    context.status_code = response.status
+    data = response.data.decode('utf-8')
+    ct = response.getheader('content-type', '')
     if ct.startswith('application/json') or\
             ct.startswith('text/yaml') or\
             ct.startswith('text/x-yaml') or\
@@ -62,13 +63,7 @@ def _set_response(context, response):
 
 @step('I issue a GET request to {url:url}')
 def do_get(context, url):
-    try:
-        r = requests.get(url)
-    except requests.exceptions.RequestException:
-        context.status_code = None
-        context.response = None
-    else:
-        _set_response(context, r)
+    do_request(context, 'GET', url, None)
 
 
 @step('I issue an empty POST request to {url:url}')
@@ -78,24 +73,18 @@ 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'}
+    data = data and json.loads(data)
     try:
-        if request_method == 'PATCH':
-            r = requests.patch(url, headers=headers, json=data)
-        else:
-            r = requests.post(url, headers=headers, json=data)
-    except requests.exceptions.RequestException:
-        context.status_code = None
-        context.response = None
+        r = request_executor.request(request_method, url, data)
+    except Exception:
+        context.status_code = context.response = None
     else:
         _set_response(context, r)
 
 
 @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()
@@ -148,8 +137,8 @@ def add_tag_to_config(context, tag, value, pg_name):
 def check_http_response(context, url, value, timeout, negate=False):
     timeout *= context.timeout_multiplier
     for _ in range(int(timeout)):
-        r = requests.get(url)
-        if (value in r.content.decode('utf-8')) != negate:
+        r = request_executor.request('GET', url)
+        if (value in r.data.decode('utf-8')) != negate:
             break
         time.sleep(1)
     else:

features/steps/slots.py

@@ -0,0 +1,36 @@
+from behave import step, then
+import psycopg2 as pg
+
+
+@step('I create a logical replication slot {slot_name} on {pg_name:w} with the {plugin:w} plugin')
+def create_logical_replication_slot(context, slot_name, pg_name, plugin):
+    try:
+        output = context.pctl.query(pg_name, ("SELECT pg_create_logical_replication_slot('{0}', '{1}'),"
+                                              " current_database()").format(slot_name, plugin))
+        print(output.fetchone())
+    except pg.Error as e:
+        print(e)
+        assert False, "Error creating slot {0} on {1} with plugin {2}".format(slot_name, pg_name, plugin)
+
+
+@then('{pg_name:w} has a logical replication slot named {slot_name} with the {plugin:w} plugin')
+def has_logical_replication_slot(context, pg_name, slot_name, plugin):
+    try:
+        row = context.pctl.query(pg_name, ("SELECT slot_type, plugin FROM pg_replication_slots"
+                                           " WHERE slot_name = '{0}'").format(slot_name)).fetchone()
+        assert row, "Couldn't find replication slot named {0}".format(slot_name)
+        assert row[0] == "logical", "Found replication slot named {0} but wasn't a logical slot".format(slot_name)
+        assert row[1] == plugin, ("Found replication slot named {0} but was using plugin "
+                                  "{1} rather than {2}").format(slot_name, row[1], plugin)
+    except pg.Error:
+        assert False, "Error looking for slot {0} on {1} with plugin {2}".format(slot_name, pg_name, plugin)
+
+
+@then('{pg_name:w} does not have a logical replication slot named {slot_name}')
+def does_not_have_logical_replication_slot(context, pg_name, slot_name):
+    try:
+        row = context.pctl.query(pg_name, ("SELECT 1 FROM pg_replication_slots"
+                                           " WHERE slot_name = '{0}'").format(slot_name)).fetchone()
+        assert not row, "Found unexpected replication slot named {0}".format(slot_name)
+    except pg.Error:
+        assert False, "Error looking for slot {0} on {1}".format(slot_name, pg_name)

features/steps/standby_cluster.py

@@ -0,0 +1,84 @@
+import os
+import sys
+import time
+
+from behave import step
+
+
+select_replication_query = """
+SELECT * FROM pg_catalog.pg_stat_replication
+WHERE application_name = '{0}'
+"""
+
+executable = sys.executable if os.name != 'nt' else sys.executable.replace('\\', '/')
+callback = executable + " features/callback2.py "
+
+
+@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": executable + " features/callback.py"
+            }
+        }
+    })
+
+
+@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": (executable + " features/backup_restore.py --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

@@ -44,31 +44,6 @@ def watchdog_was_triggered(context, name, timeout):
     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)

kubernetes/Dockerfile

@@ -1,32 +1,34 @@
-FROM postgres:9.6
+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-get install -y git curl jq python-psycopg2 python-yaml python-requests python-six python-pysocks \
-        python-dateutil python-pip python-prettytable python-wheel python-psutil python locales \
-
+    && 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 \
-
-    && pip install setuptools pip --upgrade \
-    && pip install 'git+https://github.com/zalando/patroni.git#egg=patroni[kubernetes]' \
-
-    && mkdir -p /home/postgres \
-    && chown postgres:postgres /home/postgres \
-
+    && 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 python-pip python-setuptools \
+    && 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 callback.py /
+ADD entrypoint.sh /
 
 EXPOSE 5432 8008
-ENV LC_ALL=en_US.UTF-8 LANG=en_US.UTF-8
+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/callback.py

@@ -1,62 +0,0 @@
-#!/usr/bin/env python
-
-import logging
-import os
-import socket
-import sys
-import time
-
-from kubernetes import client as k8s_client, config as k8s_config
-from urllib3.exceptions import HTTPError
-from six.moves.http_client import HTTPException
-
-logger = logging.getLogger(__name__)
-
-
-class CoreV1Api(k8s_client.CoreV1Api):
-
-    def retry(func):
-        def wrapped(*args, **kwargs):
-            count = 0
-            while True:
-                try:
-                    return func(*args, **kwargs)
-                except (HTTPException, HTTPError, socket.error, socket.timeout):
-                    if count >= 10:
-                        raise
-                    logger.info('Throttling API requests...')
-                    time.sleep(2 ** count * 0.5)
-                    count += 1
-        return wrapped
-
-    @retry
-    def patch_namespaced_endpoints(self, *args, **kwargs):
-        return super(CoreV1Api, self).patch_namespaced_endpoints(*args, **kwargs)
-
-
-def patch_master_endpoint(api, namespace, cluster):
-    addresses = [k8s_client.V1EndpointAddress(ip=os.environ['POD_IP'])]
-    ports = [k8s_client.V1EndpointPort(port=5432)]
-    subsets = [k8s_client.V1EndpointSubset(addresses=addresses, ports=ports)]
-    body = k8s_client.V1Endpoints(subsets=subsets)
-    return api.patch_namespaced_endpoints(cluster, namespace, body)
-
-
-def main():
-    logging.basicConfig(format='%(asctime)s %(levelname)s: %(message)s', level=logging.INFO)
-    if len(sys.argv) != 4 or sys.argv[1] not in ('on_start', 'on_stop', 'on_role_change'):
-        sys.exit('Usage: %s <action> <role> <cluster_name>', sys.argv[0])
-
-    action, role, cluster = sys.argv[1:4]
-
-    k8s_config.load_incluster_config()
-    k8s_api = CoreV1Api()
-
-    namespace = os.environ['KUBERNETES_NAMESPACE']
-
-    if role == 'master' and action in ('on_start', 'on_role_change'):
-        patch_master_endpoint(k8s_api, namespace, cluster)
-
-
-if __name__ == '__main__':
-    main()

kubernetes/entrypoint.sh

@@ -1,5 +1,12 @@
 #!/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:
@@ -13,24 +20,18 @@ bootstrap:
   - data-checksums
   pg_hba:
   - host all all 0.0.0.0/0 md5
-  - host replication ${PATRONI_REPLICATION_USERNAME} ${POD_IP}/16    md5
+  - host replication ${PATRONI_REPLICATION_USERNAME} ${PATRONI_KUBERNETES_POD_IP}/16 md5
 restapi:
-  connect_address: '${POD_IP}:8008'
+  connect_address: '${PATRONI_KUBERNETES_POD_IP}:8008'
 postgresql:
-  connect_address: '${POD_IP}:5432'
+  connect_address: '${PATRONI_KUBERNETES_POD_IP}:5432'
   authentication:
     superuser:
       password: '${PATRONI_SUPERUSER_PASSWORD}'
     replication:
       password: '${PATRONI_REPLICATION_PASSWORD}'
-  callbacks:
-    on_start: /callback.py
-    on_stop: /callback.py
-    on_role_change: /callback.py
 __EOF__
 
 unset PATRONI_SUPERUSER_PASSWORD PATRONI_REPLICATION_PASSWORD
-export KUBERNETES_NAMESPACE=$PATRONI_KUBERNETES_NAMESPACE
-export POD_NAME=$PATRONI_NAME
 
-exec /usr/bin/python /usr/local/bin/patroni /home/postgres/patroni.yml
+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 +1,327 @@
+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_BYPASS_API_SERVICE
+            value: 'true'
+          - 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}
+          readinessProbe:
+            httpGet:
+              scheme: HTTP
+              path: /readiness
+              port: 8008
+            initialDelaySeconds: 3
+            periodSeconds: 10
+            timeoutSeconds: 5
+            successThreshold: 1
+            failureThreshold: 3
+          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}
+# Following privileges are only required if deployed not in the "default"
+# namespace and you want Patroni to bypass kubernetes service
+# (PATRONI_KUBERNETES_BYPASS_API_SERVICE=true)
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: ClusterRole
+  metadata:
+    name: patroni-k8s-ep-access
+  rules:
+  - apiGroups:
+    - ""
+    resources:
+    - endpoints
+    resourceNames:
+    - kubernetes
+    verbs:
+    - get
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: ClusterRoleBinding
+  metadata:
+    name: ${NAMESPACE}-${SERVICE_ACCOUNT}-k8s-ep-access
+  roleRef:
+    apiGroup: rbac.authorization.k8s.io
+    kind: ClusterRole
+    name: patroni-k8s-ep-access
+  subjects:
+  - kind: ServiceAccount
+    name: ${SERVICE_ACCOUNT}
+    namespace: ${NAMESPACE}
+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 +1,355 @@
+apiVersion: v1
+kind: Template
+metadata:
+  name: patroni-pgsql-persistent
+  annotations:
+    description: |-
+      Patroni Postgresql database cluster, with persistent storage.
+    iconClass: icon-postgresql
+    openshift.io/display-name: Patroni Postgresql (Persistent)
+    openshift.io/long-description: This template deploys a a patroni postgresql HA cluster with 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: ${PATRONI_CLUSTER_NAME}
+      spec:
+        initContainers:
+        - command:
+          - sh
+          - -c 
+          - "mkdir -p /home/postgres/pgdata/pgroot/data && chmod 0700 /home/postgres/pgdata/pgroot/data"
+          image: docker-registry.default.svc:5000/${NAMESPACE}/patroni:latest
+          imagePullPolicy: IfNotPresent
+          name: fix-perms
+          resources: {}
+          terminationMessagePath: /dev/termination-log
+          terminationMessagePolicy: File
+          volumeMounts:
+          - mountPath: /home/postgres/pgdata
+            name: ${APPLICATION_NAME}
+        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_BYPASS_API_SERVICE
+            value: 'true'
+          - 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}
+          readinessProbe:
+            httpGet:
+              scheme: HTTP
+              path: /readiness
+              port: 8008
+            initialDelaySeconds: 3
+            periodSeconds: 10
+            timeoutSeconds: 5
+            successThreshold: 1
+            failureThreshold: 3
+          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}
+# Following privileges are only required if deployed not in the "default"
+# namespace and you want Patroni to bypass kubernetes service
+# (PATRONI_KUBERNETES_BYPASS_API_SERVICE=true)
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: ClusterRole
+  metadata:
+    name: patroni-k8s-ep-access
+  rules:
+  - apiGroups:
+    - ""
+    resources:
+    - endpoints
+    resourceNames:
+    - kubernetes
+    verbs:
+    - get
+- apiVersion: rbac.authorization.k8s.io/v1
+  kind: ClusterRoleBinding
+  metadata:
+    name: ${NAMESPACE}-${SERVICE_ACCOUNT}-k8s-ep-access
+  roleRef:
+    apiGroup: rbac.authorization.k8s.io
+    kind: ClusterRole
+    name: patroni-k8s-ep-access
+  subjects:
+  - kind: ServiceAccount
+    name: ${SERVICE_ACCOUNT}
+    namespace: ${NAMESPACE}
+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

@@ -1,4 +1,16 @@
-apiVersion: apps/v1beta1
+# 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/v1
 kind: StatefulSet
 metadata:
   name: &cluster_name patronidemo
@@ -19,6 +31,16 @@ spec:
       - name: *cluster_name
         image: patroni  # docker build -t patroni .
         imagePullPolicy: IfNotPresent
+        readinessProbe:
+          httpGet:
+            scheme: HTTP
+            path: /readiness
+            port: 8008
+          initialDelaySeconds: 3
+          periodSeconds: 10
+          timeoutSeconds: 5
+          successThreshold: 1
+          failureThreshold: 3
         ports:
         - containerPort: 8008
           protocol: TCP
@@ -28,7 +50,7 @@ spec:
         - mountPath: /home/postgres/pgdata
           name: pgdata
         env:
-        - name: POD_IP
+        - name: PATRONI_KUBERNETES_POD_IP
           valueFrom:
             fieldRef:
               fieldPath: status.podIP
@@ -36,6 +58,10 @@ spec:
           valueFrom:
             fieldRef:
               fieldPath: metadata.namespace
+        - name: PATRONI_KUBERNETES_BYPASS_API_SERVICE
+          value: 'true'
+        - name: PATRONI_KUBERNETES_USE_ENDPOINTS
+          value: 'true'
         - name: PATRONI_KUBERNETES_LABELS
           value: '{application: patroni, cluster-name: patronidemo}'
         - name: PATRONI_SUPERUSER_USERNAME
@@ -109,6 +135,25 @@ spec:
   - port: 5432
     targetPort: 5432
 
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: patronidemo-repl
+  labels:
+    application: patroni
+    cluster-name:  &cluster_name patronidemo
+    role: replica
+spec:
+  type: ClusterIP
+  selector:
+    application: patroni
+    cluster-name: *cluster_name
+    role: replica
+  ports:
+  - port: 5432
+    targetPort: 5432
+
 ---
 apiVersion: v1
 kind: Secret
@@ -145,8 +190,9 @@ rules:
   - patch
   - update
   - watch
-  # delete is required only for 'patronictl remove'
+  # delete and deletecollection are required only for 'patronictl remove'
   - delete
+  - deletecollection
 - apiGroups:
   - ""
   resources:
@@ -159,8 +205,9 @@ rules:
   - create
   - list
   - watch
-  # delete is required only for for 'patronictl remove'
+  # delete and deletecollection are required only for for 'patronictl remove'
   - delete
+  - deletecollection
 - apiGroups:
   - ""
   resources:
@@ -171,6 +218,16 @@ rules:
   - 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
@@ -184,3 +241,37 @@ roleRef:
 subjects:
 - kind: ServiceAccount
   name: patronidemo
+
+# Following privileges are only required if deployed not in the "default"
+# namespace and you want Patroni to bypass kubernetes service
+# (PATRONI_KUBERNETES_BYPASS_API_SERVICE=true)
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: patroni-k8s-ep-access
+rules:
+- apiGroups:
+  - ""
+  resources:
+  - endpoints
+  resourceNames:
+  - kubernetes
+  verbs:
+  - get
+
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: patroni-k8s-ep-access
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: patroni-k8s-ep-access
+subjects:
+- kind: ServiceAccount
+  name: patronidemo
+# The namespace must be specified explicitly.
+# If deploying to the different namespace you have to change it.
+  namespace: default

patroni/__init__.py

@@ -4,30 +4,35 @@ import signal
 import sys
 import time
 
+from .daemon import AbstractPatroniDaemon, abstract_main
+from .version import __version__
+
 logger = logging.getLogger(__name__)
 
+PATRONI_ENV_PREFIX = 'PATRONI_'
+KUBERNETES_ENV_PREFIX = 'KUBERNETES_'
 
-class Patroni(object):
 
-    def __init__(self):
+class Patroni(AbstractPatroniDaemon):
+
+    def __init__(self, config):
         from patroni.api import RestApiServer
-        from patroni.config import Config
         from patroni.dcs import get_dcs
         from patroni.ha import Ha
         from patroni.postgresql import Postgresql
-        from patroni.version import __version__
+        from patroni.request import PatroniRequest
         from patroni.watchdog import Watchdog
 
-        self.setup_signal_handlers()
+        super(Patroni, self).__init__(config)
 
         self.version = __version__
-        self.config = Config()
         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.request = PatroniRequest(self.config, True)
         self.ha = Ha(self)
 
         self.tags = self.get_tags()
@@ -49,6 +54,7 @@ class Patroni(object):
                 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()
@@ -62,13 +68,16 @@ class Patroni(object):
     def nosync(self):
         return bool(self.tags.get('nosync', False))
 
-    def reload_config(self):
+    def reload_config(self, sighup=False, local=False):
         try:
-            self.tags = self.get_tags()
-            self.dcs.reload_config(self.config)
+            super(Patroni, self).reload_config(sighup, local)
+            if local:
+                self.tags = self.get_tags()
+                self.request.reload_config(self.config)
+                self.api.reload_config(self.config['restapi'])
             self.watchdog.reload_config(self.config)
-            self.api.reload_config(self.config['restapi'])
-            self.postgresql.reload_config(self.config['postgresql'])
+            self.postgresql.reload_config(self.config['postgresql'], sighup)
+            self.dcs.reload_config(self.config)
         except Exception:
             logger.exception('Failed to reload config_file=%s', self.config.config_file)
 
@@ -76,14 +85,6 @@ class Patroni(object):
     def replicatefrom(self):
         return self.tags.get('replicatefrom')
 
-    def sighup_handler(self, *args):
-        self._received_sighup = True
-
-    def sigterm_handler(self, *args):
-        if not self._received_sigterm:
-            self._received_sigterm = True
-            sys.exit()
-
     @property
     def noloadbalance(self):
         return bool(self.tags.get('noloadbalance', False))
@@ -104,80 +105,71 @@ class Patroni(object):
     def run(self):
         self.api.start()
         self.next_run = time.time()
+        super(Patroni, self).run()
 
-        while not self._received_sigterm:
-            if self._received_sighup:
-                self._received_sighup = False
-                if self.config.reload_local_configuration():
-                    self.reload_config()
+    def _run_cycle(self):
+        logger.info(self.ha.run_cycle())
 
-            logger.info(self.ha.run_cycle())
+        if self.dcs.cluster and self.dcs.cluster.config and self.dcs.cluster.config.data \
+                and self.config.set_dynamic_configuration(self.dcs.cluster.config):
+            self.reload_config()
 
-            if self.dcs.cluster and self.dcs.cluster.config and self.dcs.cluster.config.data \
-                    and self.config.set_dynamic_configuration(self.dcs.cluster.config):
-                self.reload_config()
+        if self.postgresql.role != 'uninitialized':
+            self.config.save_cache()
 
-            if self.postgresql.role != 'uninitialized':
-                self.config.save_cache()
+        self.schedule_next_run()
 
-            self.schedule_next_run()
-
-    def setup_signal_handlers(self):
-        self._received_sighup = False
-        self._received_sigterm = False
-        signal.signal(signal.SIGHUP, self.sighup_handler)
-        signal.signal(signal.SIGTERM, self.sigterm_handler)
-
-    def shutdown(self):
+    def _shutdown(self):
         try:
             self.api.shutdown()
         except Exception:
             logger.exception('Exception during RestApi.shutdown')
-        self.ha.shutdown()
+        try:
+            self.ha.shutdown()
+        except Exception:
+            logger.exception('Exception during Ha.shutdown')
 
 
 def patroni_main():
-    logformat = os.environ.get('PATRONI_LOGFORMAT', '%(asctime)s %(levelname)s: %(message)s')
-    loglevel = os.environ.get('PATRONI_LOGLEVEL', 'INFO')
-    requests_loglevel = os.environ.get('PATRONI_REQUESTS_LOGLEVEL', 'WARNING')
-    logging.basicConfig(format=logformat, level=loglevel)
-    logging.getLogger('requests').setLevel(requests_loglevel)
+    from multiprocessing import freeze_support
+    from patroni.validator import schema
+
+    freeze_support()
+    abstract_main(Patroni, schema)
+
+
+def fatal(string, *args):
+    sys.stderr.write('FATAL: ' + string.format(*args) + '\n')
+    sys.exit(1)
+
+
+def check_psycopg2():
+    min_psycopg2 = (2, 5, 4)
+    min_psycopg2_str = '.'.join(map(str, min_psycopg2))
+
+    def parse_version(version):
+        for e in version.split('.'):
+            try:
+                yield int(e)
+            except ValueError:
+                break
 
-    patroni = Patroni()
     try:
-        patroni.run()
-    except KeyboardInterrupt:
-        pass
-    finally:
-        patroni.shutdown()
-
-
-def pg_ctl_start(args):
-    import subprocess
-    postmaster = subprocess.Popen(args)
-    print(postmaster.pid)
-
-
-def call_self(args, **kwargs):
-    """This function executes Patroni once again with provided arguments.
-
-    :args: list of arguments to call Patroni with.
-    :returns: `Popen` object"""
-
-    exe = [sys.executable]
-    if not getattr(sys, 'frozen', False):  # Binary distribution?
-        exe.append(sys.argv[0])
-
-    import subprocess
-    return subprocess.Popen(exe + args, **kwargs)
+        import psycopg2
+        version_str = psycopg2.__version__.split(' ')[0]
+        version = tuple(parse_version(version_str))
+        if version < min_psycopg2:
+            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():
     if os.getpid() != 1:
-        if len(sys.argv) > 5 and sys.argv[1] == 'pg_ctl_start':
-            return pg_ctl_start(sys.argv[2:])
+        check_psycopg2()
         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
@@ -188,7 +180,7 @@ 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
 
@@ -196,14 +188,18 @@ def main():
         if pid:
             os.kill(pid, signo)
 
-    signal.signal(signal.SIGCHLD, sigchld_handler)
-    signal.signal(signal.SIGHUP, passtochild)
+    if os.name != 'nt':
+        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.SIGUSR2, passtochild)
-    signal.signal(signal.SIGQUIT, passtochild)
+    signal.signal(signal.SIGABRT, 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

@@ -1,38 +1,37 @@
 import base64
-import fcntl
+import hmac
 import json
 import logging
 import psycopg2
 import time
+import traceback
 import dateutil.parser
 import datetime
+import os
+import six
+import socket
+import sys
 
-from patroni.postgresql import PostgresConnectionException, PostgresException, Postgresql
-from patroni.utils import deep_compare, parse_bool, patch_config, Retry, \
-    RetryFailedError, parse_int, split_host_port, tzutc
 from six.moves.BaseHTTPServer import BaseHTTPRequestHandler, HTTPServer
 from six.moves.socketserver import ThreadingMixIn
+from six.moves.urllib_parse import urlparse, parse_qs
 from threading import Thread
 
+from .exceptions import PostgresConnectionException, PostgresException
+from .postgresql.misc import postgres_version_to_int
+from .utils import deep_compare, enable_keepalive, parse_bool, patch_config, Retry, \
+    RetryFailedError, parse_int, split_host_port, tzutc, uri, cluster_as_json
+
 logger = logging.getLogger(__name__)
 
 
-def check_auth(func):
-    """Decorator function to check authorization header.
-
-    Usage example:
-    @check_auth
-    def do_PUT_foo():
-        pass
-    """
-    def wrapper(handler, *args, **kwargs):
-        if handler.check_auth_header():
-            return func(handler, *args, **kwargs)
-    return wrapper
-
-
 class RestApiHandler(BaseHTTPRequestHandler):
 
+    def _write_status_code_only(self, status_code):
+        message = self.responses[status_code][0]
+        self.wfile.write('{0} {1} {2}\r\n\r\n'.format(self.protocol_version, status_code, message).encode('utf-8'))
+        self.log_request(status_code)
+
     def _write_response(self, status_code, body, content_type='text/html', headers=None):
         self.send_response(status_code)
         headers = headers or {}
@@ -40,20 +39,28 @@ class RestApiHandler(BaseHTTPRequestHandler):
             headers['Content-Type'] = content_type
         for name, value in headers.items():
             self.send_header(name, value)
+        for name, value in self.server.http_extra_headers.items():
+            self.send_header(name, value)
         self.end_headers()
         self.wfile.write(body.encode('utf-8'))
 
     def _write_json_response(self, status_code, response):
         self._write_response(status_code, json.dumps(response), content_type='application/json')
 
-    def send_auth_request(self, body):
-        headers = {'WWW-Authenticate': 'Basic realm="' + self.server.patroni.__class__.__name__ + '"'}
-        self._write_response(401, body, headers=headers)
+    def check_auth(func):
+        """Decorator function to check authorization header or client certificates
 
-    def check_auth_header(self):
-        auth_header = self.headers.get('Authorization')
-        status = self.server.check_auth_header(auth_header)
-        return not status or self.send_auth_request(status)
+        Usage example:
+        @check_auth
+        def do_PUT_foo():
+            pass
+        """
+
+        def wrapper(self, *args, **kwargs):
+            if self.server.check_auth(self):
+                return func(self, *args, **kwargs)
+
+        return wrapper
 
     def _write_status_response(self, status_code, response):
         patroni = self.server.patroni
@@ -73,6 +80,12 @@ class RestApiHandler(BaseHTTPRequestHandler):
             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):
@@ -84,49 +97,82 @@ class RestApiHandler(BaseHTTPRequestHandler):
         patroni = self.server.patroni
         cluster = patroni.dcs.cluster
 
-        def is_synchronous():
-            return (cluster.is_synchronous_mode() and cluster.sync
-                    and cluster.sync.sync_standby == patroni.postgresql.name)
+        leader_optime = cluster and cluster.last_leader_operation or 0
+        replayed_location = response.get('xlog', {}).get('replayed_location', 0)
+        max_replica_lag = parse_int(self.path_query.get('lag', [sys.maxsize])[0], 'B')
+        if max_replica_lag is None:
+            max_replica_lag = sys.maxsize
+        is_lagging = leader_optime and leader_optime > replayed_location + max_replica_lag
 
-        def is_balanceable_replica():
-            return response.get('role') == 'replica' and not patroni.noloadbalance
+        replica_status_code = 200 if not patroni.noloadbalance and not is_lagging and \
+            response.get('role') == 'replica' and response.get('state') == 'running' else 503
 
-        if cluster:  # dcs available
-            if patroni.ha.is_leader():
-                status_code = 200 if 'master' in path 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 path in ('/sync', '/synchronous'):
-                status_code = 200 if is_balanceable_replica() and is_synchronous() else 503
-            elif path in ('/async', '/asynchronous'):
-                status_code = 200 if is_balanceable_replica() and not is_synchronous() else 503
-            elif response['role'] in path:  # response['role'] != 'master'
-                status_code = 503 if patroni.noloadbalance else 200
+        if not cluster and patroni.ha.is_paused():
+            primary_status_code = 200 if response.get('role') == 'master' else 503
+            standby_leader_status_code = 200 if response.get('role') == 'standby_leader' else 503
+        elif patroni.ha.is_leader():
+            if patroni.ha.is_standby_cluster():
+                primary_status_code = replica_status_code = 503
+                standby_leader_status_code = 200 if response.get('role') in ('replica', 'standby_leader') else 503
             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
+                primary_status_code = 200
+                standby_leader_status_code = 503
         else:
-            status_code = 503
+            primary_status_code = standby_leader_status_code = 503
+
+        status_code = 503
+
+        if 'standby_leader' in path or 'standby-leader' in path:
+            status_code = standby_leader_status_code
+        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 200 in (primary_status_code, standby_leader_status_code) 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 patroni.postgresql.name in cluster.sync.members
+            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]
-            self.wfile.write('{0} {1} {2}\r\n'.format(self.protocol_version, status_code, message).encode('utf-8'))
+            self._write_status_code_only(status_code)
         else:
             self._write_status_response(status_code, response)
 
     def do_OPTIONS(self):
         self.do_GET(write_status_code_only=True)
 
+    def do_GET_liveness(self):
+        self._write_status_code_only(200)
+
+    def do_GET_readiness(self):
+        patroni = self.server.patroni
+        if patroni.ha.is_leader():
+            status_code = 200
+        elif patroni.postgresql.state == 'running':
+            status_code = 200 if patroni.dcs.cluster else 503
+        else:
+            status_code = 503
+        self._write_status_code_only(status_code)
+
     def do_GET_patroni(self):
         response = self.get_postgresql_status(True)
         self._write_status_response(200, response)
 
+    def do_GET_cluster(self):
+        cluster = self.server.patroni.dcs.get_cluster(True)
+        self._write_json_response(200, cluster_as_json(cluster))
+
+    def do_GET_history(self):
+        cluster = self.server.patroni.dcs.cluster or self.server.patroni.dcs.get_cluster()
+        self._write_json_response(200, cluster.history and cluster.history.lines or [])
+
     def do_GET_config(self):
         cluster = self.server.patroni.dcs.cluster or self.server.patroni.dcs.get_cluster()
         if cluster.config:
@@ -152,7 +198,9 @@ class RestApiHandler(BaseHTTPRequestHandler):
     def do_PATCH_config(self):
         request = self._read_json_content()
         if request:
-            cluster = self.server.patroni.dcs.get_cluster()
+            cluster = self.server.patroni.dcs.get_cluster(True)
+            if not (cluster.config and cluster.config.modify_index):
+                return self.send_error(503)
             data = cluster.config.data.copy()
             if patch_config(data, request):
                 value = json.dumps(data, separators=(',', ':'))
@@ -174,18 +222,8 @@ class RestApiHandler(BaseHTTPRequestHandler):
 
     @check_auth
     def do_POST_reload(self):
-        try:
-            if self.server.patroni.config.reload_local_configuration(True):
-                status_code = 202
-                response = 'reload scheduled'
-                self.server.patroni.sighup_handler()
-            else:
-                status_code = 200
-                response = 'nothing changed'
-        except Exception as e:
-            status_code = 500
-            response = str(e)
-        self._write_response(status_code, response)
+        self.server.patroni.sighup_handler()
+        self._write_response(202, 'reload scheduled')
 
     @staticmethod
     def parse_schedule(schedule, action):
@@ -237,7 +275,7 @@ class RestApiHandler(BaseHTTPRequestHandler):
                     break
             elif k == 'postgres_version':
                 try:
-                    Postgresql.postgres_version_to_int(request[k])
+                    postgres_version_to_int(request[k])
                 except PostgresException as e:
                     status_code = 400
                     data = e.value
@@ -279,6 +317,20 @@ class RestApiHandler(BaseHTTPRequestHandler):
             code = 404
         self._write_response(code, data)
 
+    @check_auth
+    def do_DELETE_switchover(self):
+        failover = self.server.patroni.dcs.get_cluster().failover
+        if failover and failover.scheduled_at:
+            if not self.server.patroni.dcs.manual_failover('', '', index=failover.index):
+                return self.send_error(409)
+            else:
+                data = "scheduled switchover deleted"
+                code = 200
+        else:
+            data = "no switchover is scheduled"
+            code = 404
+        self._write_response(code, data)
+
     @check_auth
     def do_POST_reinitialize(self):
         request = self._read_json_content(body_is_optional=True)
@@ -318,13 +370,13 @@ class RestApiHandler(BaseHTTPRequestHandler):
         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:
+            if action == 'switchover' and cluster.is_synchronous_mode() and candidate not in cluster.sync.members:
                 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]
+            members = [m for m in cluster.members if m.name in cluster.sync.members]
             if not members:
                 return action + ' is not possible: can not find sync_standby'
         else:
@@ -401,6 +453,9 @@ class RestApiHandler(BaseHTTPRequestHandler):
 
         ret = BaseHTTPRequestHandler.parse_request(self)
         if ret:
+            urlpath = urlparse(self.path)
+            self.path = urlpath.path
+            self.path_query = parse_qs(urlpath.query) or {}
             mname = self.path.lstrip('/').split('/')[0]
             mname = self.command + ('_' + mname if mname else '')
             if hasattr(self, 'do_' + mname):
@@ -414,69 +469,75 @@ class RestApiHandler(BaseHTTPRequestHandler):
         return retry(self.server.query, sql, *params)
 
     def get_postgresql_status(self, retry=False):
+        postgresql = self.server.patroni.postgresql
         try:
-            if self.server.patroni.postgresql.state not in ('running', 'restarting', 'starting'):
-                raise RetryFailedError('')
-            stmt = ("WITH replication_info AS ("
-                    "SELECT usename, application_name, client_addr, state, sync_state, sync_priority"
-                    " FROM pg_stat_replication) SELECT"
-                    " to_char(pg_postmaster_start_time(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),"
-                    " CASE WHEN pg_is_in_recovery() THEN 0"
-                    " ELSE ('x' || SUBSTR(pg_{0}file_name(pg_current_{0}_{1}()), 1, 8))::bit(32)::int END,"
-                    " CASE WHEN pg_is_in_recovery() THEN 0"
-                    " ELSE pg_{0}_{1}_diff(pg_current_{0}_{1}(), '0/0')::bigint END,"
-                    " pg_{0}_{1}_diff(COALESCE(pg_last_{0}_receive_{1}(), pg_last_{0}_replay_{1}()), '0/0')::bigint,"
-                    " pg_{0}_{1}_diff(pg_last_{0}_replay_{1}(), '0/0')::bigint,"
-                    " to_char(pg_last_xact_replay_timestamp(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),"
-                    " pg_is_in_recovery() AND pg_is_{0}_replay_paused(),"
-                    " (SELECT array_to_json(array_agg(row_to_json(ri))) FROM replication_info ri)")
+            cluster = self.server.patroni.dcs.cluster
 
-            row = self.query(stmt.format(self.server.patroni.postgresql.wal_name,
-                                         self.server.patroni.postgresql.lsn_name), retry=retry)[0]
+            if postgresql.state not in ('running', 'restarting', 'starting'):
+                raise RetryFailedError('')
+            stmt = ("SELECT " + postgresql.POSTMASTER_START_TIME + ", " + postgresql.TL_LSN + ","
+                    " pg_catalog.to_char(pg_catalog.pg_last_xact_replay_timestamp(), 'YYYY-MM-DD HH24:MI:SS.MS TZ'),"
+                    " pg_catalog.array_to_json(pg_catalog.array_agg(pg_catalog.row_to_json(ri))) "
+                    "FROM (SELECT (SELECT rolname FROM pg_authid WHERE oid = usesysid) AS usename,"
+                    " application_name, client_addr, w.state, sync_state, sync_priority"
+                    " FROM pg_catalog.pg_stat_get_wal_senders() w, pg_catalog.pg_stat_get_activity(pid)) AS ri")
+
+            row = self.query(stmt.format(postgresql.wal_name, postgresql.lsn_name), retry=retry)[0]
 
             result = {
-                'state': self.server.patroni.postgresql.state,
+                'state': postgresql.state,
                 'postmaster_start_time': row[0],
                 'role': 'replica' if row[1] == 0 else 'master',
-                'server_version': self.server.patroni.postgresql.server_version,
+                'server_version': 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] == 0 else {
+                    'received_location': row[4] or row[3],
+                    'replayed_location': row[3],
+                    'replayed_timestamp': row[6],
+                    'paused': row[5]} if row[1] == 0 else {
                     'location': row[2]
                 })
             }
 
+            if result['role'] == 'replica' and self.server.patroni.ha.is_standby_cluster():
+                result['role'] = postgresql.role
+
             if row[1] > 0:
                 result['timeline'] = row[1]
             else:
-                cluster = self.server.patroni.dcs.cluster
                 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)
+                result['timeline'] = postgresql.replica_cached_timeline(leader_timeline)
 
             if row[7]:
                 result['replication'] = row[7]
 
             return result
         except (psycopg2.Error, RetryFailedError, PostgresConnectionException):
-            state = self.server.patroni.postgresql.state
+            state = postgresql.state
             if state == 'running':
                 logger.exception('get_postgresql_status')
                 state = 'unknown'
-            return {'state': state, 'role': self.server.patroni.postgresql.role}
+            return {'state': state, 'role': postgresql.role}
+
+    def handle_one_request(self):
+        self.__start_time = time.time()
+        BaseHTTPRequestHandler.handle_one_request(self)
 
     def log_message(self, fmt, *args):
-        logger.debug("API thread: %s - - [%s] %s", self.client_address[0], self.log_date_time_string(), fmt % args)
+        latency = 1000.0 * (time.time() - self.__start_time)
+        logger.debug("API thread: %s - - %s latency: %0.3f ms", self.client_address[0], fmt % args, latency)
 
 
 class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
+    # On 3.7+ the `ThreadingMixIn` gathers all non-daemon worker threads in order to join on them at server close.
+    daemon_threads = True  # Make worker threads "fire and forget" to prevent a memory leak.
 
     def __init__(self, patroni, config):
         self.patroni = patroni
         self.__listen = None
-        self.__initialize(config)
-        self.__set_config_parameters(config)
+        self.__ssl_options = None
+        self.http_extra_headers = {}
+        self.reload_config(config)
         self.daemon = True
 
     def query(self, sql, *params):
@@ -492,11 +553,13 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
 
     @staticmethod
     def _set_fd_cloexec(fd):
-        flags = fcntl.fcntl(fd, fcntl.F_GETFD)
-        fcntl.fcntl(fd, fcntl.F_SETFD, flags | fcntl.FD_CLOEXEC)
+        if os.name != 'nt':
+            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
+        return hmac.compare_digest(self.__auth_key, key.encode('utf-8'))
 
     def check_auth_header(self, auth_header):
         if self.__auth_key:
@@ -505,47 +568,132 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
             if not auth_header.startswith('Basic ') or not self.check_basic_auth_key(auth_header[6:]):
                 return 'not authenticated'
 
+    def check_auth(self, rh):
+        if not hasattr(rh.request, 'getpeercert') or not rh.request.getpeercert():  # valid client cert isn't present
+            if self.__protocol == 'https' and self.__ssl_options.get('verify_client') in ('required', 'optional'):
+                return rh._write_response(403, 'client certificate required')
+
+        reason = self.check_auth_header(rh.headers.get('Authorization'))
+        if reason:
+            headers = {'WWW-Authenticate': 'Basic realm="' + self.patroni.__class__.__name__ + '"'}
+            return rh._write_response(401, reason, headers=headers)
+        return True
+
     @staticmethod
-    def __get_ssl_options(config):
-        return {option: config[option] for option in ['certfile', 'keyfile'] if option in config}
+    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 __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.connection_string = '{0}://{1}/patroni'.format(self.__protocol,
-                                                            config.get('connect_address') or self.__listen)
+    def __httpserver_init(self, host, port):
+        dual_stack = self.__has_dual_stack()
+        if host in ('', '*'):
+            host = None
 
-    def __initialize(self, config):
+        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]
         try:
-            host, port = split_host_port(config['listen'], None)
+            HTTPServer.__init__(self, info[0][-1][:2], RestApiHandler)
+        except socket.error:
+            logger.error(
+                    "Couldn't start a service on '%s:%s', please check your `restapi.listen` configuration", host, port)
+            raise
+
+    def __initialize(self, listen, ssl_options):
+        try:
+            host, port = split_host_port(listen, None)
         except Exception:
             raise ValueError('Invalid "restapi" config: expected <HOST>:<PORT> for "listen", but got "{0}"'
-                             .format(config['listen']))
+                             .format(listen))
 
-        if self.__listen is not None:  # changing config in runtime
+        reloading_config = self.__listen is not None  # changing config in runtime
+        if reloading_config:
             self.shutdown()
 
-        self.__listen = config['listen']
-        self.__ssl_options = self.__get_ssl_options(config)
+        self.__listen = listen
+        self.__ssl_options = ssl_options
 
-        HTTPServer.__init__(self, (host, port), RestApiHandler)
+        self.__httpserver_init(host, port)
         Thread.__init__(self, target=self.serve_forever)
         self._set_fd_cloexec(self.socket)
 
-        self.__protocol = 'http'
-
         # wrap socket with ssl if 'certfile' is defined in a config.yaml
         # Sometime it's also needed to pass reference to a 'keyfile'.
-        if self.__ssl_options.get('certfile'):
+        self.__protocol = 'https' if ssl_options.get('certfile') else 'http'
+        if self.__protocol == 'https':
             import ssl
-            self.socket = ssl.wrap_socket(self.socket, server_side=True, **self.__ssl_options)
-            self.__protocol = 'https'
-        return True
+            ctx = ssl.create_default_context(ssl.Purpose.CLIENT_AUTH, cafile=ssl_options.get('cafile'))
+            if ssl_options.get('ciphers'):
+                ctx.set_ciphers(ssl_options['ciphers'])
+            ctx.load_cert_chain(certfile=ssl_options['certfile'], keyfile=ssl_options.get('keyfile'),
+                                password=ssl_options.get('keyfile_password'))
+            verify_client = ssl_options.get('verify_client')
+            if verify_client:
+                modes = {'none': ssl.CERT_NONE, 'optional': ssl.CERT_OPTIONAL, 'required': ssl.CERT_REQUIRED}
+                if verify_client in modes:
+                    ctx.verify_mode = modes[verify_client]
+                else:
+                    logger.error('Bad value in the "restapi.verify_client": %s', verify_client)
+            self.socket = ctx.wrap_socket(self.socket, server_side=True)
+        if reloading_config:
+            self.start()
+
+    def process_request_thread(self, request, client_address):
+        if isinstance(request, tuple):
+            sock, newsock = request
+            try:
+                request = sock.context.wrap_socket(newsock, do_handshake_on_connect=sock.do_handshake_on_connect,
+                                                   suppress_ragged_eofs=sock.suppress_ragged_eofs, server_side=True)
+            except socket.error:
+                return
+        super(RestApiServer, self).process_request_thread(request, client_address)
+
+    def get_request(self):
+        sock = self.socket
+        newsock, addr = socket.socket.accept(sock)
+        enable_keepalive(newsock, 10, 3)
+        if hasattr(sock, 'context'):  # SSLSocket, we want to do the deferred handshake from a thread
+            newsock = (sock, newsock)
+        return newsock, addr
+
+    def shutdown_request(self, request):
+        if isinstance(request, tuple):
+            _, request = request  # SSLSocket
+        return super(RestApiServer, self).shutdown_request(request)
 
     def reload_config(self, config):
         if 'listen' not in config:  # changing config in runtime
             raise ValueError('Can not find "restapi.listen" 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)
+        ssl_options = {n: config[n] for n in ('certfile', 'keyfile', 'keyfile_password',
+                                              'cafile', 'ciphers') if n in config}
+
+        self.http_extra_headers = config.get('http_extra_headers') or {}
+        self.http_extra_headers.update((config.get('https_extra_headers') or {}) if ssl_options.get('certfile') else {})
+
+        if isinstance(config.get('verify_client'), six.string_types):
+            ssl_options['verify_client'] = config['verify_client'].lower()
+
+        if self.__listen != config['listen'] or self.__ssl_options != ssl_options:
+            self.__initialize(config['listen'], ssl_options)
+
+        self.__auth_key = base64.b64encode(config['auth'].encode('utf-8')) if 'auth' in config else None
+        self.connection_string = uri(self.__protocol, config.get('connect_address') or self.__listen, 'patroni')
+
+    @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

@@ -52,8 +52,8 @@ class CriticalTask(object):
 
 class AsyncExecutor(object):
 
-    def __init__(self, state_handler, ha_wakeup):
-        self.state_handler = state_handler
+    def __init__(self, cancellable, ha_wakeup):
+        self._cancellable = cancellable
         self._ha_wakeup = ha_wakeup
         self._thread_lock = RLock()
         self._scheduled_action = None
@@ -92,7 +92,7 @@ class AsyncExecutor(object):
                     return
                 self._finish_event.clear()
 
-            self.state_handler.reset_is_cancelled()
+            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
@@ -110,6 +110,12 @@ class AsyncExecutor(object):
     def run_async(self, func, args=()):
         Thread(target=self.run, args=(func, args)).start()
 
+    def try_run_async(self, action, func, args=()):
+        prev = self.schedule(action)
+        if prev is None:
+            return self.run_async(func, args)
+        return 'Failed to run {0}, {1} is already in progress'.format(action, prev)
+
     def cancel(self):
         with self:
             with self._scheduled_action_lock:
@@ -118,7 +124,7 @@ class AsyncExecutor(object):
                 logger.warning('Cancelling long running task %s', self._scheduled_action)
             self._is_cancelled = True
 
-        self.state_handler.cancel()
+        self._cancellable.cancel()
         self._finish_event.wait()
 
         with self:

patroni/callback_executor.py

@@ -1,37 +0,0 @@
-import logging
-import subprocess
-from threading import Event, Lock, Thread
-
-logger = logging.getLogger(__name__)
-
-
-class CallbackExecutor(Thread):
-
-    def __init__(self):
-        super(CallbackExecutor, self).__init__()
-        self.daemon = True
-        self._lock = Lock()
-        self._cmd = None
-        self._process = None
-        self._callback_event = Event()
-        self.start()
-
-    def call(self, cmd):
-        with self._lock:
-            if self._process and self._process.poll() is None:
-                self._process.kill()
-                logger.warning('Killed the old callback process because it was still running: %s', self._cmd)
-        self._cmd = cmd
-        self._callback_event.set()
-
-    def run(self):
-        while True:
-            self._callback_event.wait()
-            self._callback_event.clear()
-            with self._lock:
-                try:
-                    self._process = subprocess.Popen(self._cmd, close_fds=True)
-                except Exception:
-                    logger.exception('Failed to execute %s',  self._cmd)
-                    continue
-            self._process.wait()

patroni/config.py

@@ -1,19 +1,38 @@
 import json
 import logging
 import os
-import sys
+import shutil
 import tempfile
 import yaml
 
 from collections import defaultdict
 from copy import deepcopy
+from patroni import PATRONI_ENV_PREFIX
+from patroni.exceptions import ConfigParseError
 from patroni.dcs import ClusterConfig
-from patroni.postgresql import Postgresql
+from patroni.postgresql.config import CaseInsensitiveDict, ConfigHandler
 from patroni.utils import deep_compare, parse_bool, parse_int, patch_config
-from requests.structures import CaseInsensitiveDict
 
 logger = logging.getLogger(__name__)
 
+_AUTH_ALLOWED_PARAMETERS = (
+    'username',
+    'password',
+    'sslmode',
+    'sslcert',
+    'sslkey',
+    'sslpassword',
+    'sslrootcert',
+    'sslcrl',
+    'gssencmode',
+    'channel_binding'
+)
+
+
+def default_validator(conf):
+    if not conf:
+        return "Config is empty."
+
 
 class Config(object):
     """
@@ -35,47 +54,59 @@ class Config(object):
          to work with it as with the old `config` object.
     """
 
-    PATRONI_ENV_PREFIX = 'PATRONI_'
     PATRONI_CONFIG_VARIABLE = PATRONI_ENV_PREFIX + 'CONFIGURATION'
 
     __CACHE_FILENAME = 'patroni.dynamic.json'
     __DEFAULT_CONFIG = {
         'ttl': 30, 'loop_wait': 10, 'retry_timeout': 10,
         'maximum_lag_on_failover': 1048576,
+        'maximum_lag_on_syncnode': -1,
+        'check_timeline': False,
         'master_start_timeout': 300,
+        'master_stop_timeout': 0,
         'synchronous_mode': False,
         'synchronous_mode_strict': False,
+        'synchronous_node_count': 1,
+        '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': CaseInsensitiveDict({p: v[0] for p, v in Postgresql.CMDLINE_OPTIONS.items()})
+            'parameters': CaseInsensitiveDict({p: v[0] for p, v in ConfigHandler.CMDLINE_OPTIONS.items()
+                                               if p not in ('wal_keep_segments', 'wal_keep_size')})
         },
         'watchdog': {
             'mode': 'automatic',
         }
     }
 
-    def __init__(self):
+    def __init__(self, configfile, validator=default_validator):
         self._modify_index = -1
         self._dynamic_configuration = {}
 
         self.__environment_configuration = self._build_environment_configuration()
 
         # Patroni reads the configuration from the command-line argument if it exists, otherwise from the environment
-        self._config_file = len(sys.argv) >= 2 and os.path.isfile(sys.argv[1]) and sys.argv[1]
+        self._config_file = configfile and os.path.exists(configfile) and configfile
         if self._config_file:
             self._local_configuration = self._load_config_file()
         else:
             config_env = os.environ.pop(self.PATRONI_CONFIG_VARIABLE, None)
             self._local_configuration = config_env and yaml.safe_load(config_env) or self.__environment_configuration
-            if not self._local_configuration:
-                print('Usage: {0} config.yml'.format(sys.argv[0]))
-                print('\tPatroni may also read the configuration from the {0} environment variable'.
-                      format(self.PATRONI_CONFIG_VARIABLE))
-                sys.exit(1)
+        if validator:
+            error = validator(self._local_configuration)
+            if error:
+                raise ConfigParseError(error)
 
         self.__effective_configuration = self._build_effective_configuration({}, self._local_configuration)
-        self._data_dir = self.__effective_configuration['postgresql']['data_dir']
+        self._data_dir = self.__effective_configuration.get('postgresql', {}).get('data_dir', "")
         self._cache_file = os.path.join(self._data_dir, self.__CACHE_FILENAME)
         self._load_cache()
         self._cache_needs_saving = False
@@ -91,12 +122,32 @@ class Config(object):
     def check_mode(self, mode):
         return bool(parse_bool(self._dynamic_configuration.get(mode)))
 
+    def _load_config_path(self, path):
+        """
+        If path is a file, loads the yml file pointed to by path.
+        If path is a directory, loads all yml files in that directory in alphabetical order
+        """
+        if os.path.isfile(path):
+            files = [path]
+        elif os.path.isdir(path):
+            files = [os.path.join(path, f) for f in sorted(os.listdir(path))
+                     if (f.endswith('.yml') or f.endswith('.yaml')) and os.path.isfile(os.path.join(path, f))]
+        else:
+            logger.error('config path %s is neither directory nor file', path)
+            raise ConfigParseError('invalid config path')
+
+        overall_config = {}
+        for fname in files:
+            with open(fname) as f:
+                config = yaml.safe_load(f)
+                patch_config(overall_config, config)
+        return overall_config
+
     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:
-            config = yaml.safe_load(f)
-            patch_config(config, self.__environment_configuration)
-            return config
+        config = self._load_config_path(self._config_file)
+        patch_config(config, self.__environment_configuration)
+        return config
 
     def _load_cache(self):
         if os.path.isfile(self._cache_file):
@@ -114,7 +165,7 @@ 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)
@@ -147,31 +198,25 @@ class Config(object):
             except Exception:
                 logger.exception('Exception when setting dynamic_configuration')
 
-    def reload_local_configuration(self, dry_run=False):
+    def reload_local_configuration(self):
         if self.config_file:
             try:
                 configuration = self._load_config_file()
                 if not deep_compare(self._local_configuration, configuration):
                     new_configuration = self._build_effective_configuration(self._dynamic_configuration, configuration)
-                    if dry_run:
-                        return not deep_compare(new_configuration, self.__effective_configuration)
                     self._local_configuration = configuration
                     self.__effective_configuration = new_configuration
                     return True
                 else:
-                    logger.info('No configuration items changed, nothing to reload.')
+                    logger.info('No local configuration items changed.')
             except Exception:
                 logger.exception('Exception when reloading local configuration from %s', self.config_file)
-                if dry_run:
-                    raise
 
     @staticmethod
     def _process_postgresql_parameters(parameters, is_local=False):
-        ret = {}
-        for name, value in (parameters or {}).items():
-            if name not in Postgresql.CMDLINE_OPTIONS or not is_local and Postgresql.CMDLINE_OPTIONS[name][1](value):
-                ret[name] = value
-        return ret
+        return {name: value for name, value in (parameters or {}).items()
+                if name not in ConfigHandler.CMDLINE_OPTIONS or
+                not is_local and ConfigHandler.CMDLINE_OPTIONS[name][1](value)}
 
     def _safe_copy_dynamic_configuration(self, dynamic_configuration):
         config = deepcopy(self.__DEFAULT_CONFIG)
@@ -183,6 +228,10 @@ 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 in ('synchronous_mode', 'synchronous_mode_strict'):
                     config[name] = value
@@ -195,25 +244,54 @@ class Config(object):
         ret = defaultdict(dict)
 
         def _popenv(name):
-            return os.environ.pop(Config.PATRONI_ENV_PREFIX + name.upper(), None)
+            return os.environ.pop(PATRONI_ENV_PREFIX + name.upper(), None)
 
         for param in ('name', 'namespace', 'scope'):
             value = _popenv(param)
             if value:
                 ret[param] = value
 
+        def _fix_log_env(name, oldname):
+            value = _popenv(oldname)
+            name = 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)
                 if value:
                     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('restapi', ['listen', 'connect_address', 'certfile', 'keyfile', 'keyfile_password',
+                                        'cafile', 'ciphers', 'verify_client', 'http_extra_headers',
+                                        'https_extra_headers'])
+        _set_section_values('ctl', ['insecure', 'cacert', 'certfile', 'keyfile'])
+        _set_section_values('postgresql', ['listen', 'connect_address', 'config_dir', 'data_dir', 'pgpass', 'bin_dir'])
+        _set_section_values('log', ['level', 'traceback_level', 'format', 'dateformat', 'max_queue_size',
+                                    'dir', 'file_size', 'file_num', 'loggers'])
 
-        def _get_auth(name):
+        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, params=None):
             ret = {}
-            for param in ('username', 'password'):
+            for param in params or _AUTH_ALLOWED_PARAMETERS[:2]:
                 value = _popenv(name + '_' + param)
                 if value:
                     ret[param] = value
@@ -224,16 +302,14 @@ class Config(object):
             ret['restapi']['authentication'] = restapi_auth
 
         authentication = {}
-        for user_type in ('replication', 'superuser'):
-            entry = _get_auth(user_type)
+        for user_type in ('replication', 'superuser', 'rewind'):
+            entry = _get_auth(user_type, _AUTH_ALLOWED_PARAMETERS)
             if entry:
                 authentication[user_type] = entry
 
         if authentication:
             ret['postgresql']['authentication'] = authentication
 
-        users = {}
-
         def _parse_list(value):
             if not (value.strip().startswith('-') or '[' in value):
                 value = '[{0}]'.format(value)
@@ -243,39 +319,48 @@ class Config(object):
                 logger.exception('Exception when parsing list %s', value)
                 return None
 
+        _set_section_values('raft', ['data_dir', 'self_addr', 'partner_addrs', 'password', 'bind_addr'])
+        if 'raft' in ret and 'partner_addrs' in ret['raft']:
+            ret['raft']['partner_addrs'] = _parse_list(ret['raft']['partner_addrs'])
+
         for param in list(os.environ.keys()):
-            if param.startswith(Config.PATRONI_ENV_PREFIX):
+            if param.startswith(PATRONI_ENV_PREFIX):
+                # PATRONI_(ETCD|CONSUL|ZOOKEEPER|EXHIBITOR|...)_(HOSTS?|PORT|..)
                 name, suffix = (param[8:].split('_', 1) + [''])[:2]
-                if name and suffix:
-                    # PATRONI_(ETCD|CONSUL|ZOOKEEPER|EXHIBITOR|...)_(HOSTS?|PORT|..)
-                    if suffix in ('HOST', 'HOSTS', 'PORT', 'SRV', 'URL', 'PROXY', 'CACERT', 'CERT',
-                                  'KEY', 'VERIFY', 'TOKEN', 'CHECKS', 'DC', 'NAMESPACE', 'CONTEXT',
-                                  'USE_ENDPOINTS', 'SCOPE_LABEL', 'ROLE_LABEL', 'POD_IP', 'PORTS', 'LABELS'):
-                        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':
-                            if not value.strip().startswith('{'):
-                                value = '{{{0}}}'.format(value)
-                            try:
-                                value = yaml.safe_load(value)
-                            except Exception:
-                                logger.exception('Exception when parsing dict %s', value)
-                                value = None
-                        if value:
-                            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 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',
+                              'BYPASS_API_SERVICE', 'KEY_PASSWORD', 'USE_SSL') 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', 'USE_ENDPOINTS', 'BYPASS_API_SERVICE'):
+                        value = parse_bool(value)
+                    if value:
+                        ret[name.lower()][suffix.lower()] = value
+        for dcs in ('etcd', 'etcd3'):
+            if dcs in ret:
+                ret[dcs].update(_get_auth(dcs))
+
+        users = {}
+        for param in list(os.environ.keys()):
+            if param.startswith(PATRONI_ENV_PREFIX):
+                name, suffix = (param[8:].rsplit('_', 1) + [''])[:2]
+                # PATRONI_<username>_PASSWORD=<password>, PATRONI_<username>_OPTIONS=<option1,option2,...>
+                # CREATE USER "<username>" WITH <OPTIONS> PASSWORD '<password>'
+                if name and 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
 
@@ -294,7 +379,7 @@ class Config(object):
                 config[name] = deepcopy(value) if value else {}
 
         # restapi server expects to get restapi.auth = 'username:password'
-        if 'authentication' in config['restapi']:
+        if 'restapi' in config and 'authentication' in config['restapi']:
             config['restapi']['auth'] = '{username}:{password}'.format(**config['restapi']['authentication'])
 
         # special treatment for old config
@@ -313,12 +398,26 @@ class Config(object):
         if 'superuser' not in pg_config['authentication'] and 'pg_rewind' in pg_config:
             pg_config['authentication']['superuser'] = pg_config['pg_rewind']
 
+        # handle setting additional connection parameters that may be available
+        # in the configuration file, such as SSL connection parameters
+        for name, value in pg_config['authentication'].items():
+            pg_config['authentication'][name] = {n: v for n, v in value.items() if n in _AUTH_ALLOWED_PARAMETERS}
+
         # no 'name' in config
         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',
-                          'synchronous_mode', 'maximum_lag_on_failover') if p in config})
+        updated_fields = (
+            'name',
+            'scope',
+            'retry_timeout',
+            'synchronous_mode',
+            'synchronous_mode_strict',
+            'synchronous_node_count',
+            'maximum_lag_on_syncnode'
+        )
+
+        pg_config.update({p: config[p] for p in updated_fields if p in config})
 
         return config
 

patroni/ctl.py

@@ -2,39 +2,43 @@
 Patroni Control
 '''
 
-import base64
 import click
 import codecs
 import datetime
 import dateutil.parser
-import cdiff
+import dateutil.tz
 import copy
 import difflib
 import io
 import json
 import logging
 import os
-import psycopg2
 import random
-import requests
+import six
 import subprocess
 import sys
 import tempfile
 import time
-import tzlocal
 import yaml
 
 from click import ClickException
+from collections import defaultdict
 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 patch_config, polling_loop
-from patroni.version import __version__
-from prettytable import PrettyTable
+from prettytable import ALL, FRAME, PrettyTable
 from six.moves.urllib_parse import urlparse
-from six import text_type
+
+try:
+    from ydiff import markup_to_pager, PatchStream
+except ImportError:  # pragma: no cover
+    from cdiff import markup_to_pager, PatchStream
+
+from .dcs import get_dcs as _get_dcs
+from .exceptions import PatroniException
+from .postgresql import Postgresql
+from .postgresql.misc import postgres_version_to_int
+from .utils import cluster_as_json, find_executable, patch_config, polling_loop
+from .request import PatroniRequest
+from .version import __version__
 
 CONFIG_DIR_PATH = click.get_app_dir('patroni')
 CONFIG_FILE_PATH = os.path.join(CONFIG_DIR_PATH, 'patronictl.yaml')
@@ -48,14 +52,42 @@ class PatroniCtlException(ClickException):
     pass
 
 
+class PatronictlPrettyTable(PrettyTable):
+
+    def __init__(self, header, *args, **kwargs):
+        PrettyTable.__init__(self, *args, **kwargs)
+        self.__table_header = header
+        self.__hline_num = 0
+        self.__hline = None
+
+    def _is_first_hline(self):
+        return self.__hline_num == 0
+
+    def _set_hline(self, value):
+        self.__hline = value
+
+    def _get_hline(self):
+        ret = self.__hline
+
+        # Inject nice table header
+        if self._is_first_hline() and self.__table_header:
+            header = self.__table_header[:len(ret) - 2]
+            ret = "".join([ret[0], header, ret[1 + len(header):]])
+
+        self.__hline_num += 1
+        return ret
+
+    _hrule = property(_get_hline, _set_hline)
+
+
 def parse_dcs(dcs):
     if dcs is None:
         return None
+    elif '//' not in dcs:
+        dcs = '//' + dcs
 
     parsed = urlparse(dcs)
     scheme = parsed.scheme
-    if scheme == '' and parsed.netloc == '':
-        parsed = urlparse('//' + dcs)
     port = int(parsed.port) if parsed.port else None
 
     if scheme == '':
@@ -63,23 +95,22 @@ def parse_dcs(dcs):
     elif scheme not in DCS_DEFAULTS:
         raise PatroniCtlException('Unknown dcs scheme: {}'.format(scheme))
 
-    dcs_info = DCS_DEFAULTS[scheme]
-    return yaml.load(dcs_info['template'].format(host=parsed.hostname or 'localhost', port=port or dcs_info['port']))
+    default = DCS_DEFAULTS[scheme]
+    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)
-    config = {}
-    old_argv = list(sys.argv)
-    try:
-        sys.argv[1] = path
-        if Config.PATRONI_CONFIG_VARIABLE not in os.environ:
-            for p in ('PATRONI_RESTAPI_LISTEN', 'PATRONI_POSTGRESQL_DATA_DIR'):
-                if p not in os.environ:
-                    os.environ[p] = '.'
-        config = Config().copy()
-    finally:
-        sys.argv = old_argv
+    from patroni.config import Config
+
+    if not (os.path.exists(path) and os.access(path, os.R_OK)):
+        if path != CONFIG_FILE_PATH:    # bail if non-default config location specified but file not found / readable
+            raise PatroniCtlException('Provided config file {0} not existing or no read rights.'
+                                      ' Check the -c/--config-file parameter'.format(path))
+        else:
+            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 = Config(path, validator=None).copy()
 
     dcs = parse_dcs(dcs) or parse_dcs(config.get('dcs_api')) or {}
     if dcs:
@@ -97,21 +128,30 @@ def store_config(config, path):
         yaml.dump(config, fd)
 
 
-option_format = click.option('--format', '-f', 'fmt', help='Output format (pretty, json, yaml)', default='pretty')
+option_format = click.option('--format', '-f', 'fmt', help='Output format (pretty, tsv, 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):
-    logging.basicConfig(format='%(asctime)s - %(levelname)s - %(message)s', level=os.environ.get('LOGLEVEL', 'WARNING'))
+def ctl(ctx, config_file, dcs, insecure):
+    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):
@@ -123,49 +163,42 @@ def get_dcs(config, scope):
         raise PatroniCtlException(str(e))
 
 
-def auth_header(config):
-    if config.get('restapi', {}).get('auth', ''):
-        return {'Authorization': 'Basic ' + base64.b64encode(config['restapi']['auth'].encode('utf-8')).decode('utf-8')}
+def request_patroni(member, method='GET', endpoint=None, data=None):
+    ctx = click.get_current_context()  # the current click context
+    request_executor = ctx.obj.get('__request_patroni')
+    if not request_executor:
+        request_executor = ctx.obj['__request_patroni'] = PatroniRequest(ctx.obj)
+    return request_executor(member, method, endpoint, data)
 
 
-def request_patroni(member, request_type, endpoint, content=None, headers=None):
-    headers = headers or {}
-    url_parts = urlparse(member.api_url)
-    logging.debug(url_parts)
-    if 'Content-Type' not in headers:
-        headers['Content-Type'] = 'application/json'
+def print_output(columns, rows, alignment=None, fmt='pretty', header=None, delimiter='\t'):
+    if fmt in {'json', 'yaml', 'yml'}:
+        elements = [{k: v for k, v in zip(columns, r) if not header or str(v)} for r in rows]
+        func = json.dumps if fmt == 'json' else format_config_for_editing
+        click.echo(func(elements))
+    elif fmt in {'pretty', 'tsv', 'topology'}:
+        list_cluster = bool(header and columns and columns[0] == 'Cluster')
+        if list_cluster and 'Tags' in columns:  # we want to format member tags as YAML
+            i = columns.index('Tags')
+            for row in rows:
+                if row[i]:
+                    row[i] = format_config_for_editing(row[i], fmt != 'pretty').strip()
+        if list_cluster and fmt != 'tsv':  # skip cluster name if pretty-printing
+            columns = columns[1:] if columns else []
+            rows = [row[1:] for row in rows]
 
-    url = '{0}://{1}/{2}'.format(url_parts.scheme, url_parts.netloc, endpoint)
-
-    return getattr(requests, request_type)(url, headers=headers,
-                                           data=json.dumps(content) if content else None, timeout=60)
-
-
-def print_output(columns, rows=None, alignment=None, fmt='pretty', header=True, delimiter='\t'):
-    rows = rows or []
-    if fmt == 'pretty':
-        t = PrettyTable(columns)
-        for k, v in (alignment or {}).items():
-            t.align[k] = v
-        for r in rows:
-            t.add_row(r)
-        click.echo(t)
-        return
-
-    if fmt in ['json', 'yaml', 'yml']:
-        elements = [dict(zip(columns, r)) for r in rows]
-        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')
-
-        for r in rows:
-            c = [str(c) for c in r]
-            click.echo(delimiter.join(c))
+        if fmt == 'tsv':
+            for r in ([columns] if columns else []) + rows:
+                click.echo(delimiter.join(map(str, r)))
+        else:
+            hrules = ALL if any(any(isinstance(c, six.string_types) and '\n' in c for c in r) for r in rows) else FRAME
+            table = PatronictlPrettyTable(header, columns, hrules=hrules)
+            table.align = 'l'
+            for k, v in (alignment or {}).items():
+                table.align[k] = v
+            for r in rows:
+                table.add_row(r)
+            click.echo(table)
 
 
 def watching(w, watch, max_count=None, clear=True):
@@ -198,7 +231,7 @@ def watching(w, watch, max_count=None, clear=True):
 
 def get_all_members(cluster, role='master'):
     if role == 'master':
-        if cluster.leader is not None:
+        if cluster.leader is not None and cluster.leader.name:
             yield cluster.leader
         return
 
@@ -215,6 +248,15 @@ def get_any_member(cluster, role='master', member=None):
             return m
 
 
+def get_all_members_leader_first(cluster):
+    leader_name = cluster.leader.member.name if cluster.leader and cluster.leader.member.api_url else None
+    if leader_name:
+        yield cluster.leader.member
+    for member in cluster.members:
+        if member.api_url and member.name != leader_name:
+            yield member
+
+
 def get_cursor(cluster, connect_parameters, role='master', member=None):
     member = get_any_member(cluster, role=role, member=member)
     if member is None:
@@ -227,13 +269,14 @@ 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':
@@ -244,10 +287,12 @@ 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, ask_confirmation=True):
     candidates = {m.name: m for m in cluster.members}
 
     if not force or role:
+        if not member_names and not candidates:
+            raise PatroniCtlException('{0} cluster doesn\'t have any members'.format(cluster_name))
         output_members(cluster, cluster_name)
 
     if role:
@@ -267,12 +312,25 @@ def get_members(cluster, cluster_name, member_names, role, force, action):
         if member_name not in candidates:
             raise PatroniCtlException('{0} is not a member of cluster'.format(member_name))
 
-    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))
+    members = [candidates[n] for n in member_names]
+    if ask_confirmation:
+        confirm_members_action(members, force, action)
+    return members
 
-    return [candidates[n] for n in member_names]
+
+def confirm_members_action(members, force, action, scheduled_at=None):
+    if scheduled_at:
+        if not force:
+            confirm = click.confirm('Are you sure you want to schedule {0} of members {1} at {2}?'
+                                    .format(action, ', '.join([m.name for m in members]), 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([m.name for m in members])))
+            if not confirm:
+                raise PatroniCtlException('Aborted {0}'.format(action))
 
 
 @ctl.command('dsn', help='Generate a dsn for the provided member, defaults to a dsn of the master')
@@ -298,8 +356,7 @@ def dsn(obj, cluster_name, role, member):
 
 @ctl.command('query', help='Query a Patroni PostgreSQL member')
 @arg_cluster_name
-@option_format
-@click.option('--format', 'fmt', help='Output format (pretty, json)', default='tsv')
+@click.option('--format', 'fmt', help='Output format (pretty, tsv, json, yaml)', default='tsv')
 @click.option('--file', '-f', 'p_file', help='Execute the SQL commands from this file', type=click.File('rb'))
 @click.option('--password', help='force password prompt', is_flag=True)
 @click.option('-U', '--username', help='database user name', type=str)
@@ -356,11 +413,12 @@ def query(
         if cursor is None:
             cluster = dcs.get_cluster()
 
-        output, cursor = query_member(cluster, cursor, member, role, command, connect_parameters)
-        print_output(None, output, fmt=fmt, delimiter=delimiter)
+        output, header = query_member(cluster, cursor, member, role, command, connect_parameters)
+        print_output(header, output, fmt=fmt, delimiter=delimiter)
 
 
 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)
@@ -373,15 +431,8 @@ 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()')
-        in_recovery = cursor.fetchone()[0]
-
-        if in_recovery and role == 'master' or not in_recovery and role == 'replica':
-            cursor.connection.close()
-            return None, None
-
         cursor.execute(command)
-        return cursor.fetchall(), cursor
+        return cursor.fetchall(), [d.name for d in cursor.description]
     except (psycopg2.OperationalError, psycopg2.DatabaseError) as oe:
         logging.debug(oe)
         if cursor is not None and not cursor.connection.closed:
@@ -421,9 +472,9 @@ def remove(obj, cluster_name, fmt):
 
 
 def check_response(response, member_name, action_name, silent_success=False):
-    if response.status_code >= 400:
+    if response.status >= 400:
         click.echo('Failed: {0} for member {1}, status code={2}, ({3})'.format(
-            action_name, member_name, response.status_code, response.text
+            action_name, member_name, response.status, response.data.decode('utf-8')
         ))
         return False
     elif not silent_success:
@@ -436,7 +487,7 @@ def parse_scheduled(scheduled):
         try:
             scheduled_at = dateutil.parser.parse(scheduled)
             if scheduled_at.tzinfo is None:
-                scheduled_at = tzlocal.get_localzone().localize(scheduled_at)
+                scheduled_at = scheduled_at.replace(tzinfo=dateutil.tz.tzlocal())
         except (ValueError, TypeError):
             message = 'Unable to parse scheduled timestamp ({0}). It should be in an unambiguous format (e.g. ISO 8601)'
             raise PatroniCtlException(message.format(scheduled))
@@ -457,18 +508,17 @@ def reload(obj, cluster_name, member_names, force, role):
 
     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:
+        r = request_patroni(member, 'post', 'reload')
+        if r.status == 200:
             click.echo('No changes to apply on member {0}'.format(member.name))
-        elif r.status_code == 202:
+        elif r.status == 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)
+                member.name, r.status, r.data.decode('utf-8'))
             )
 
 
@@ -490,7 +540,15 @@ def reload(obj, cluster_name, member_names, force, role):
 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')
+    members = get_members(cluster, cluster_name, member_names, role, force, 'restart', False)
+    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)
+    confirm_members_action(members, force, 'restart', scheduled_at)
+
     if p_any:
         random.shuffle(members)
         members = members[:1]
@@ -505,16 +563,12 @@ def restart(obj, cluster_name, member_names, force, role, p_any, scheduled, vers
 
     if version:
         try:
-            Postgresql.postgres_version_to_int(version)
+            postgres_version_to_int(version)
         except PatroniException as e:
             raise PatroniCtlException(e.value)
 
         content['postgres_version'] = version
 
-    if scheduled is None and not force:
-        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")
@@ -526,19 +580,19 @@ def restart(obj, cluster_name, member_names, force, role, p_any, scheduled, vers
     for member in members:
         if 'schedule' in content:
             if force and member.data.get('scheduled_restart'):
-                r = request_patroni(member, 'delete', 'restart', headers=auth_header(obj))
+                r = request_patroni(member, 'delete', 'restart')
                 check_response(r, member.name, 'flush scheduled restart', True)
 
-        r = request_patroni(member, 'post', 'restart', content, auth_header(obj))
-        if r.status_code == 200:
+        r = request_patroni(member, 'post', 'restart', content)
+        if r.status == 200:
             click.echo('Success: restart on member {0}'.format(member.name))
-        elif r.status_code == 202:
+        elif r.status == 202:
             click.echo('Success: restart scheduled on member {0}'.format(member.name))
-        elif r.status_code == 409:
+        elif r.status == 409:
             click.echo('Failed: another restart is already scheduled on member {0}'.format(member.name))
         else:
             click.echo('Failed: restart for member {0}, status code={1}, ({2})'.format(
-                member.name, r.status_code, r.text)
+                member.name, r.status, r.data.decode('utf-8'))
             )
 
 
@@ -546,20 +600,39 @@ def restart(obj, cluster_name, member_names, force, role, p_any, scheduled, vers
 @click.argument('cluster_name')
 @click.argument('member_names', nargs=-1)
 @option_force
+@click.option('--wait', help='Wait until reinitialization completes', is_flag=True)
 @click.pass_obj
-def reinit(obj, cluster_name, member_names, force):
+def reinit(obj, cluster_name, member_names, force, wait):
     cluster = get_dcs(obj, cluster_name).get_cluster()
     members = get_members(cluster, cluster_name, member_names, None, force, 'reinitialize')
 
+    wait_on_members = []
     for member in members:
         body = {'force': force}
         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') \
+            r = request_patroni(member, 'post', 'reinitialize', body)
+            started = check_response(r, member.name, 'reinitialize')
+            if not started and r.data.endswith(b' already in progress') \
                     and not force and click.confirm('Do you want to cancel it and reinitialize anyway?'):
                 body['force'] = True
                 continue
             break
+        if started and wait:
+            wait_on_members.append(member)
+
+    last_display = []
+    while wait_on_members:
+        if wait_on_members != last_display:
+            click.echo('Waiting for reinitialize to complete on: {0}'.format(
+                ", ".join(member.name for member in wait_on_members))
+            )
+            last_display[:] = wait_on_members
+        time.sleep(2)
+        for member in wait_on_members:
+            data = json.loads(request_patroni(member, 'get', 'patroni').data.decode('utf-8'))
+            if data.get('state') != 'creating replica':
+                click.echo('Reinitialize is completed on: {0}'.format(member.name))
+                wait_on_members.remove(member)
 
 
 def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, force, scheduled=None):
@@ -573,7 +646,7 @@ def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, for
     dcs = get_dcs(obj, cluster_name)
     cluster = dcs.get_cluster()
 
-    if action == 'switchover' and cluster.leader is None:
+    if action == 'switchover' and (cluster.leader is None or not cluster.leader.name):
         raise PatroniCtlException('This cluster has no master')
 
     if master is None:
@@ -585,7 +658,8 @@ def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, for
     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()
 
@@ -605,10 +679,12 @@ def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, for
         raise PatroniCtlException('Member {0} does not exist in cluster {1}'.format(candidate, cluster_name))
 
     scheduled_at_str = None
+    scheduled_at = None
 
     if action == 'switchover':
         if scheduled is None and not force:
-            scheduled = click.prompt('When should the switchover take place (e.g. 2015-10-01T14:30) ',
+            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)
@@ -627,27 +703,32 @@ def _do_failover_or_switchover(obj, action, cluster_name, master, candidate, for
 
     if not force:
         demote_msg = ', demoting current master ' + master if master 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)
+        if scheduled_at_str:
+            if not click.confirm('Are you sure you want to schedule {0} of cluster {1} at {2}{3}?'
+                                 .format(action, cluster_name, scheduled_at_str, demote_msg)):
+                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 cluster.get_member(candidate, False)
 
-        r = request_patroni(member, 'post', action, failover_value, auth_header(obj))
+        r = request_patroni(member, 'post', action, failover_value)
 
         # 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))
+        if r.status == 501 and action == 'switchover' and b'Server does not support this operation' in r.data:
+            r = request_patroni(member, 'post', 'failover', failover_value)
 
-        if r.status_code in (200, 202):
+        if r.status in (200, 202):
             logging.debug(r)
             cluster = dcs.get_cluster()
             logging.debug(cluster)
-            click.echo('{0} {1}'.format(timestamp(), r.text))
+            click.echo('{0} {1}'.format(timestamp(), r.data.decode('utf-8')))
         else:
-            click.echo('{0} failed, details: {1}, {2}'.format(action.title(), r.status_code, r.text))
+            click.echo('{0} failed, details: {1}, {2}'.format(action.title(), r.status, r.data.decode('utf-8')))
             return
     except Exception:
         logging.exception(r)
@@ -681,75 +762,84 @@ def switchover(obj, cluster_name, master, candidate, force, scheduled):
     _do_failover_or_switchover(obj, 'switchover', cluster_name, master, candidate, force, scheduled)
 
 
+def generate_topology(level, member, topology):
+    members = topology.get(member['name'], [])
+
+    if level > 0:
+        member['name'] = '{0}+ {1}'.format((' ' * (level - 1) * 2), member['name'])
+
+    if member['name']:
+        yield member
+
+    for member in members:
+        for member in generate_topology(level + 1, member, topology):
+            yield member
+
+
+def topology_sort(members):
+    topology = defaultdict(list)
+    leader = next((m for m in members if m['role'].endswith('leader')), {'name': None})
+    replicas = set(member['name'] for member in members if not member['role'].endswith('leader'))
+    for member in members:
+        if not member['role'].endswith('leader'):
+            parent = member.get('tags', {}).get('replicatefrom')
+            parent = parent if parent and parent != member['name'] and parent in replicas else leader['name']
+            topology[parent].append(member)
+    for member in generate_topology(0, leader, topology):
+        yield member
+
+
 def output_members(cluster, name, extended=False, fmt='pretty'):
     rows = []
     logging.debug(cluster)
-    leader_name = None
-    if cluster.leader:
-        leader_name = cluster.leader.name
+    initialize = {None: 'uninitialized', '': 'initializing'}.get(cluster.initialize, cluster.initialize)
+    cluster = cluster_as_json(cluster)
 
-    xlog_location_cluster = cluster.last_leader_operation or 0
+    columns = ['Cluster', 'Member', 'Host', 'Role', 'State', 'TL', 'Lag in MB']
+    for c in ('Pending restart', 'Scheduled restart', 'Tags'):
+        if extended or any(m.get(c.lower().replace(' ', '_')) for m in cluster['members']):
+            columns.append(c)
 
-    # Mainly for consistent pretty printing and watching we sort the output
-    cluster.members.sort(key=lambda x: x.name)
+    # Show Host as 'host:port' if somebody is running on non-standard port or two nodes are running on the same host
+    members = [m for m in cluster['members'] if 'host' in m]
+    append_port = any('port' in m and m['port'] != 5432 for m in members) or\
+        len(set(m['host'] for m in members)) < len(members)
 
-    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)
-
-    for m in cluster.members:
+    sort = topology_sort if fmt == 'topology' else iter
+    for m in sort(cluster['members']):
         logging.debug(m)
 
-        role = ''
-        if m.name == leader_name:
-            role = 'Leader'
-        elif m.name == cluster.sync.sync_standby:
-            role = 'Sync standby'
+        lag = m.get('lag', '')
+        m.update(cluster=name, member=m['name'], host=m.get('host', ''), tl=m.get('timeline', ''),
+                 role=m['role'].replace('_', ' ').title(),
+                 lag_in_mb=round(lag/1024/1024) if isinstance(lag, six.integer_types) else lag,
+                 pending_restart='*' if m.get('pending_restart') else '')
 
-        xlog_location = m.data.get('xlog_location')
-        lag = ''
-        if xlog_location is None:
-            lag = 'unknown'
-        elif xlog_location_cluster >= xlog_location:
-            lag = round((xlog_location_cluster - xlog_location)/1024/1024)
+        if append_port and m['host'] and m.get('port'):
+            m['host'] = ':'.join([m['host'], str(m['port'])])
 
-        row = [name, m.name, m.conn_kwargs()['host'], role, m.data.get('state', ''), lag]
+        if 'scheduled_restart' in m:
+            value = m['scheduled_restart']['schedule']
+            if 'postgres_version' in m['scheduled_restart']:
+                value += ' if version < {0}'.format(m['scheduled_restart']['postgres_version'])
+            m['scheduled_restart'] = value
 
-        if extended or has_pending_restarts:
-            row.append('*' if m.data.get('pending_restart') else '')
+        rows.append([m.get(n.lower().replace(' ', '_'), '') for n in columns])
 
-        if extended or has_scheduled_restarts:
-            value = ''
-            scheduled_restart = m.data.get('scheduled_restart')
-            if scheduled_restart:
-                value = scheduled_restart['schedule']
-                if 'postgres_version' in scheduled_restart:
-                    value += ' if version < {0}'.format(scheduled_restart['postgres_version'])
+    print_output(columns, rows, {'Lag in MB': 'r', 'TL': 'r'}, fmt, ' Cluster: {0} ({1}) '.format(name, initialize))
 
-            row.append(value)
-
-        rows.append(row)
-
-    columns = ['Cluster', 'Member', 'Host', 'Role', 'State', 'Lag in MB']
-    alignment = {'Lag in MB': 'r'}
-
-    if extended or has_pending_restarts:
-        columns.append('Pending restart')
-
-    if extended or has_scheduled_restarts:
-        columns.append('Scheduled restart')
-
-    print_output(columns, rows, alignment, fmt)
+    if fmt not in ('pretty', 'topology'):  # Omit service info when using machine-readable formats
+        return
 
     service_info = []
-    if cluster.is_paused():
+    if cluster.get('pause'):
         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
+    if 'scheduled_switchover' in cluster:
+        info = 'Switchover scheduled at: ' + cluster['scheduled_switchover']['at']
+        for name in ('from', 'to'):
+            if name in cluster['scheduled_switchover']:
+                info += '\n{0:>24}: {1}'.format(name, cluster['scheduled_switchover'][name])
         service_info.append(info)
 
     if service_info:
@@ -782,6 +872,16 @@ def members(obj, cluster_names, fmt, watch, w, extended, ts):
             output_members(cluster, cluster_name, extended, fmt)
 
 
+@ctl.command('topology', help='Prints ASCII topology for given cluster')
+@click.argument('cluster_names', nargs=-1)
+@option_watch
+@option_watchrefresh
+@click.pass_obj
+@click.pass_context
+def topology(ctx, obj, cluster_names, watch, w):
+    ctx.forward(members, fmt='topology')
+
+
 def timestamp(precision=6):
     return datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S.%f')[:precision - 7]
 
@@ -849,25 +949,45 @@ def scaffold(obj, cluster_name, sysid):
     click.echo("Cluster {0} has been created successfully".format(cluster_name))
 
 
-@ctl.command('flush', help='Flush scheduled events')
+@ctl.command('flush', help='Discard scheduled events')
 @click.argument('cluster_name')
 @click.argument('member_names', nargs=-1)
-@click.argument('target', type=click.Choice(['restart']))
+@click.argument('target', type=click.Choice(['restart', 'switchover']))
 @click.option('--role', '-r', help='Flush only members with this role', default='any',
               type=click.Choice(['master', 'replica', 'any']))
 @option_force
 @click.pass_obj
 def flush(obj, cluster_name, member_names, force, role, target):
-    cluster = get_dcs(obj, cluster_name).get_cluster()
+    dcs = get_dcs(obj, cluster_name)
+    cluster = dcs.get_cluster()
 
-    members = get_members(cluster, cluster_name, member_names, role, force, 'flush')
-    for member in members:
-        if target == 'restart':
+    if target == 'restart':
+        for member in get_members(cluster, cluster_name, member_names, role, force, 'flush'):
             if member.data.get('scheduled_restart'):
-                r = request_patroni(member, 'delete', 'restart', None, auth_header(obj))
+                r = request_patroni(member, 'delete', 'restart')
                 check_response(r, member.name, 'flush scheduled restart')
             else:
                 click.echo('No scheduled restart for member {0}'.format(member.name))
+    elif target == 'switchover':
+        failover = cluster.failover
+        if not failover or not failover.scheduled_at:
+            return click.echo('No pending scheduled switchover')
+        for member in get_all_members_leader_first(cluster):
+            try:
+                r = request_patroni(member, 'delete', 'switchover')
+                if r.status in (200, 404):
+                    prefix = 'Success' if r.status == 200 else 'Failed'
+                    return click.echo('{0}: {1}'.format(prefix, r.data.decode('utf-8')))
+            except Exception as err:
+                logging.warning(str(err))
+                logging.warning('Member %s is not accessible', member.name)
+
+            click.echo('Failed: member={0}, status_code={1}, ({2})'.format(
+                member.name, r.status, r.data.decode('utf-8')))
+
+        logging.warning('Failing over to DCS')
+        click.echo('{0} Could not find any accessible member of cluster {1}'.format(timestamp(), cluster_name))
+        dcs.manual_failover('', '', index=failover.index)
 
 
 def wait_until_pause_is_applied(dcs, paused, old_cluster):
@@ -894,26 +1014,22 @@ def toggle_pause(config, cluster_name, paused, wait):
     if cluster.is_paused() == paused:
         raise PatroniCtlException('Cluster is {0} paused'.format(paused and 'already' or 'not'))
 
-    members = []
-    if cluster.leader:
-        members.append(cluster.leader.member)
-    members.extend([m for m in cluster.members if m.api_url and (not members or members[0].name != m.name)])
-
-    for member in members:
+    for member in get_all_members_leader_first(cluster):
         try:
-            r = request_patroni(member, 'patch', 'config', {'pause': paused or None}, auth_header(config))
-        except Exception:
+            r = request_patroni(member, 'patch', 'config', {'pause': paused or None})
+        except Exception as err:
+            logging.warning(str(err))
             logging.warning('Member %s is not accessible', member.name)
             continue
 
-        if r.status_code == 200:
+        if r.status == 200:
             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))
+                       paused and 'pause' or 'resume', r.status, r.data.decode('utf-8')))
         break
     else:
         raise PatroniCtlException('Can not find accessible cluster member')
@@ -960,7 +1076,7 @@ def show_diff(before_editing, after_editing):
     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')]
+        return [line + '\n' for line in string.rstrip('\n').split('\n')]
 
     unified_diff = difflib.unified_diff(listify(before_editing), listify(after_editing))
 
@@ -968,25 +1084,32 @@ def show_diff(before_editing, after_editing):
         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.write(six.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)
+            wrap = True
+            if find_executable('less'):
+                pager = None
+            else:
+                pager = 'more.com' if sys.platform == 'win32' else 'more'
+            pager_options = None
+
+        markup_to_pager(PatchStream(buf), opts)
     else:
         for line in unified_diff:
             click.echo(line.rstrip('\n'))
 
 
-def format_config_for_editing(data):
+def format_config_for_editing(data, default_flow_style=False):
     """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)
+    return yaml.safe_dump(data, default_flow_style=default_flow_style, encoding=None, allow_unicode=True, width=200)
 
 
 def apply_config_changes(before_editing, data, kvpairs):
@@ -1056,9 +1179,16 @@ def invoke_editor(before_editing, cluster_name):
     :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:
-        raise PatroniCtlException('EDITOR environment variable is not set')
+        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',
@@ -1152,8 +1282,8 @@ def version(obj, cluster_name, member_names):
         if m.api_url:
             if not member_names or m.name in member_names:
                 try:
-                    response = request_patroni(m, 'get', 'patroni')
-                    data = response.json()
+                    response = request_patroni(m)
+                    data = json.loads(response.data.decode('utf-8'))
                     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 ""
@@ -1162,6 +1292,19 @@ def version(obj, cluster_name, member_names):
                     click.echo("{0}: failed to get version: {1}".format(m.name, e))
 
 
+@ctl.command('history', help="Show the history of failovers/switchovers")
+@arg_cluster_name
+@option_format
+@click.pass_obj
+def history(obj, cluster_name, fmt):
+    cluster = get_dcs(obj, cluster_name).get_cluster()
+    history = cluster.history and cluster.history.lines or []
+    for line in history:
+        if len(line) < 4:
+            line.append('')
+    print_output(['TL', 'LSN', 'Reason', 'Timestamp'], history, {'TL': 'r', 'LSN': 'r'}, fmt)
+
+
 def format_pg_version(version):
     if version < 100000:
         return "{0}.{1}.{2}".format(version // 10000, version // 100 % 100, version % 100)

patroni/daemon.py

@@ -0,0 +1,104 @@
+import abc
+import os
+import signal
+import six
+import sys
+
+from threading import Lock
+
+
+@six.add_metaclass(abc.ABCMeta)
+class AbstractPatroniDaemon(object):
+
+    def __init__(self, config):
+        from patroni.log import PatroniLogger
+
+        self.setup_signal_handlers()
+
+        self.logger = PatroniLogger()
+        self.config = config
+        AbstractPatroniDaemon.reload_config(self, local=True)
+
+    def sighup_handler(self, *args):
+        self._received_sighup = True
+
+    def sigterm_handler(self, *args):
+        with self._sigterm_lock:
+            if not self._received_sigterm:
+                self._received_sigterm = True
+                sys.exit()
+
+    def setup_signal_handlers(self):
+        self._received_sighup = False
+        self._sigterm_lock = Lock()
+        self._received_sigterm = False
+        if os.name != 'nt':
+            signal.signal(signal.SIGHUP, self.sighup_handler)
+        signal.signal(signal.SIGTERM, self.sigterm_handler)
+
+    @property
+    def received_sigterm(self):
+        with self._sigterm_lock:
+            return self._received_sigterm
+
+    def reload_config(self, sighup=False, local=False):
+        if local:
+            self.logger.reload_config(self.config.get('log', {}))
+
+    @abc.abstractmethod
+    def _run_cycle(self):
+        """_run_cycle"""
+
+    def run(self):
+        self.logger.start()
+        while not self.received_sigterm:
+            if self._received_sighup:
+                self._received_sighup = False
+                self.reload_config(True, self.config.reload_local_configuration())
+
+            self._run_cycle()
+
+    @abc.abstractmethod
+    def _shutdown(self):
+        """_shutdown"""
+
+    def shutdown(self):
+        with self._sigterm_lock:
+            self._received_sigterm = True
+        self._shutdown()
+        self.logger.shutdown()
+
+
+def abstract_main(cls, validator=None):
+    import argparse
+
+    from .config import Config, ConfigParseError
+    from .version import __version__
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--version', action='version', version='%(prog)s {0}'.format(__version__))
+    if validator:
+        parser.add_argument('--validate-config', action='store_true', help='Run config validator and exit')
+    parser.add_argument('configfile', nargs='?', default='',
+                        help='Patroni may also read the configuration from the {0} environment variable'
+                        .format(Config.PATRONI_CONFIG_VARIABLE))
+    args = parser.parse_args()
+    try:
+        if validator and args.validate_config:
+            Config(args.configfile, validator=validator)
+            sys.exit()
+
+        config = Config(args.configfile)
+    except ConfigParseError as e:
+        if e.value:
+            print(e.value)
+        parser.print_help()
+        sys.exit(1)
+
+    controller = cls(config)
+    try:
+        controller.run()
+    except KeyboardInterrupt:
+        pass
+    finally:
+        controller.shutdown()

patroni/dcs/__init__.py

@@ -6,19 +6,39 @@ import json
 import logging
 import os
 import pkgutil
+import re
 import six
 import sys
+import time
 
-from collections import namedtuple
-from patroni.exceptions import PatroniException
-from patroni.utils import parse_bool
+from collections import defaultdict, namedtuple
+from copy import deepcopy
+from patroni.exceptions import PatroniFatalException
+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:
         postgres://{username}:{password}@{connect_address}/postgres
@@ -46,33 +66,44 @@ def dcs_modules():
     module_prefix = __package__ + '.'
 
     if getattr(sys, 'frozen', False):
-        importer = pkgutil.get_importer(dcs_dirname)
-        return [module for module in list(importer.toc) if module.startswith(module_prefix) and module.count('.') == 2]
+        toc = set()
+        for importer in pkgutil.iter_importers(dcs_dirname):
+            if hasattr(importer, 'toc'):
+                toc |= importer.toc
+        return [module for module in toc if module.startswith(module_prefix) and module.count('.') == 2]
     else:
         return [module_prefix + name for _, name, is_pkg in pkgutil.iter_modules([dcs_dirname]) if not is_pkg]
 
 
 def get_dcs(config):
-    available_implementations = set()
-    for module_name in dcs_modules():
-        try:
-            module = importlib.import_module(module_name)
-            for name in filter(lambda name: not name.startswith('__'), dir(module)):  # iterate through module content
-                item = getattr(module, name)
-                name = name.lower()
-                # try to find implementation of AbstractDCS interface, class name must match with module_name
-                if inspect.isclass(item) and issubclass(item, AbstractDCS) and __package__ + '.' + name == module_name:
-                    available_implementations.add(name)
-                    if name in config:  # which has configuration section in the config file
+    modules = dcs_modules()
+
+    for module_name in modules:
+        name = module_name.split('.')[-1]
+        if name in config:  # we will try to import only modules which have configuration section in the config file
+            try:
+                module = importlib.import_module(module_name)
+                for key, item in module.__dict__.items():  # iterate through the module content
+                    # try to find implementation of AbstractDCS interface, class name must match with module_name
+                    if key.lower() == name and inspect.isclass(item) and issubclass(item, AbstractDCS):
                         # propagate some parameters
                         config[name].update({p: config[p] for p in ('namespace', 'name', 'scope', 'loop_wait',
                                              'patronictl', 'ttl', 'retry_timeout') if p in config})
                         return item(config[name])
+            except ImportError:
+                logger.debug('Failed to import %s', module_name)
+
+    available_implementations = []
+    for module_name in modules:
+        name = module_name.split('.')[-1]
+        try:
+            module = importlib.import_module(module_name)
+            available_implementations.extend(name for key, item in module.__dict__.items() if key.lower() == name
+                                             and inspect.isclass(item) and issubclass(item, AbstractDCS))
         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))
+            logger.info('Failed to import %s', module_name)
+    raise PatroniFatalException("""Can not find suitable configuration of distributed configuration store
+Available implementations: """ + ', '.join(sorted(set(available_implementations))))
 
 
 class Member(namedtuple('Member', 'index,name,session,data')):
@@ -108,14 +139,31 @@ 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')
+        if conn_url:
+            return conn_url
+
+        conn_kwargs = self.data.get('conn_kwargs')
+        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": None,
+            "port": None,
+            "database": None
+        }
         ret = self.data.get('conn_kwargs')
         if ret:
-            ret = ret.copy()
+            defaults.update(ret)
+            ret = defaults
         else:
-            r = urlparse(self.conn_url)
+            conn_url = self.conn_url
+            if not conn_url:
+                return {}  # due to the invalid conn_url we don't care about authentication parameters
+            r = urlparse(conn_url)
             ret = {
                 'host': r.hostname,
                 'port': r.port or 5432,
@@ -123,11 +171,11 @@ class Member(namedtuple('Member', 'index,name,session,data')):
             }
             self.data['conn_kwargs'] = ret.copy()
 
+        # apply any remaining authentication parameters
         if auth and isinstance(auth, dict):
+            ret.update({k: v for k, v in auth.items() if v is not None})
             if 'username' in auth:
-                ret['user'] = auth['username']
-            if 'password' in auth:
-                ret['password'] = auth['password']
+                ret['user'] = ret.pop('username')
         return ret
 
     @property
@@ -159,6 +207,26 @@ 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.
@@ -178,9 +246,27 @@ class Leader(namedtuple('Leader', 'index,session,member')):
     def conn_url(self):
         return self.member.conn_url
 
+    @property
+    def data(self):
+        return self.member.data
+
     @property
     def timeline(self):
-        return self.member.data.get('timeline')
+        return self.data.get('timeline')
+
+    @property
+    def checkpoint_after_promote(self):
+        """
+        >>> Leader(1, '', Member.from_node(1, '', '', '{"version":"z"}')).checkpoint_after_promote
+        """
+        version = self.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.data['role'] == 'master' and 'checkpoint_after_promote' not in self.data
+            except Exception:
+                logger.debug('Failed to parse Patroni version %s', version)
 
 
 class Failover(namedtuple('Failover', 'index,leader,candidate,scheduled_at')):
@@ -236,14 +322,32 @@ 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
-        return ClusterConfig(index, data, modify_index or index)
+            data = None
+            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 {}
+
+    @property
+    def ignore_slots_matchers(self):
+        return isinstance(self.data, dict) and self.data.get('ignore_slots') or []
+
+    @property
+    def max_timelines_history(self):
+        return self.data.get('max_timelines_history', 0)
 
 
 class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
@@ -251,7 +355,7 @@ class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
 
     :param index: modification index of a synchronization key in a Configuration Store
     :param leader: reference to member that was leader
-    :param sync_standby: standby that was last synchronized to leader
+    :param sync_standby: synchronous standby list (comma delimited) which are last synchronized to leader
     """
 
     @staticmethod
@@ -283,15 +387,22 @@ class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
             data = {}
         return SyncState(index, data.get('leader'), data.get('sync_standby'))
 
+    @property
+    def members(self):
+        """ Returns sync_standby in list """
+        return self.sync_standby and self.sync_standby.split(',') or []
+
     def matches(self, name):
         """
         Returns if a node name matches one of the nodes in the sync state
 
-        >>> s = SyncState(1, 'foo', 'bar')
+        >>> s = SyncState(1, 'foo', 'bar,zoo')
         >>> s.matches('foo')
         True
         >>> s.matches('bar')
         True
+        >>> s.matches('zoo')
+        True
         >>> s.matches('baz')
         False
         >>> s.matches(None)
@@ -299,10 +410,10 @@ class SyncState(namedtuple('SyncState', 'index,leader,sync_standby')):
         >>> SyncState(1, None, None).matches('foo')
         False
         """
-        return name is not None and name in (self.leader, self.sync_standby)
+        return name is not None and name in [self.leader] + self.members
 
 
-class TimelineHistory(namedtuple('TimelineHistory', 'index,lines')):
+class TimelineHistory(namedtuple('TimelineHistory', 'index,value,lines')):
     """Object representing timeline history file"""
 
     @staticmethod
@@ -318,7 +429,7 @@ class TimelineHistory(namedtuple('TimelineHistory', 'index,lines')):
             lines = None
         if not isinstance(lines, list):
             lines = []
-        return TimelineHistory(index, lines)
+        return TimelineHistory(index, value, lines)
 
 
 class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operation,members,failover,sync,history')):
@@ -359,6 +470,87 @@ class Cluster(namedtuple('Cluster', 'initialize,config,leader,last_leader_operat
     def is_synchronous_mode(self):
         return self.check_mode('synchronous_mode')
 
+    def get_replication_slots(self, my_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
+        use_slots = self.config and self.config.data.get('postgresql', {}).get('use_slots', True)
+        if role in ('master', 'standby_leader'):
+            slot_members = [m.name for m in self.members if use_slots and m.name != my_name and
+                            (m.replicatefrom is None or m.replicatefrom == my_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 use_slots and
+                            m.replicatefrom == my_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
+
+            value = deepcopy(value) if value else {'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':
+                    # Don't try to create permanent physical replication slot for yourself
+                    if name != slot_name_from_member_name(my_name):
+                        slots[name] = value
+                    continue
+                elif value['type'] == 'logical' and value.get('database') and value.get('plugin'):
+                    if name in slots:
+                        logger.error("Permanent logical replication slot {'%s': %s} is conflicting with" +
+                                     " physical replication slot for cluster member", name, value)
+                    else:
+                        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)
 class AbstractDCS(object):
@@ -379,11 +571,12 @@ class AbstractDCS(object):
             i.e.: `zookeeper` for zookeeper, `etcd` for etcd, etc...
         """
         self._name = config['name']
-        self._base_path = os.path.join('/', config.get('namespace', '/service/').strip('/'), config['scope'])
+        self._base_path = re.sub('/+', '/', '/'.join(['', config.get('namespace', 'service'), 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()
@@ -431,6 +624,10 @@ 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"""
@@ -457,23 +654,32 @@ class AbstractDCS(object):
            If the current node was running as a master and exception raised,
            instance would be demoted."""
 
-    def get_cluster(self):
+    def _bypass_caches(self):
+        """Used only in zookeeper"""
+
+    def get_cluster(self, force=False):
+        if force:
+            self._bypass_caches()
+        try:
+            cluster = self._load_cluster()
+        except Exception:
+            self.reset_cluster()
+            raise
+
         with self._cluster_thread_lock:
-            try:
-                self._load_cluster()
-            except Exception:
-                self._cluster = None
-                raise
-            return self._cluster
+            self._cluster = cluster
+            self._cluster_valid_till = time.time() + self.ttl
+            return 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):
@@ -495,7 +701,7 @@ 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):
+    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
@@ -539,7 +745,7 @@ 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.
@@ -569,10 +775,19 @@ class AbstractDCS(object):
         otherwise it should return `!False`"""
 
     @abc.abstractmethod
-    def delete_leader(self):
-        """Voluntarily remove leader key from DCS
+    def _delete_leader(self):
+        """Remove leader key from DCS.
         This method should remove leader key if current instance is the leader"""
 
+    def delete_leader(self, last_operation=None):
+        """Update optime/leader and voluntarily remove leader key from DCS.
+        This method should remove leader key if current instance is the leader.
+        :param last_operation: latest checkpoint location in bytes"""
+
+        if last_operation:
+            self.write_leader_optime(last_operation)
+        return self._delete_leader()
+
     @abc.abstractmethod
     def cancel_initialization(self):
         """ Removes the initialize key for a cluster """
@@ -583,8 +798,10 @@ class AbstractDCS(object):
 
     @staticmethod
     def sync_state(leader, sync_standby):
-        """Build sync_state dict"""
-        return {'leader': leader, 'sync_standby': sync_standby}
+        """Build sync_state dict
+           sync_standby dictionary key being kept for backward compatibility
+        """
+        return {'leader': leader, 'sync_standby': sync_standby and ','.join(sorted(sync_standby)) or None}
 
     def write_sync_state(self, leader, sync_standby, index=None):
         sync_value = self.sync_state(leader, sync_standby)

patroni/dcs/consul.py

@@ -2,19 +2,22 @@ from __future__ import absolute_import
 import json
 import logging
 import os
+import re
 import socket
 import ssl
 import time
 import urllib3
 
+from collections import namedtuple
 from consul import ConsulException, NotFound, base
-from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
-from patroni.exceptions import DCSError
-from patroni.utils import deep_compare, parse_bool, Retry, RetryFailedError, split_host_port
 from urllib3.exceptions import HTTPError
-from six.moves.urllib.parse import urlencode, urlparse
+from six.moves.urllib.parse import urlencode, urlparse, quote
 from six.moves.http_client import HTTPException
 
+from . import AbstractDCS, Cluster, ClusterConfig, Failover, Leader, Member, SyncState, TimelineHistory
+from ..exceptions import DCSError
+from ..utils import deep_compare, parse_bool, Retry, RetryFailedError, split_host_port, uri, USER_AGENT
+
 logger = logging.getLogger(__name__)
 
 
@@ -26,16 +29,23 @@ class ConsulInternalError(ConsulException):
     """An internal Consul server error occurred"""
 
 
-class InvalidSessionTTL(ConsulInternalError):
+class InvalidSessionTTL(ConsulException):
     """Session TTL is too small or too big"""
 
 
+class InvalidSession(ConsulException):
+    """invalid session"""
+
+
+Response = namedtuple('Response', 'code,headers,body,content')
+
+
 class HTTPClient(object):
 
     def __init__(self, host='127.0.0.1', port=8500, token=None, scheme='http', verify=True, cert=None, ca_cert=None):
         self.token = token
         self._read_timeout = 10
-        self.base_uri = '{0}://{1}:{2}'.format(scheme, host, port)
+        self.base_uri = uri(scheme, (host, port))
         kwargs = {}
         if cert:
             if isinstance(cert, tuple):
@@ -47,9 +57,8 @@ class HTTPClient(object):
                 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)
+        kwargs['cert_reqs'] = ssl.CERT_REQUIRED if verify or ca_cert else ssl.CERT_NONE
+        self.http = urllib3.PoolManager(num_pools=10, maxsize=10, **kwargs)
         self._ttl = None
 
     def set_read_timeout(self, timeout):
@@ -66,14 +75,17 @@ class HTTPClient(object):
 
     @staticmethod
     def response(response):
-        data = response.data.decode('utf-8')
+        content = response.data
+        body = content.decode('utf-8')
         if response.status == 500:
-            msg = '{0} {1}'.format(response.status, data)
-            if data.startswith('Invalid Session TTL'):
+            msg = '{0} {1}'.format(response.status, body)
+            if body.startswith('Invalid Session TTL'):
                 raise InvalidSessionTTL(msg)
+            elif body.startswith('invalid session'):
+                raise InvalidSession(msg)
             else:
                 raise ConsulInternalError(msg)
-        return base.Response(response.status, response.headers, data)
+        return Response(response.status, response.headers, body, content)
 
     def uri(self, path, params=None):
         return '{0}{1}{2}'.format(self.base_uri, path, params and '?' + urlencode(params) or '')
@@ -82,7 +94,7 @@ class HTTPClient(object):
         if method not in ('get', 'post', 'put', 'delete'):
             raise AttributeError("HTTPClient instance has no attribute '{0}'".format(method))
 
-        def wrapper(callback, path, params=None, data=''):
+        def wrapper(callback, path, params=None, data='', headers=None):
             # python-consul doesn't allow to specify ttl smaller then 10 seconds
             # because session_ttl_min defaults to 10s, so we have to do this ugly dirty hack...
             if method == 'put' and path == '/v1/session/create':
@@ -95,12 +107,19 @@ class HTTPClient(object):
                 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
+            kwargs['headers'] = (headers or {}).copy()
+            kwargs['headers'].update(urllib3.make_headers(user_agent=USER_AGENT))
             token = params.pop('token', self.token) if isinstance(params, dict) else self.token
             if token:
-                kwargs['headers'] = {'X-Consul-Token': token}
+                kwargs['headers']['X-Consul-Token'] = token
             return callback(self.response(self.http.request(method.upper(), self.uri(path, params), **kwargs)))
         return wrapper
 
@@ -110,19 +129,27 @@ class ConsulClient(base.Consul):
     def __init__(self, *args, **kwargs):
         self._cert = kwargs.pop('cert', None)
         self._ca_cert = kwargs.pop('ca_cert', None)
-        self._token = kwargs.get('token')
+        self.token = kwargs.get('token')
         super(ConsulClient, self).__init__(*args, **kwargs)
 
-    def connect(self, *args, **kwargs):
+    def http_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
+        if self.token:
+            kwargs['token'] = self.token
         return HTTPClient(**kwargs)
 
+    def connect(self, *args, **kwargs):
+        return self.http_connect(*args, **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):
     def wrapper(*args, **kwargs):
@@ -133,6 +160,31 @@ 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):
@@ -144,7 +196,6 @@ class Consul(AbstractDCS):
                             retry_exceptions=(ConsulInternalError, HTTPException,
                                               HTTPError, socket.error, socket.timeout))
 
-        self._my_member_data = {}
         kwargs = {}
         if 'url' in config:
             r = urlparse(config['url'])
@@ -161,7 +212,7 @@ class Consul(AbstractDCS):
         if config.get('key') and config.get('cert'):
             config['cert'] = (config['cert'], config['key'])
 
-        config_keys = ('host', 'port', 'token', 'scheme', 'cert', 'ca_cert', 'dc')
+        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')
@@ -174,7 +225,15 @@ class Consul(AbstractDCS):
         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.__session_checks = config.get('checks', [])
+        self._register_service = config.get('register_service', False)
+        if self._register_service:
+            self._service_tags = config.get('service_tags', [])
+            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()
 
@@ -189,11 +248,19 @@ 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)
@@ -256,7 +323,7 @@ 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)
@@ -299,48 +366,117 @@ 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, history)
+            return Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
         except NotFound:
-            self._cluster = Cluster(None, None, None, None, [], None, None, None)
+            return Cluster(None, None, None, None, [], None, None, None)
         except Exception:
             logger.exception('get_cluster')
             raise ConsulError('Consul is not responding properly')
 
-    def touch_member(self, data, ttl=None, permanent=False):
+    @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 = not permanent and self.refresh_session()
+
+        try:
+            create_member = not permanent and self.refresh_session()
+        except DCSError:
+            return False
 
         if member and (create_member or member.session != self._session):
-            try:
-                self._client.kv.delete(self.member_path)
-                create_member = True
-            except Exception:
-                return False
+            self._client.kv.delete(self.member_path)
+            create_member = True
 
-        if not create_member and member and deep_compare(data, self._my_member_data):
+        if not create_member and member and deep_compare(data, member.data):
             return True
 
         try:
             args = {} if permanent else {'acquire': self._session}
             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 _do_attempt_to_acquire_leader(self, kwargs):
-        return self.retry(self._client.kv.put, self.leader_path, self._name, **kwargs)
+    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))
+        tags = self._service_tags[:]
+        tags.append(role)
+        params = {
+            'service_id': '{0}/{1}'.format(self._scope, self._name),
+            'address': conn_parts.hostname,
+            'port': conn_parts.port,
+            'check': check,
+            'tags': tags
+        }
+
+        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()
 
-        ret = self._do_attempt_to_acquire_leader({} if permanent else {'acquire': self._session})
+        ret = self._do_attempt_to_acquire_leader(permanent)
         if not ret:
             logger.info('Could not take out TTL lock')
+
         return ret
 
     def take_leader(self):
@@ -383,7 +519,7 @@ class Consul(AbstractDCS):
         return self._client.kv.put(self.history_path, value)
 
     @catch_consul_errors
-    def delete_leader(self):
+    def _delete_leader(self):
         cluster = self.cluster
         if cluster and isinstance(cluster.leader, Leader) and cluster.leader.name == self._name:
             return self._client.kv.delete(self.leader_path, cas=cluster.leader.index)
@@ -397,6 +533,7 @@ class Consul(AbstractDCS):
         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
@@ -408,7 +545,7 @@ 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

@@ -1,32 +1,34 @@
 from __future__ import absolute_import
+import abc
 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, TimelineHistory
-from patroni.exceptions import DCSError
-from patroni.utils import Retry, RetryFailedError, split_host_port
-from urllib3.exceptions import HTTPError, ReadTimeoutError
-from requests.exceptions import RequestException
+from urllib3 import Timeout
+from urllib3.exceptions import HTTPError, ReadTimeoutError, ProtocolError
 from six.moves.queue import Queue
 from six.moves.http_client import HTTPException
 from six.moves.urllib_parse import urlparse
 from threading import Thread
 
+from . import AbstractDCS, Cluster, ClusterConfig, Failover, Leader, Member, SyncState, TimelineHistory
+from ..exceptions import DCSError
+from ..request import get as requests_get
+from ..utils import Retry, RetryFailedError, split_host_port, uri, USER_AGENT
+
 logger = logging.getLogger(__name__)
 
 
-def uri(protocol, host, port, endpoint=''):
-    return '{0}://{1}:{2}{3}'.format(protocol, host, port, endpoint)
+class EtcdRaftInternal(etcd.EtcdException):
+    """Raft Internal Error"""
 
 
 class EtcdError(DCSError):
@@ -69,16 +71,20 @@ class DnsCachingResolver(Thread):
     def resolve_async(self, host, port, attempt=0):
         self._resolve_queue.put(((host, port), attempt))
 
+    def remove(self, host, port):
+        self._cache.pop((host, port), None)
+
     @staticmethod
     def _do_resolve(host, port):
         try:
             return socket.getaddrinfo(host, port, 0, socket.SOCK_STREAM, socket.IPPROTO_TCP)
-        except socket.gaierror:
-            logger.warning('failed to resolve host %s', host)
+        except Exception as e:
+            logger.warning('failed to resolve host %s: %s', host, e)
             return []
 
 
-class Client(etcd.Client):
+@six.add_metaclass(abc.ABCMeta)
+class AbstractEtcdClientWithFailover(etcd.Client):
 
     def __init__(self, config, dns_resolver, cache_ttl=300):
         self._dns_resolver = dns_resolver
@@ -86,31 +92,81 @@ class Client(etcd.Client):
         self._machines_cache_updated = 0
         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)
+        super(AbstractEtcdClientWithFailover, 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._initial_machines_cache = []
         self._load_machines_cache()
-        self._allow_reconnect = not self._use_proxies
+        self._allow_reconnect = True
+        # allow passing retry argument to api_execute in params
+        self._comparison_conditions.add('retry')
+        self._read_options.add('retry')
+        self._del_conditions.add('retry')
 
-    def _build_request_parameters(self):
-        kwargs = {'headers': self._get_headers(), 'redirect': self.allow_redirect}
+    def _calculate_timeouts(self, etcd_nodes, timeout=None):
+        """Calculate a request timeout and number of retries per single etcd node.
+        In case if the timeout per node is too small (less than one second) we will reduce the number of nodes.
+        For the cluster with only one node we will try to do 2 retries.
+        For clusters with 2 nodes we will try to do 1 retry for every node.
+        No retries for clusters with 3 or more nodes. We better rely on switching to a different node."""
 
-        # calculate the number of retries and timeout *per node*
-        # actual number of retries depends on the number of nodes
-        etcd_nodes = len(self._machines_cache) + 1
-        kwargs['retries'] = 0 if etcd_nodes > 3 else (1 if etcd_nodes > 1 else 2)
+        per_node_timeout = timeout = float(timeout or self.read_timeout)
 
-        # if etcd_nodes > 3:
-        #     kwargs.update({'retries': 0, 'timeout': float(self.read_timeout)/etcd_nodes})
-        # elif etcd_nodes > 1:
-        #     kwargs.update({'retries': 1, 'timeout': self.read_timeout/2.0/etcd_nodes})
-        # else:
-        #     kwargs.update({'retries': 2, 'timeout': self.read_timeout/3.0})
-        kwargs['timeout'] = self.read_timeout/float(kwargs['retries'] + 1)/etcd_nodes
+        max_retries = 4 - min(etcd_nodes, 3)
+        per_node_retries = 1
+        min_timeout = 1.0
+
+        while etcd_nodes > 0:
+            per_node_timeout = float(timeout) / etcd_nodes
+            if per_node_timeout >= min_timeout:
+                # for small clusters we will try to do more than on try on every node
+                while per_node_retries < max_retries and per_node_timeout / (per_node_retries + 1) >= min_timeout:
+                    per_node_retries += 1
+                per_node_timeout /= per_node_retries
+                break
+            # if the timeout per one node is to small try to reduce number of nodes
+            etcd_nodes -= 1
+            max_retries = 1
+
+        return etcd_nodes, per_node_timeout, per_node_retries - 1
+
+    def reload_config(self, config):
+        self.username = config.get('username')
+        self.password = config.get('password')
+
+    def _get_headers(self):
+        basic_auth = ':'.join((self.username, self.password)) if self.username and self.password else None
+        return urllib3.make_headers(basic_auth=basic_auth, user_agent=USER_AGENT)
+
+    def _prepare_common_parameters(self, etcd_nodes, timeout=None):
+        kwargs = {'headers': self._get_headers(), 'redirect': self.allow_redirect, 'preload_content': False}
+
+        if timeout is not None:
+            kwargs.update(retries=0, timeout=timeout)
+        else:
+            _, per_node_timeout, per_node_retries = self._calculate_timeouts(etcd_nodes)
+            connect_timeout = max(1, per_node_timeout/2)
+            kwargs.update(timeout=Timeout(connect=connect_timeout, total=per_node_timeout), retries=per_node_retries)
         return kwargs
 
     def set_machines_cache_ttl(self, cache_ttl):
         self._machines_cache_ttl = cache_ttl
 
+    @abc.abstractmethod
+    def _prepare_get_members(self, etcd_nodes):
+        """returns: request parameters"""
+
+    @abc.abstractmethod
+    def _get_members(self, base_uri, **kwargs):
+        """returns: list of clientURLs"""
+
+    @property
+    def machines_cache(self):
+        base_uri, cache = self._base_uri, self._machines_cache
+        return ([base_uri] if base_uri in cache else []) + [machine for machine in cache if machine != base_uri]
+
     @property
     def machines(self):
         """Original `machines` method(property) of `etcd.Client` class raise exception
@@ -123,101 +179,101 @@ class Client(etcd.Client):
         Also this method implements the same timeout-retry logic as `api_execute`, because
         the original method was retrying 2 times with the `read_timeout` on each node."""
 
-        kwargs = self._build_request_parameters()
+        machines_cache = self.machines_cache
+        kwargs = self._prepare_get_members(len(machines_cache))
 
-        while True:
+        for base_uri in machines_cache:
             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(',')]
+                machines = list(self._get_members(base_uri, **kwargs))
                 logger.debug("Retrieved list of machines: %s", machines)
-                random.shuffle(machines)
-                for url in machines:
-                    r = urlparse(url)
-                    port = r.port or (443 if r.scheme == 'https' else 80)
-                    self._dns_resolver.resolve_async(r.hostname, port)
-                return machines
+                if machines:
+                    random.shuffle(machines)
+                    self._update_dns_cache(self._dns_resolver.resolve_async, machines)
+                    return machines
             except Exception as e:
-                # We can't get the list of machines, if one server is in the
-                # machines cache, try on it
-                logger.error("Failed to get list of machines from %s%s: %r", self._base_uri, self.version_prefix, e)
-                if self._machines_cache:
-                    self._base_uri = self._machines_cache.pop(0)
-                    logger.info("Retrying on %s", self._base_uri)
-                elif self._update_machines_cache:
-                    raise etcd.EtcdException("Could not get the list of servers, "
-                                             "maybe you provided the wrong "
-                                             "host(s) to connect to?")
-                else:
-                    return []
+                self.http.clear()
+                logger.error("Failed to get list of machines from %s%s: %r", base_uri, self.version_prefix, e)
+
+        raise etcd.EtcdConnectionFailed('No more machines in the cluster')
 
     def set_read_timeout(self, timeout):
         self._read_timeout = timeout
 
-    def _do_http_request(self, request_executor, method, url, fields=None, **kwargs):
-        try:
-            response = request_executor(method, url, fields=fields, **kwargs)
-            response.data.decode('utf-8')
-            self._check_cluster_id(response)
-        except (HTTPError, HTTPException, socket.error, socket.timeout) as e:
-            if (isinstance(fields, dict) and fields.get("wait") == "true" and
-                    isinstance(e, ReadTimeoutError)):
-                logger.debug("Watch timed out.")
-                raise etcd.EtcdWatchTimedOut("Watch timed out: {0}".format(e), cause=e)
-            logger.error("Request to server %s failed: %r", self._base_uri, e)
-            logger.info("Reconnection allowed, looking for another server.")
-            self._base_uri = self._next_server(cause=e)
-            response = False
-        return response
+    def _do_http_request(self, retry, machines_cache, request_executor, method, path, fields=None, **kwargs):
+        if fields is not None:
+            kwargs['fields'] = fields
+        some_request_failed = False
+        for i, base_uri in enumerate(machines_cache):
+            if i > 0:
+                logger.info("Retrying on %s", base_uri)
+            try:
+                response = request_executor(method, base_uri + path, **kwargs)
+                response.data.decode('utf-8')
+                if some_request_failed:
+                    self.set_base_uri(base_uri)
+                    self._refresh_machines_cache()
+                return response
+            except (HTTPError, HTTPException, socket.error, socket.timeout) as e:
+                self.http.clear()
+                # switch to the next etcd node because we don't know exactly what happened,
+                # whether the key didn't received an update or there is a network problem.
+                if not retry and i + 1 < len(machines_cache):
+                    self.set_base_uri(machines_cache[i + 1])
+                if (isinstance(fields, dict) and fields.get("wait") == "true" and
+                        isinstance(e, (ReadTimeoutError, ProtocolError))):
+                    logger.debug("Watch timed out.")
+                    raise etcd.EtcdWatchTimedOut("Watch timed out: {0}".format(e), cause=e)
+                logger.error("Request to server %s failed: %r", base_uri, e)
+                logger.info("Reconnection allowed, looking for another server.")
+                if not retry:
+                    raise etcd.EtcdException('{0} {1} request failed'.format(method, path))
+                some_request_failed = True
+
+        raise etcd.EtcdConnectionFailed('No more machines in the cluster')
+
+    @abc.abstractmethod
+    def _prepare_request(self, kwargs, params=None, method=None):
+        """returns: request_executor"""
 
     def api_execute(self, path, method, params=None, timeout=None):
-        if not path.startswith('/'):
-            raise ValueError('Path does not start with /')
-
-        kwargs = {'fields': params, 'preload_content': False}
-
-        if method in [self._MGET, self._MDELETE]:
-            request_executor = self.http.request
-        elif method in [self._MPUT, self._MPOST]:
-            request_executor = self.http.request_encode_body
-            kwargs['encode_multipart'] = False
-        else:
-            raise etcd.EtcdException('HTTP method {0} not supported'.format(method))
+        retry = params.pop('retry', None) if isinstance(params, dict) else None
 
         # 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:
-            self._machines_cache = self.machines
-            if self._base_uri in self._machines_cache:
-                self._machines_cache.remove(self._base_uri)
-            self._machines_cache_updated = time.time()
+        elif not self._use_proxies and time.time() - self._machines_cache_updated > self._machines_cache_ttl:
+            self._refresh_machines_cache()
 
-        kwargs.update(self._build_request_parameters())
+        machines_cache = self.machines_cache
+        etcd_nodes = len(machines_cache)
 
-        if timeout is not None:
-            kwargs.update({'retries': 0, 'timeout': timeout})
+        kwargs = self._prepare_common_parameters(etcd_nodes, timeout)
+        request_executor = self._prepare_request(kwargs, params, method)
 
-        response = False
-
-        try:
-            some_request_failed = False
-            while not response:
-                response = self._do_http_request(request_executor, method, self._base_uri + path, **kwargs)
-
-                if response is False:
-                    some_request_failed = True
-            if some_request_failed and not self._use_proxies:
-                self._machines_cache = self.machines
-                if self._base_uri in self._machines_cache:
-                    self._machines_cache.remove(self._base_uri)
-        except etcd.EtcdConnectionFailed as e:
-            if isinstance(e, etcd.EtcdWatchTimedOut) and self._machines_cache:
-                self._base_uri = self._next_server()
-            else:
-                self._update_machines_cache = True
-            if not response:
+        while True:
+            try:
+                response = self._do_http_request(retry, machines_cache, request_executor, method, path, **kwargs)
+                return self._handle_server_response(response)
+            except etcd.EtcdWatchTimedOut:
                 raise
-        return self._handle_server_response(response)
+            except etcd.EtcdConnectionFailed as ex:
+                try:
+                    if self._load_machines_cache():
+                        machines_cache = self.machines_cache
+                        etcd_nodes = len(machines_cache)
+                except Exception as e:
+                    logger.debug('Failed to update list of etcd nodes: %r', e)
+                sleeptime = retry.sleeptime
+                remaining_time = retry.stoptime - sleeptime - time.time()
+                nodes, timeout, retries = self._calculate_timeouts(etcd_nodes, remaining_time)
+                if nodes == 0:
+                    self._update_machines_cache = True
+                    raise ex
+                retry.sleep_func(sleeptime)
+                retry.update_delay()
+                # We still have some time left. Partially reduce `machines_cache` and retry request
+                kwargs.update(timeout=Timeout(connect=max(1, timeout/2), total=timeout), retries=retries)
+                machines_cache = machines_cache[:nodes]
 
     @staticmethod
     def get_srv_record(host):
@@ -236,15 +292,15 @@ 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 = uri(protocol, host, port, endpoint)
+                url = uri(protocol, (host, port), endpoint)
                 if endpoint:
                     try:
-                        response = requests.get(url, timeout=self.read_timeout, verify=False)
-                        if response.ok:
-                            for member in response.json():
+                        response = requests_get(url, timeout=self.read_timeout, verify=False)
+                        if response.status < 400:
+                            for member in json.loads(response.data.decode('utf-8')):
                                 ret.extend(member['clientURLs'])
                             break
-                    except RequestException:
+                    except Exception:
                         logger.exception('GET %s', url)
                 else:
                     ret.append(url)
@@ -258,15 +314,32 @@ 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 = []
-            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(uri(self.protocol, host, port))
+            ret = map(lambda res: uri(self.protocol, res[-1][:2]), self._dns_resolver.resolve(host, port))
             if ret:
                 return list(set(ret))
-        return [uri(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
+
+    @staticmethod
+    def _update_dns_cache(func, machines):
+        for url in machines:
+            r = urlparse(url)
+            port = r.port or (443 if r.scheme == 'https' else 80)
+            func(r.hostname, port)
 
     def _load_machines_cache(self):
         """This method should fill up `_machines_cache` from scratch.
@@ -279,50 +352,95 @@ class Client(etcd.Client):
         if 'srv' not in self._config and 'host' not in self._config and 'hosts' not in self._config:
             raise Exception('Neither srv, hosts, host nor url are defined in etcd section of config')
 
-        if self._use_proxies:
-            self._machines_cache = [uri(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 'hosts' in self._config:
-                self._machines_cache = list(self._config['hosts'])
-
-            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'])
-
+        machines_cache = self._get_machines_cache_from_config()
         # Can not bootstrap list of etcd-cluster members, giving up
-        if not self._machines_cache:
+        if not machines_cache:
             raise etcd.EtcdException
 
-        # 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
+        # enforce resolving dns name,they might get new ips
+        self._update_dns_cache(self._dns_resolver.remove, machines_cache)
 
-        if self._base_uri in self._machines_cache:
-            self._machines_cache.remove(self._base_uri)
+        # The etcd cluster could change its topology over time and depending on how we resolve the initial
+        # topology (list of hosts in the Patroni config or DNS records, A or SRV) we might get into the situation
+        # the the real topology doesn't match anymore with the topology resolved from the configuration file.
+        # In case if the "initial" topology is the same as before we will not override the `_machines_cache`.
+        ret = set(machines_cache) != set(self._initial_machines_cache)
+        if ret:
+            self._initial_machines_cache = self._machines_cache = machines_cache
+
+            # After filling up the initial list of machines_cache we should ask etcd-cluster about actual list
+            self._refresh_machines_cache(True)
 
         self._update_machines_cache = False
+        return ret
+
+    def _refresh_machines_cache(self, updating_cache=False):
+        if self._use_proxies:
+            self._machines_cache = self._get_machines_cache_from_config()
+        else:
+            try:
+                self._machines_cache = self.machines
+            except etcd.EtcdConnectionFailed:
+                if updating_cache:
+                    raise etcd.EtcdException("Could not get the list of servers, "
+                                             "maybe you provided the wrong "
+                                             "host(s) to connect to?")
+                return
+
+        if self._base_uri not in self._machines_cache:
+            self.set_base_uri(self._machines_cache[0])
         self._machines_cache_updated = time.time()
 
+    def set_base_uri(self, value):
+        logger.info('Selected new etcd server %s', value)
+        self._base_uri = value
 
-class Etcd(AbstractDCS):
 
-    def __init__(self, config):
-        super(Etcd, self).__init__(config)
-        self._ttl = int(config.get('ttl') or 30)
+class EtcdClient(AbstractEtcdClientWithFailover):
+
+    ERROR_CLS = EtcdError
+
+    def __del__(self):
+        if self.http is not None:
+            try:
+                self.http.clear()
+            except (ReferenceError, TypeError, AttributeError):
+                pass
+
+    def _prepare_get_members(self, etcd_nodes):
+        return self._prepare_common_parameters(etcd_nodes)
+
+    def _get_members(self, base_uri, **kwargs):
+        response = self.http.request(self._MGET, base_uri + self.version_prefix + '/machines', **kwargs)
+        data = self._handle_server_response(response).data.decode('utf-8')
+        return [m.strip() for m in data.split(',') if m.strip()]
+
+    def _prepare_request(self, kwargs, params=None, method=None):
+        kwargs['fields'] = params
+        if method in (self._MPOST, self._MPUT):
+            kwargs['encode_multipart'] = False
+        return self.http.request
+
+
+class AbstractEtcd(AbstractDCS):
+
+    def __init__(self, config, client_cls, retry_errors_cls):
+        super(AbstractEtcd, self).__init__(config)
         self._retry = Retry(deadline=config['retry_timeout'], max_delay=1, max_tries=-1,
-                            retry_exceptions=(etcd.EtcdLeaderElectionInProgress,
-                                              etcd.EtcdWatcherCleared,
-                                              etcd.EtcdEventIndexCleared))
-        self._client = self.get_etcd_client(config)
+                            retry_exceptions=retry_errors_cls)
+        self._ttl = int(config.get('ttl') or 30)
+        self._client = self.get_etcd_client(config, client_cls)
         self.__do_not_watch = False
         self._has_failed = False
 
+    def reload_config(self, config):
+        super(AbstractEtcd, self).reload_config(config)
+        self._client.reload_config(config.get(self.__class__.__name__.lower(), {}))
+
     def retry(self, *args, **kwargs):
-        return self._retry.copy()(*args, **kwargs)
+        retry = self._retry.copy()
+        kwargs['retry'] = retry
+        return retry(*args, **kwargs)
 
     def _handle_exception(self, e, name='', do_sleep=False, raise_ex=None):
         if not self._has_failed:
@@ -335,22 +453,13 @@ class Etcd(AbstractDCS):
         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):
+    def set_socket_options(sock, socket_options):
+        if socket_options:
+            for opt in socket_options:
+                sock.setsockopt(*opt)
+
+    def get_etcd_client(self, config, client_cls):
         if 'proxy' in config:
             config['use_proxies'] = True
             config['url'] = config['proxy']
@@ -370,7 +479,7 @@ class Etcd(AbstractDCS):
             config['hosts'] = []
             for value in hosts:
                 if isinstance(value, six.string_types):
-                    config['hosts'].append(uri(protocol, *split_host_port(value, default_port)))
+                    config['hosts'].append(uri(protocol, split_host_port(value.strip(), default_port)))
         elif 'host' in config:
             host, port = split_host_port(config['host'], 2379)
             config['host'] = host
@@ -399,9 +508,7 @@ class Etcd(AbstractDCS):
                 sock = None
                 try:
                     sock = socket.socket(af, socktype, proto)
-                    if socket_options:
-                        for opt in socket_options:
-                            sock.setsockopt(*opt)
+                    self.set_socket_options(sock, socket_options)
                     if timeout is not socket._GLOBAL_DEFAULT_TIMEOUT:
                         sock.settimeout(timeout)
                     if source_address:
@@ -425,7 +532,9 @@ class Etcd(AbstractDCS):
         client = None
         while not client:
             try:
-                client = Client(config, dns_resolver)
+                client = client_cls(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)
@@ -433,22 +542,53 @@ class Etcd(AbstractDCS):
 
     def set_ttl(self, ttl):
         ttl = int(ttl)
-        self.__do_not_watch = self._ttl != ttl
+        ret = self._ttl != ttl
         self._ttl = ttl
         self._client.set_machines_cache_ttl(ttl*10)
+        return ret
+
+    @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)
 
+
+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=self._client.ERROR_CLS('unexpected error'))
+
+    return wrapper
+
+
+class Etcd(AbstractEtcd):
+
+    def __init__(self, config):
+        super(Etcd, self).__init__(config, EtcdClient, (etcd.EtcdLeaderElectionInProgress, EtcdRaftInternal))
+        self.__do_not_watch = False
+
+    def set_ttl(self, ttl):
+        self.__do_not_watch = super(Etcd, self).set_ttl(ttl)
+
     @staticmethod
     def member(node):
         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)
@@ -486,21 +626,22 @@ 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, history)
+            cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
         except etcd.EtcdKeyNotFound:
-            self._cluster = Cluster(None, None, None, None, [], None, None, None)
+            cluster = Cluster(None, None, None, None, [], None, None, None)
         except Exception as e:
             self._handle_exception(e, 'get_cluster', raise_ex=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):
         data = json.dumps(data, separators=(',', ':'))
-        return self.retry(self._client.set, self.member_path, data, None if permanent else ttl or self._ttl)
+        return self._client.set(self.member_path, data, None if permanent else self._ttl)
 
     @catch_etcd_errors
     def take_leader(self):
-        return self.retry(self._client.set, self.leader_path, self._name, self._ttl)
+        return self.retry(self._client.write, self.leader_path, self._name, ttl=self._ttl)
 
     def attempt_to_acquire_leader(self, permanent=False):
         try:
@@ -529,14 +670,14 @@ class Etcd(AbstractDCS):
 
     @catch_etcd_errors
     def _update_leader(self):
-        return self.retry(self._client.test_and_set, self.leader_path, self._name, self._name, self._ttl)
+        return self.retry(self._client.write, self.leader_path, self._name, prevValue=self._name, ttl=self._ttl)
 
     @catch_etcd_errors
     def initialize(self, create_new=True, sysid=""):
         return self.retry(self._client.write, self.initialize_path, sysid, prevExist=(not create_new))
 
     @catch_etcd_errors
-    def delete_leader(self):
+    def _delete_leader(self):
         return self._client.delete(self.leader_path, prevValue=self._name)
 
     @catch_etcd_errors
@@ -569,13 +710,14 @@ 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)
+                    result = self._client.watch(self.leader_path, index=leader_index, timeout=timeout + 0.5)
                     self._has_failed = False
+                    if result.action == 'compareAndSwap':
+                        time.sleep(0.01)
                     # 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
@@ -590,3 +732,6 @@ class Etcd(AbstractDCS):
             return super(Etcd, self).watch(None, timeout)
         finally:
             self.event.clear()
+
+
+etcd.EtcdError.error_exceptions[300] = EtcdRaftInternal

patroni/dcs/etcd3.py

@@ -0,0 +1,799 @@
+from __future__ import absolute_import
+import base64
+import etcd
+import json
+import logging
+import os
+import six
+import socket
+import sys
+import time
+import urllib3
+
+from threading import Condition, Lock, Thread
+
+from . import ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
+from .etcd import AbstractEtcdClientWithFailover, AbstractEtcd, catch_etcd_errors
+from ..exceptions import DCSError, PatroniException
+from ..utils import deep_compare, enable_keepalive, iter_response_objects, RetryFailedError, USER_AGENT
+
+logger = logging.getLogger(__name__)
+
+
+class Etcd3Error(DCSError):
+    pass
+
+
+class UnsupportedEtcdVersion(PatroniException):
+    pass
+
+
+# google.golang.org/grpc/codes
+GRPCCode = type('Enum', (), {'OK': 0, 'Canceled': 1, 'Unknown': 2, 'InvalidArgument': 3, 'DeadlineExceeded': 4,
+                             'NotFound': 5, 'AlreadyExists': 6, 'PermissionDenied': 7, 'ResourceExhausted': 8,
+                             'FailedPrecondition': 9, 'Aborted': 10, 'OutOfRange': 11, 'Unimplemented': 12,
+                             'Internal': 13, 'Unavailable': 14, 'DataLoss': 15, 'Unauthenticated': 16})
+GRPCcodeToText = {v: k for k, v in GRPCCode.__dict__.items() if not k.startswith('__') and isinstance(v, int)}
+
+
+class Etcd3Exception(etcd.EtcdException):
+    pass
+
+
+class Etcd3ClientError(Etcd3Exception):
+
+    def __init__(self, code=None, error=None, status=None):
+        if not hasattr(self, 'error'):
+            self.error = error and error.strip()
+        self.codeText = GRPCcodeToText.get(code)
+        self.status = status
+
+    def __repr__(self):
+        return "<{0} error: '{1}', code: {2}>".format(self.__class__.__name__, self.error, self.code)
+
+    __str__ = __repr__
+
+    def as_dict(self):
+        return {'error': self.error, 'code': self.code, 'codeText': self.codeText, 'status': self.status}
+
+    @classmethod
+    def get_subclasses(cls):
+        for subclass in cls.__subclasses__():
+            for subsubclass in subclass.get_subclasses():
+                yield subsubclass
+            yield subclass
+
+
+class Unknown(Etcd3ClientError):
+    code = GRPCCode.Unknown
+
+
+class InvalidArgument(Etcd3ClientError):
+    code = GRPCCode.InvalidArgument
+
+
+class DeadlineExceeded(Etcd3ClientError):
+    code = GRPCCode.DeadlineExceeded
+    error = "context deadline exceeded"
+
+
+class NotFound(Etcd3ClientError):
+    code = GRPCCode.NotFound
+
+
+class FailedPrecondition(Etcd3ClientError):
+    code = GRPCCode.FailedPrecondition
+
+
+class Unavailable(Etcd3ClientError):
+    code = GRPCCode.Unavailable
+
+
+# https://github.com/etcd-io/etcd/blob/master/etcdserver/api/v3rpc/rpctypes/error.go
+class LeaseNotFound(NotFound):
+    error = "etcdserver: requested lease not found"
+
+
+class UserEmpty(InvalidArgument):
+    error = "etcdserver: user name is empty"
+
+
+class AuthFailed(InvalidArgument):
+    error = "etcdserver: authentication failed, invalid user ID or password"
+
+
+class PermissionDenied(Etcd3ClientError):
+    code = GRPCCode.PermissionDenied
+    error = "etcdserver: permission denied"
+
+
+class AuthNotEnabled(FailedPrecondition):
+    error = "etcdserver: authentication is not enabled"
+
+
+class InvalidAuthToken(Etcd3ClientError):
+    code = GRPCCode.Unauthenticated
+    error = "etcdserver: invalid auth token"
+
+
+errStringToClientError = {s.error: s for s in Etcd3ClientError.get_subclasses() if hasattr(s, 'error')}
+errCodeToClientError = {s.code: s for s in Etcd3ClientError.__subclasses__()}
+
+
+def _raise_for_data(data, status_code=None):
+    try:
+        error = data.get('error') or data.get('Error')
+        if isinstance(error, dict):  # streaming response
+            status_code = error.get('http_code')
+            code = error['grpc_code']
+            error = error['message']
+        else:
+            code = data.get('code') or data.get('Code')
+    except Exception:
+        error = str(data)
+        code = GRPCCode.Unknown
+    err = errStringToClientError.get(error) or errCodeToClientError.get(code) or Unknown
+    raise err(code, error, status_code)
+
+
+def to_bytes(v):
+    return v if isinstance(v, bytes) else v.encode('utf-8')
+
+
+def prefix_range_end(v):
+    v = bytearray(to_bytes(v))
+    for i in range(len(v) - 1, -1, -1):
+        if v[i] < 0xff:
+            v[i] += 1
+            break
+    return bytes(v)
+
+
+def base64_encode(v):
+    return base64.b64encode(to_bytes(v)).decode('utf-8')
+
+
+def base64_decode(v):
+    return base64.b64decode(v).decode('utf-8')
+
+
+def build_range_request(key, range_end=None):
+    fields = {'key': base64_encode(key)}
+    if range_end:
+        fields['range_end'] = base64_encode(range_end)
+    return fields
+
+
+class Etcd3Client(AbstractEtcdClientWithFailover):
+
+    ERROR_CLS = Etcd3Error
+
+    def __init__(self, config, dns_resolver, cache_ttl=300):
+        self._token = None
+        self._cluster_version = None
+        self.version_prefix = '/v3beta'
+        super(Etcd3Client, self).__init__(config, dns_resolver, cache_ttl)
+
+        if six.PY2:  # pragma: no cover
+            # Old grpc-gateway sometimes sends double 'transfer-encoding: chunked' headers,
+            # what breaks the old (python2.7) httplib.HTTPConnection (it closes the socket).
+            def dedup_addheader(httpm, key, value):
+                prev = httpm.dict.get(key)
+                if prev is None:
+                    httpm.dict[key] = value
+                elif key != 'transfer-encoding' or prev != value:
+                    combined = ", ".join((prev, value))
+                    httpm.dict[key] = combined
+
+            import httplib
+            httplib.HTTPMessage.addheader = dedup_addheader
+
+        try:
+            self.authenticate()
+        except AuthFailed as e:
+            logger.fatal('Etcd3 authentication failed: %r', e)
+            sys.exit(1)
+
+    def _get_headers(self):
+        headers = urllib3.make_headers(user_agent=USER_AGENT)
+        if self._token and self._cluster_version >= (3, 3, 0):
+            headers['authorization'] = self._token
+        return headers
+
+    def _prepare_request(self, kwargs, params=None, method=None):
+        if params is not None:
+            kwargs['body'] = json.dumps(params)
+            kwargs['headers']['Content-Type'] = 'application/json'
+        return self.http.urlopen
+
+    @staticmethod
+    def _handle_server_response(response):
+        data = response.data
+        try:
+            data = data.decode('utf-8')
+            data = json.loads(data)
+        except (TypeError, ValueError, UnicodeError) as e:
+            if response.status < 400:
+                raise etcd.EtcdException('Server response was not valid JSON: %r' % e)
+        if response.status < 400:
+            return data
+        _raise_for_data(data, response.status)
+
+    def _ensure_version_prefix(self, base_uri, **kwargs):
+        if self.version_prefix != '/v3':
+            response = self.http.urlopen(self._MGET, base_uri + '/version', **kwargs)
+            response = self._handle_server_response(response)
+
+            server_version_str = response['etcdserver']
+            server_version = tuple(int(x) for x in server_version_str.split('.'))
+            cluster_version_str = response['etcdcluster']
+            self._cluster_version = tuple(int(x) for x in cluster_version_str.split('.'))
+
+            if self._cluster_version < (3, 0) or server_version < (3, 0, 4):
+                raise UnsupportedEtcdVersion('Detected Etcd version {0} is lower than 3.0.4'.format(server_version_str))
+
+            if self._cluster_version < (3, 3):
+                if self.version_prefix != '/v3alpha':
+                    if self._cluster_version < (3, 1):
+                        logger.warning('Detected Etcd version %s is lower than 3.1.0, watches are not supported',
+                                       cluster_version_str)
+                    if self.username and self.password:
+                        logger.warning('Detected Etcd version %s is lower than 3.3.0, authentication is not supported',
+                                       cluster_version_str)
+                    self.version_prefix = '/v3alpha'
+            elif self._cluster_version < (3, 4):
+                self.version_prefix = '/v3beta'
+            else:
+                self.version_prefix = '/v3'
+
+    def _prepare_get_members(self, etcd_nodes):
+        kwargs = self._prepare_common_parameters(etcd_nodes)
+        self._prepare_request(kwargs, {})
+        return kwargs
+
+    def _get_members(self, base_uri, **kwargs):
+        self._ensure_version_prefix(base_uri, **kwargs)
+        resp = self.http.urlopen(self._MPOST, base_uri + self.version_prefix + '/cluster/member/list', **kwargs)
+        members = self._handle_server_response(resp)['members']
+        return set(url for member in members for url in member.get('clientURLs', []))
+
+    def call_rpc(self, method, fields, retry=None):
+        fields['retry'] = retry
+        return self.api_execute(self.version_prefix + method, self._MPOST, fields)
+
+    def authenticate(self):
+        if self._cluster_version >= (3, 3) and self.username and self.password:
+            logger.info('Trying to authenticate on Etcd...')
+            old_token, self._token = self._token, None
+            try:
+                response = self.call_rpc('/auth/authenticate', {'name': self.username, 'password': self.password})
+            except AuthNotEnabled:
+                logger.info('Etcd authentication is not enabled')
+                self._token = None
+            except Exception:
+                self._token = old_token
+                raise
+            else:
+                self._token = response.get('token')
+            return old_token != self._token
+
+    def _handle_auth_errors(func):
+        def wrapper(self, *args, **kwargs):
+            def retry(ex):
+                if self.username and self.password:
+                    self.authenticate()
+                    return func(self, *args, **kwargs)
+                else:
+                    logger.fatal('Username or password not set, authentication is not possible')
+                    raise ex
+
+            try:
+                return func(self, *args, **kwargs)
+            except (UserEmpty, PermissionDenied) as e:  # no token provided
+                # PermissionDenied is raised on 3.0 and 3.1
+                if self._cluster_version < (3, 3) and (not isinstance(e, PermissionDenied)
+                                                       or self._cluster_version < (3, 2)):
+                    raise UnsupportedEtcdVersion('Authentication is required by Etcd cluster but not '
+                                                 'supported on version lower than 3.3.0. Cluster version: '
+                                                 '{0}'.format('.'.join(map(str, self._cluster_version))))
+                return retry(e)
+            except InvalidAuthToken as e:
+                logger.error('Invalid auth token: %s', self._token)
+                return retry(e)
+
+        return wrapper
+
+    @_handle_auth_errors
+    def range(self, key, range_end=None, retry=None):
+        params = build_range_request(key, range_end)
+        params['serializable'] = True  # For better performance. We can tolerate stale reads.
+        return self.call_rpc('/kv/range', params, retry)
+
+    def prefix(self, key, retry=None):
+        return self.range(key, prefix_range_end(key), retry)
+
+    def lease_grant(self, ttl, retry=None):
+        return self.call_rpc('/lease/grant', {'TTL': ttl}, retry)['ID']
+
+    def lease_keepalive(self, ID, retry=None):
+        return self.call_rpc('/lease/keepalive', {'ID': ID}, retry).get('result', {}).get('TTL')
+
+    def txn(self, compare, success, retry=None):
+        return self.call_rpc('/kv/txn', {'compare': [compare], 'success': [success]}, retry).get('succeeded')
+
+    @_handle_auth_errors
+    def put(self, key, value, lease=None, create_revision=None, mod_revision=None, retry=None):
+        fields = {'key': base64_encode(key), 'value': base64_encode(value)}
+        if lease:
+            fields['lease'] = lease
+        if create_revision is not None:
+            compare = {'target': 'CREATE', 'create_revision': create_revision}
+        elif mod_revision is not None:
+            compare = {'target': 'MOD', 'mod_revision': mod_revision}
+        else:
+            return self.call_rpc('/kv/put', fields, retry)
+        compare['key'] = fields['key']
+        return self.txn(compare, {'request_put': fields}, retry)
+
+    @_handle_auth_errors
+    def deleterange(self, key, range_end=None, mod_revision=None, retry=None):
+        fields = build_range_request(key, range_end)
+        if mod_revision is None:
+            return self.call_rpc('/kv/deleterange', fields, retry)
+        compare = {'target': 'MOD', 'mod_revision': mod_revision, 'key': fields['key']}
+        return self.txn(compare, {'request_delete_range': fields}, retry)
+
+    def deleteprefix(self, key, retry=None):
+        return self.deleterange(key, prefix_range_end(key), retry=retry)
+
+    def watchrange(self, key, range_end=None, start_revision=None, filters=None):
+        """returns: response object"""
+        params = build_range_request(key, range_end)
+        if start_revision is not None:
+            params['start_revision'] = start_revision
+        params['filters'] = filters or []
+        kwargs = self._prepare_common_parameters(1, self.read_timeout)
+        request_executor = self._prepare_request(kwargs, {'create_request': params})
+        kwargs.update(timeout=urllib3.Timeout(connect=kwargs['timeout']), retries=0)
+        return request_executor(self._MPOST, self._base_uri + self.version_prefix + '/watch', **kwargs)
+
+    def watchprefix(self, key, start_revision=None, filters=None):
+        return self.watchrange(key, prefix_range_end(key), start_revision, filters)
+
+
+class KVCache(Thread):
+
+    def __init__(self, dcs, client):
+        Thread.__init__(self)
+        self.daemon = True
+        self._dcs = dcs
+        self._client = client
+        self.condition = Condition()
+        self._config_key = base64_encode(dcs.config_path)
+        self._leader_key = base64_encode(dcs.leader_path)
+        self._optime_key = base64_encode(dcs.leader_optime_path)
+        self._name = base64_encode(dcs._name)
+        self._is_ready = False
+        self._response = None
+        self._response_lock = Lock()
+        self._object_cache = {}
+        self._object_cache_lock = Lock()
+        self.start()
+
+    def set(self, value, overwrite=False):
+        with self._object_cache_lock:
+            name = value['key']
+            old_value = self._object_cache.get(name)
+            ret = not old_value or int(old_value['mod_revision']) < int(value['mod_revision'])
+            if ret or overwrite and old_value['mod_revision'] == value['mod_revision']:
+                self._object_cache[name] = value
+        return ret, old_value
+
+    def delete(self, name, mod_revision):
+        with self._object_cache_lock:
+            old_value = self._object_cache.get(name)
+            ret = old_value and int(old_value['mod_revision']) < int(mod_revision)
+            if ret:
+                del self._object_cache[name]
+        return not old_value or ret, old_value
+
+    def copy(self):
+        with self._object_cache_lock:
+            return [v.copy() for v in self._object_cache.values()]
+
+    def get(self, name):
+        with self._object_cache_lock:
+            return self._object_cache.get(name)
+
+    def _process_event(self, event):
+        kv = event['kv']
+        key = kv['key']
+        if event.get('type') == 'DELETE':
+            success, old_value = self.delete(key, kv['mod_revision'])
+        else:
+            success, old_value = self.set(kv, True)
+
+        if success:
+            old_value = old_value and old_value.get('value')
+            new_value = kv.get('value')
+
+            value_changed = old_value != new_value and \
+                (key == self._leader_key or key == self._optime_key and new_value is not None or
+                 key == self._config_key and old_value is not None and new_value is not None)
+
+            if value_changed:
+                logger.debug('%s changed from %s to %s', key, old_value, new_value)
+
+            # We also want to wake up HA loop on replicas if leader optime was updated
+            if value_changed and (key != self._optime_key or self.get(self._leader_key) != self._name):
+                self._dcs.event.set()
+
+    def _process_message(self, message):
+        logger.debug('Received message: %s', message)
+        if 'error' in message:
+            _raise_for_data(message)
+        for event in message.get('result', {}).get('events', []):
+            self._process_event(event)
+
+    @staticmethod
+    def _finish_response(response):
+        try:
+            response.close()
+        finally:
+            response.release_conn()
+
+    def _do_watch(self, revision):
+        with self._response_lock:
+            self._response = None
+        response = self._client.watchprefix(self._dcs.cluster_prefix, revision)
+        with self._response_lock:
+            if self._response is None:
+                self._response = response
+
+        if not self._response:
+            return self._finish_response(response)
+
+        for message in iter_response_objects(response):
+            self._process_message(message)
+
+    def _build_cache(self):
+        result = self._dcs.retry(self._client.prefix, self._dcs.cluster_prefix)
+        with self._object_cache_lock:
+            self._object_cache = {node['key']: node for node in result.get('kvs', [])}
+        with self.condition:
+            self._is_ready = True
+            self.condition.notify()
+
+        try:
+            self._do_watch(result['header']['revision'])
+        except Exception as e:
+            logger.error('watchprefix failed: %r', e)
+        finally:
+            with self.condition:
+                self._is_ready = False
+            with self._response_lock:
+                response, self._response = self._response, None
+            if response:
+                self._finish_response(response)
+
+    def run(self):
+        while True:
+            try:
+                self._build_cache()
+            except Exception as e:
+                logger.error('KVCache.run %r', e)
+                time.sleep(1)
+
+    def kill_stream(self):
+        sock = None
+        with self._response_lock:
+            if self._response:
+                try:
+                    sock = self._response.connection.sock
+                except Exception:
+                    sock = None
+            else:
+                self._response = False
+        if sock:
+            try:
+                sock.shutdown(socket.SHUT_RDWR)
+                sock.close()
+            except Exception as e:
+                logger.debug('Error on socket.shutdown: %r', e)
+
+    def is_ready(self):
+        """Must be called only when holding the lock on `condition`"""
+        return self._is_ready
+
+
+class PatroniEtcd3Client(Etcd3Client):
+
+    def __init__(self, *args, **kwargs):
+        self._kv_cache = None
+        super(PatroniEtcd3Client, self).__init__(*args, **kwargs)
+
+    def configure(self, etcd3):
+        self._etcd3 = etcd3
+
+    def start_watcher(self):
+        if self._cluster_version >= (3, 1):
+            self._kv_cache = KVCache(self._etcd3, self)
+
+    def _restart_watcher(self):
+        if self._kv_cache:
+            self._kv_cache.kill_stream()
+
+    def set_base_uri(self, value):
+        super(PatroniEtcd3Client, self).set_base_uri(value)
+        self._restart_watcher()
+
+    def authenticate(self):
+        ret = super(PatroniEtcd3Client, self).authenticate()
+        if ret:
+            self._restart_watcher()
+        return ret
+
+    def _wait_cache(self, timeout):
+        stop_time = time.time() + timeout
+        while not self._kv_cache.is_ready():
+            timeout = stop_time - time.time()
+            if timeout <= 0:
+                raise RetryFailedError('Exceeded retry deadline')
+            self._kv_cache.condition.wait(timeout)
+
+    def get_cluster(self):
+        if self._kv_cache:
+            with self._kv_cache.condition:
+                self._wait_cache(self._etcd3._retry.deadline)
+                return self._kv_cache.copy()
+        else:
+            return self._etcd3.retry(self.prefix, self._etcd3.cluster_prefix).get('kvs', [])
+
+    def call_rpc(self, method, fields, retry=None):
+        ret = super(PatroniEtcd3Client, self).call_rpc(method, fields, retry)
+
+        if self._kv_cache:
+            value = delete = None
+            if method == '/kv/txn' and ret.get('succeeded'):
+                on_success = fields['success'][0]
+                value = on_success.get('request_put')
+                delete = on_success.get('request_delete_range')
+            elif method == '/kv/put' and ret:
+                value = fields
+            elif method == '/kv/deleterange' and ret:
+                delete = fields
+
+            if value:
+                value['mod_revision'] = ret['header']['revision']
+                self._kv_cache.set(value)
+            elif delete and 'range_end' not in delete:
+                self._kv_cache.delete(delete['key'], ret['header']['revision'])
+
+        return ret
+
+
+class Etcd3(AbstractEtcd):
+
+    def __init__(self, config):
+        super(Etcd3, self).__init__(config, PatroniEtcd3Client, (DeadlineExceeded, Unavailable, FailedPrecondition))
+        self.__do_not_watch = False
+        self._lease = None
+        self._last_lease_refresh = 0
+
+        self._client.configure(self)
+        if not self._ctl:
+            self._client.start_watcher()
+            self.create_lease()
+
+    def set_socket_options(self, sock, socket_options):
+        enable_keepalive(sock, self.ttl, int(self.loop_wait + self._retry.deadline))
+
+    def set_ttl(self, ttl):
+        self.__do_not_watch = super(Etcd3, self).set_ttl(ttl)
+        if self.__do_not_watch:
+            self._lease = None
+
+    def _do_refresh_lease(self, retry=None):
+        if self._lease and self._last_lease_refresh + self._loop_wait > time.time():
+            return False
+
+        if self._lease and not self._client.lease_keepalive(self._lease, retry):
+            self._lease = None
+
+        ret = not self._lease
+        if ret:
+            self._lease = self._client.lease_grant(self._ttl, retry)
+
+        self._last_lease_refresh = time.time()
+        return ret
+
+    def refresh_lease(self):
+        try:
+            return self.retry(self._do_refresh_lease)
+        except (Etcd3ClientError, RetryFailedError):
+            logger.exception('refresh_lease')
+        raise Etcd3Error('Failed ro keepalive/grant lease')
+
+    def create_lease(self):
+        while not self._lease:
+            try:
+                self.refresh_lease()
+            except Etcd3Error:
+                logger.info('waiting on etcd')
+                time.sleep(5)
+
+    @property
+    def cluster_prefix(self):
+        return self.client_path('')
+
+    @staticmethod
+    def member(node):
+        return Member.from_node(node['mod_revision'], os.path.basename(node['key']), node['lease'], node['value'])
+
+    def _load_cluster(self):
+        cluster = None
+        try:
+            path_len = len(self.cluster_prefix)
+
+            nodes = {}
+            for node in self._client.get_cluster():
+                node['key'] = base64_decode(node['key'])
+                node['value'] = base64_decode(node.get('value', ''))
+                node['lease'] = node.get('lease')
+                nodes[node['key'][path_len:].lstrip('/')] = node
+
+            # get initialize flag
+            initialize = nodes.get(self._INITIALIZE)
+            initialize = initialize and initialize['value']
+
+            # get global dynamic configuration
+            config = nodes.get(self._CONFIG)
+            config = config and ClusterConfig.from_node(config['mod_revision'], config['value'])
+
+            # get timeline history
+            history = nodes.get(self._HISTORY)
+            history = history and TimelineHistory.from_node(history['mod_revision'], 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'])
+
+            # get list of members
+            members = [self.member(n) for k, n in nodes.items() if k.startswith(self._MEMBERS) and k.count('/') == 1]
+
+            # get leader
+            leader = nodes.get(self._LEADER)
+            if not self._ctl and leader and leader['value'] == self._name and self._lease != leader.get('lease'):
+                logger.warning('I am the leader but not owner of the lease')
+
+            if leader:
+                member = Member(-1, leader['value'], None, {})
+                member = ([m for m in members if m.name == leader['value']] or [member])[0]
+                leader = Leader(leader['mod_revision'], leader['lease'], member)
+
+            # failover key
+            failover = nodes.get(self._FAILOVER)
+            if failover:
+                failover = Failover.from_node(failover['mod_revision'], failover['value'])
+
+            # get synchronization state
+            sync = nodes.get(self._SYNC)
+            sync = SyncState.from_node(sync and sync['mod_revision'], sync and sync['value'])
+
+            cluster = Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
+        except UnsupportedEtcdVersion:
+            raise
+        except Exception as e:
+            self._handle_exception(e, 'get_cluster', raise_ex=Etcd3Error('Etcd is not responding properly'))
+        self._has_failed = False
+        return cluster
+
+    @catch_etcd_errors
+    def touch_member(self, data, permanent=False):
+        if not permanent:
+            try:
+                self.refresh_lease()
+            except Etcd3Error:
+                return False
+
+        cluster = self.cluster
+        member = cluster and cluster.get_member(self._name, fallback_to_leader=False)
+
+        if member and member.session == self._lease and deep_compare(data, member.data):
+            return True
+
+        data = json.dumps(data, separators=(',', ':'))
+        try:
+            return self._client.put(self.member_path, data, None if permanent else self._lease)
+        except LeaseNotFound:
+            self._lease = None
+            logger.error('Our lease disappeared from Etcd, can not "touch_member"')
+
+    @catch_etcd_errors
+    def take_leader(self):
+        return self.retry(self._client.put, self.leader_path, self._name, self._lease)
+
+    @catch_etcd_errors
+    def _do_attempt_to_acquire_leader(self, permanent):
+        try:
+            return self.retry(self._client.put, self.leader_path, self._name, None if permanent else self._lease, 0)
+        except LeaseNotFound:
+            self._lease = None
+            logger.error('Our lease disappeared from Etcd. Will try to get a new one and retry attempt')
+            self.refresh_lease()
+            return self.retry(self._client.put, self.leader_path, self._name, None if permanent else self._lease, 0)
+
+    def attempt_to_acquire_leader(self, permanent=False):
+        if not self._lease and not permanent:
+            self.refresh_lease()
+
+        ret = self._do_attempt_to_acquire_leader(permanent)
+        if not ret:
+            logger.info('Could not take out TTL lock')
+        return ret
+
+    @catch_etcd_errors
+    def set_failover_value(self, value, index=None):
+        return self._client.put(self.failover_path, value, mod_revision=index)
+
+    @catch_etcd_errors
+    def set_config_value(self, value, index=None):
+        return self._client.put(self.config_path, value, mod_revision=index)
+
+    @catch_etcd_errors
+    def _write_leader_optime(self, last_operation):
+        return self._client.put(self.leader_optime_path, last_operation)
+
+    @catch_etcd_errors
+    def _update_leader(self):
+        if not self._lease:
+            self.refresh_lease()
+        elif self.retry(self._client.lease_keepalive, self._lease):
+            self._last_lease_refresh = time.time()
+
+        if self._lease:
+            cluster = self.cluster
+            leader_lease = cluster and isinstance(cluster.leader, Leader) and cluster.leader.session
+            if leader_lease != self._lease:
+                self.take_leader()
+        return bool(self._lease)
+
+    @catch_etcd_errors
+    def initialize(self, create_new=True, sysid=""):
+        return self.retry(self._client.put, self.initialize_path, sysid, None, 0 if create_new else None)
+
+    @catch_etcd_errors
+    def _delete_leader(self):
+        cluster = self.cluster
+        if cluster and isinstance(cluster.leader, Leader) and cluster.leader.name == self._name:
+            return self._client.deleterange(self.leader_path, mod_revision=cluster.leader.index)
+
+    @catch_etcd_errors
+    def cancel_initialization(self):
+        return self.retry(self._client.deleterange, self.initialize_path)
+
+    @catch_etcd_errors
+    def delete_cluster(self):
+        return self.retry(self._client.deleteprefix, self.cluster_prefix)
+
+    @catch_etcd_errors
+    def set_history_value(self, value):
+        return self._client.put(self.history_path, value)
+
+    @catch_etcd_errors
+    def set_sync_state_value(self, value, index=None):
+        return self.retry(self._client.put, self.sync_path, value, mod_revision=index)
+
+    @catch_etcd_errors
+    def delete_sync_state(self, index=None):
+        return self.retry(self._client.deleterange, self.sync_path, mod_revision=index)
+
+    def watch(self, leader_index, timeout):
+        if self.__do_not_watch:
+            self.__do_not_watch = False
+            return True
+
+        try:
+            return super(Etcd3, self).watch(None, timeout)
+        finally:
+            self.event.clear()

patroni/dcs/exhibitor.py

@@ -1,10 +1,11 @@
+import json
 import logging
 import random
-import requests
 import time
 
 from patroni.dcs.zookeeper import ZooKeeper
-from requests.exceptions import RequestException
+from patroni.request import get as requests_get
+from patroni.utils import uri
 
 logger = logging.getLogger(__name__)
 
@@ -46,12 +47,11 @@ 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)
-                return response.json()
-            except RequestException:
-                pass
+                response = requests_get(uri('http', (host, self._exhibitor_port), self._uri_path), timeout=self.TIMEOUT)
+                return json.loads(response.data.decode('utf-8'))
+            except Exception:
+                logging.debug('Request to %s failed', host)
         return None
 
     @property

patroni/dcs/raft.py

@@ -0,0 +1,422 @@
+import json
+import logging
+import os
+import threading
+import time
+
+from patroni.dcs import AbstractDCS, ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
+from ..utils import validate_directory
+from pysyncobj import SyncObj, SyncObjConf, replicated, FAIL_REASON
+from pysyncobj.transport import Node, TCPTransport, CONNECTION_STATE
+
+logger = logging.getLogger(__name__)
+
+
+class MessageNode(Node):
+
+    def __init__(self, address):
+        self.address = address
+
+
+class UtilityTransport(TCPTransport):
+
+    def __init__(self, syncObj, selfNode, otherNodes):
+        super(UtilityTransport, self).__init__(syncObj, selfNode, otherNodes)
+        self._selfIsReadonlyNode = False
+
+    def _connectIfNecessarySingle(self, node):
+        pass
+
+    def connectionState(self, node):
+        return self._connections[node].state
+
+    def isDisconnected(self, node):
+        return self.connectionState(node) == CONNECTION_STATE.DISCONNECTED
+
+    def connectIfRequiredSingle(self, node):
+        if self.isDisconnected(node):
+            return self._connections[node].connect(node.ip, node.port)
+
+    def disconnectSingle(self, node):
+        self._connections[node].disconnect()
+
+
+class SyncObjUtility(SyncObj):
+
+    def __init__(self, otherNodes, conf):
+        autoTick = conf.autoTick
+        conf.autoTick = False
+        super(SyncObjUtility, self).__init__(None, otherNodes, conf, transportClass=UtilityTransport)
+        conf.autoTick = autoTick
+        self._SyncObj__transport.setOnMessageReceivedCallback(self._onMessageReceived)
+        self.__result = None
+
+    def setPartnerNode(self, partner):
+        self.__node = partner
+
+    def sendMessage(self, message):
+        # Abuse the fact that node address is send as a first message
+        self._SyncObj__transport._selfNode = MessageNode(message)
+        self._SyncObj__transport.connectIfRequiredSingle(self.__node)
+        while not self._SyncObj__transport.isDisconnected(self.__node):
+            self._poller.poll(0.5)
+        return self.__result
+
+    def _onMessageReceived(self, _, message):
+        self.__result = message
+        self._SyncObj__transport.disconnectSingle(self.__node)
+
+
+class MyTCPTransport(TCPTransport):
+
+    def _onIncomingMessageReceived(self, conn, message):
+        if self._syncObj.encryptor and not conn.sendRandKey:
+            conn.sendRandKey = message
+            conn.recvRandKey = os.urandom(32)
+            conn.send(conn.recvRandKey)
+            return
+
+        # Utility messages
+        if isinstance(message, list) and message[0] == 'members':
+            conn.send(self._syncObj._get_members())
+            return True
+
+        return super(MyTCPTransport, self)._onIncomingMessageReceived(conn, message)
+
+
+class DynMemberSyncObj(SyncObj):
+
+    def __init__(self, selfAddress, partnerAddrs, conf):
+        add_self = False
+        utility = SyncObjUtility(partnerAddrs, conf)
+        for node in utility._SyncObj__otherNodes:
+            utility.setPartnerNode(node)
+            response = utility.sendMessage(['members'])
+            if response:
+                partnerAddrs = [member['addr'] for member in response if member['addr'] != selfAddress]
+                add_self = selfAddress and len(partnerAddrs) == len(response)
+                break
+
+        super(DynMemberSyncObj, self).__init__(selfAddress, partnerAddrs, conf, transportClass=MyTCPTransport)
+        if add_self:
+            threading.Thread(target=utility.sendMessage, args=(['add', selfAddress],)).start()
+
+    def _get_members(self):
+        ret = [{'addr': node.id, 'leader': node == self._getLeader(),
+                'status': CONNECTION_STATE.CONNECTED if node in self._SyncObj__connectedNodes
+                else CONNECTION_STATE.DISCONNECTED} for node in self._SyncObj__otherNodes]
+        ret.append({'addr': self._SyncObj__selfNode.id, 'leader': self._isLeader(),
+                    'status': CONNECTION_STATE.CONNECTED})
+        return ret
+
+    def _SyncObj__doChangeCluster(self, request, reverse=False):
+        ret = False
+        if not self._SyncObj__selfNode or request[0] != 'add' or reverse or request[1] != self._SyncObj__selfNode.id:
+            ret = super(DynMemberSyncObj, self)._SyncObj__doChangeCluster(request, reverse)
+            if ret:
+                self.forceLogCompaction()
+        return ret
+
+
+class KVStoreTTL(DynMemberSyncObj):
+
+    def __init__(self, selfAddress, partnerAddrs, conf, on_set=None, on_delete=None):
+        self.__on_set = on_set
+        self.__on_delete = on_delete
+        self.__limb = {}
+        self.__retry_timeout = None
+        self.__early_apply_local_log = selfAddress is not None
+        self.applied_local_log = False
+        super(KVStoreTTL, self).__init__(selfAddress, partnerAddrs, conf)
+        self.__data = {}
+
+    @staticmethod
+    def __check_requirements(old_value, **kwargs):
+        return ('prevExist' not in kwargs or bool(kwargs['prevExist']) == bool(old_value)) and \
+            ('prevValue' not in kwargs or old_value and old_value['value'] == kwargs['prevValue']) and \
+            (not kwargs.get('prevIndex') or old_value and old_value['index'] == kwargs['prevIndex'])
+
+    def set_retry_timeout(self, retry_timeout):
+        self.__retry_timeout = retry_timeout
+
+    def retry(self, func, *args, **kwargs):
+        event = threading.Event()
+        ret = {'result': None, 'error': -1}
+
+        def callback(result, error):
+            ret.update(result=result, error=error)
+            event.set()
+
+        kwargs['callback'] = callback
+        timeout = kwargs.pop('timeout', None) or self.__retry_timeout
+        deadline = timeout and time.time() + timeout
+
+        while True:
+            event.clear()
+            func(*args, **kwargs)
+            event.wait(timeout)
+            if ret['error'] == FAIL_REASON.SUCCESS:
+                return ret['result']
+            elif ret['error'] == FAIL_REASON.REQUEST_DENIED:
+                break
+            elif deadline:
+                timeout = deadline - time.time()
+                if timeout <= 0:
+                    break
+            time.sleep(1)
+        return False
+
+    @replicated
+    def _set(self, key, value, **kwargs):
+        old_value = self.__data.get(key, {})
+        if not self.__check_requirements(old_value, **kwargs):
+            return False
+
+        if old_value and old_value['created'] != value['created']:
+            value['created'] = value['updated']
+        value['index'] = self._SyncObj__raftLastApplied + 1
+
+        self.__data[key] = value
+        if self.__on_set:
+            self.__on_set(key, value)
+        return True
+
+    def set(self, key, value, ttl=None, **kwargs):
+        old_value = self.__data.get(key, {})
+        if not self.__check_requirements(old_value, **kwargs):
+            return False
+
+        value = {'value': value, 'updated': time.time()}
+        value['created'] = old_value.get('created', value['updated'])
+        if ttl:
+            value['expire'] = value['updated'] + ttl
+        return self.retry(self._set, key, value, **kwargs)
+
+    def __pop(self, key):
+        self.__data.pop(key)
+        if self.__on_delete:
+            self.__on_delete(key)
+
+    @replicated
+    def _delete(self, key, recursive=False, **kwargs):
+        if recursive:
+            for k in list(self.__data.keys()):
+                if k.startswith(key):
+                    self.__pop(k)
+        elif not self.__check_requirements(self.__data.get(key, {}), **kwargs):
+            return False
+        else:
+            self.__pop(key)
+        return True
+
+    def delete(self, key, recursive=False, **kwargs):
+        if not recursive and not self.__check_requirements(self.__data.get(key, {}), **kwargs):
+            return False
+        return self.retry(self._delete, key, recursive=recursive, **kwargs)
+
+    @staticmethod
+    def __values_match(old, new):
+        return all(old.get(n) == new.get(n) for n in ('created', 'updated', 'expire', 'value'))
+
+    @replicated
+    def _expire(self, key, value, callback=None):
+        current = self.__data.get(key)
+        if current and self.__values_match(current, value):
+            self.__pop(key)
+
+    def __expire_keys(self):
+        for key, value in self.__data.items():
+            if value and 'expire' in value and value['expire'] <= time.time() and \
+                    not (key in self.__limb and self.__values_match(self.__limb[key], value)):
+                self.__limb[key] = value
+
+                def callback(*args):
+                    if key in self.__limb and self.__values_match(self.__limb[key], value):
+                        self.__limb.pop(key)
+                self._expire(key, value, callback=callback)
+
+    def get(self, key, recursive=False):
+        if not recursive:
+            return self.__data.get(key)
+        return {k: v for k, v in self.__data.items() if k.startswith(key)}
+
+    def _onTick(self, timeToWait=0.0):
+        # The SyncObj starts applying the local log only when there is at least one node connected.
+        # We want to change this behavior and apply the local log even when there is nobody except us.
+        # It gives us at least some picture about the last known cluster state.
+        if self.__early_apply_local_log and not self.applied_local_log and self._SyncObj__needLoadDumpFile:
+            self._SyncObj__raftCommitIndex = self._SyncObj__getCurrentLogIndex()
+            self._SyncObj__raftCurrentTerm = self._SyncObj__getCurrentLogTerm()
+
+        super(KVStoreTTL, self)._onTick(timeToWait)
+
+        # The SyncObj calls onReady callback only when cluster got the leader and is ready for writes.
+        # In some cases for us it is safe to "signal" the Raft object when the local log is fully applied.
+        # We are using the `applied_local_log` property for that, but not calling the callback function.
+        if self.__early_apply_local_log and not self.applied_local_log and self._SyncObj__raftCommitIndex != 1 and \
+                self._SyncObj__raftLastApplied == self._SyncObj__raftCommitIndex:
+            self.applied_local_log = True
+
+        if self._isLeader():
+            self.__expire_keys()
+        else:
+            self.__limb.clear()
+
+
+class Raft(AbstractDCS):
+
+    def __init__(self, config):
+        super(Raft, self).__init__(config)
+        self._ttl = int(config.get('ttl') or 30)
+
+        self_addr = config.get('self_addr')
+        partner_addrs = config.get('partner_addrs', [])
+        if self._ctl:
+            if self_addr:
+                partner_addrs.append(self_addr)
+            self_addr = None
+
+        # Create raft data_dir if necessary
+        raft_data_dir = config.get('data_dir', '')
+        if raft_data_dir != '':
+            validate_directory(raft_data_dir)
+
+        ready_event = threading.Event()
+        file_template = os.path.join(config.get('data_dir', ''), (self_addr or ''))
+        conf = SyncObjConf(password=config.get('password'), appendEntriesUseBatch=False,
+                           bindAddress=config.get('bind_addr'), commandsWaitLeader=False,
+                           fullDumpFile=(file_template + '.dump' if self_addr else None),
+                           journalFile=(file_template + '.journal' if self_addr else None),
+                           onReady=ready_event.set, dynamicMembershipChange=True)
+
+        self._sync_obj = KVStoreTTL(self_addr, partner_addrs, conf, self._on_set, self._on_delete)
+        while True:
+            ready_event.wait(5)
+            if ready_event.isSet() or self._sync_obj.applied_local_log:
+                break
+            else:
+                logger.info('waiting on raft')
+        self._sync_obj.forceLogCompaction()
+        self.set_retry_timeout(int(config.get('retry_timeout') or 10))
+
+    def _on_set(self, key, value):
+        leader = (self._sync_obj.get(self.leader_path) or {}).get('value')
+        if key == value['created'] == value['updated'] and \
+                (key.startswith(self.members_path) or key == self.leader_path and leader != self._name) or \
+                key == self.leader_optime_path and leader != self._name or key in (self.config_path, self.sync_path):
+            self.event.set()
+
+    def _on_delete(self, key):
+        if key == self.leader_path:
+            self.event.set()
+
+    def set_ttl(self, ttl):
+        self._ttl = ttl
+
+    @property
+    def ttl(self):
+        return self._ttl
+
+    def set_retry_timeout(self, retry_timeout):
+        self._sync_obj.set_retry_timeout(retry_timeout)
+
+    @staticmethod
+    def member(key, value):
+        return Member.from_node(value['index'], os.path.basename(key), None, value['value'])
+
+    def _load_cluster(self):
+        prefix = self.client_path('')
+        response = self._sync_obj.get(prefix, recursive=True)
+        if not response:
+            return Cluster(None, None, None, None, [], None, None, None)
+        nodes = {os.path.relpath(key, prefix).replace('\\', '/'): value for key, value in response.items()}
+
+        # get initialize flag
+        initialize = nodes.get(self._INITIALIZE)
+        initialize = initialize and initialize['value']
+
+        # get global dynamic configuration
+        config = nodes.get(self._CONFIG)
+        config = config and ClusterConfig.from_node(config['index'], config['value'])
+
+        # get timeline history
+        history = nodes.get(self._HISTORY)
+        history = history and TimelineHistory.from_node(history['index'], 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'])
+
+        # get list of members
+        members = [self.member(k, n) for k, n in nodes.items() if k.startswith(self._MEMBERS) and k.count('/') == 1]
+
+        # get leader
+        leader = nodes.get(self._LEADER)
+        if leader:
+            member = Member(-1, leader['value'], None, {})
+            member = ([m for m in members if m.name == leader['value']] or [member])[0]
+            leader = Leader(leader['index'], None, member)
+
+        # failover key
+        failover = nodes.get(self._FAILOVER)
+        if failover:
+            failover = Failover.from_node(failover['index'], failover['value'])
+
+        # get synchronization state
+        sync = nodes.get(self._SYNC)
+        sync = SyncState.from_node(sync and sync['index'], sync and sync['value'])
+
+        return Cluster(initialize, config, leader, last_leader_operation, members, failover, sync, history)
+
+    def _write_leader_optime(self, last_operation):
+        return self._sync_obj.set(self.leader_optime_path, last_operation, timeout=1)
+
+    def _update_leader(self):
+        ret = self._sync_obj.set(self.leader_path, self._name, ttl=self._ttl, prevValue=self._name)
+        if not ret and self._sync_obj.get(self.leader_path) is None:
+            ret = self.attempt_to_acquire_leader()
+        return ret
+
+    def attempt_to_acquire_leader(self, permanent=False):
+        return self._sync_obj.set(self.leader_path, self._name, prevExist=False,
+                                  ttl=None if permanent else self._ttl)
+
+    def set_failover_value(self, value, index=None):
+        return self._sync_obj.set(self.failover_path, value, prevIndex=index)
+
+    def set_config_value(self, value, index=None):
+        return self._sync_obj.set(self.config_path, value, prevIndex=index)
+
+    def touch_member(self, data, permanent=False):
+        data = json.dumps(data, separators=(',', ':'))
+        return self._sync_obj.set(self.member_path, data, None if permanent else self._ttl, timeout=2)
+
+    def take_leader(self):
+        return self._sync_obj.set(self.leader_path, self._name, ttl=self._ttl)
+
+    def initialize(self, create_new=True, sysid=''):
+        return self._sync_obj.set(self.initialize_path, sysid, prevExist=(not create_new))
+
+    def _delete_leader(self):
+        return self._sync_obj.delete(self.leader_path, prevValue=self._name, timeout=1)
+
+    def cancel_initialization(self):
+        return self._sync_obj.delete(self.initialize_path)
+
+    def delete_cluster(self):
+        return self._sync_obj.delete(self.client_path(''), recursive=True)
+
+    def set_history_value(self, value):
+        return self._sync_obj.set(self.history_path, value)
+
+    def set_sync_state_value(self, value, index=None):
+        return self._sync_obj.set(self.sync_path, value, prevIndex=index)
+
+    def delete_sync_state(self, index=None):
+        return self._sync_obj.delete(self.sync_path, prevIndex=index)
+
+    def watch(self, leader_index, timeout):
+        try:
+            return super(Raft, self).watch(leader_index, timeout)
+        finally:
+            self.event.clear()

patroni/dcs/zookeeper.py

@@ -1,5 +1,6 @@
 import json
 import logging
+import select
 import time
 
 from kazoo.client import KazooClient, KazooState, KazooRetry
@@ -37,12 +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):
 
@@ -53,21 +63,25 @@ class ZooKeeper(AbstractDCS):
         if isinstance(hosts, list):
             hosts = ','.join(hosts)
 
+        mapping = {'use_ssl': 'use_ssl', 'verify': 'verify_certs', 'cacert': 'ca',
+                   'cert': 'certfile', 'key': 'keyfile', 'key_password': 'keyfile_password'}
+        kwargs = {v: config[k] for k, v in mapping.items() if k in config}
+
         self._client = KazooClient(hosts, handler=PatroniSequentialThreadingHandler(config['retry_timeout']),
                                    timeout=config['ttl'], connection_retry=KazooRetry(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))
+                                   max_delay=1, max_tries=-1, sleep_func=time.sleep), **kwargs)
         self._client.add_listener(self.session_listener)
 
-        self._my_member_data = {}
         self._fetch_cluster = True
+        self._fetch_optime = True
 
         self._orig_kazoo_connect = self._client._connection._connect
         self._client._connection._connect = self._kazoo_connect
 
         self._client.start()
 
-    def _kazoo_connect(self, host, port):
+    def _kazoo_connect(self, *args):
         """Kazoo is using Ping's to determine health of connection to zookeeper. If there is no
         response on Ping after Ping interval (1/2 from read_timeout) it will consider current
         connection dead and try to connect to another node. Without this "magic" it was taking
@@ -79,16 +93,20 @@ class ZooKeeper(AbstractDCS):
         than loop_wait, because we can spend up to 2 seconds when calling `touch_member()` and
         `write_leader_optime()` methods, which also may hang..."""
 
-        ret = self._orig_kazoo_connect(host, port)
+        ret = self._orig_kazoo_connect(*args)
         return max(self.loop_wait - 2, 2)*1000, ret[1]
 
     def session_listener(self, state):
         if state in [KazooState.SUSPENDED, KazooState.LOST]:
             self.cluster_watcher(None)
 
+    def optime_watcher(self, event):
+        self._fetch_optime = True
+        self.event.set()
+
     def cluster_watcher(self, event):
         self._fetch_cluster = True
-        self.event.set()
+        self.optime_watcher(event)
 
     def reload_config(self, config):
         self.set_retry_timeout(config['retry_timeout'])
@@ -105,18 +123,23 @@ class ZooKeeper(AbstractDCS):
         # the same time, set_ttl method will reestablish connection and return
         # `!True`, otherwise we will close existing connection and let kazoo
         # open the new one.
-        if not self.set_ttl(int(config['ttl'] * 1000)) and loop_wait_changed:
+        if not self.set_ttl(config['ttl']) and loop_wait_changed:
             self._client._connection._socket.close()
 
     def set_ttl(self, ttl):
         """It is not possible to change ttl (session_timeout) in zookeeper without
         destroying old session and creating the new one. This method returns `!True`
         if session_timeout has been changed (`restart()` has been called)."""
+        ttl = int(ttl * 1000)
         if self._client._session_timeout != ttl:
             self._client._session_timeout = ttl
             self._client.restart()
             return True
 
+    @property
+    def ttl(self):
+        return self._client._session_timeout / 1000.0
+
     def set_retry_timeout(self, retry_timeout):
         retry = self._client.retry if isinstance(self._client.retry, KazooRetry) else self._client._retry
         retry.deadline = retry_timeout
@@ -128,6 +151,12 @@ class ZooKeeper(AbstractDCS):
         except NoNodeError:
             return None
 
+    def get_leader_optime(self, leader):
+        watch = self.optime_watcher if not leader or leader.name != self._name else None
+        optime = self.get_node(self.leader_optime_path, watch)
+        self._fetch_optime = False
+        return optime and int(optime[0]) or 0
+
     @staticmethod
     def member(name, value, znode):
         return Member.from_node(znode.version, name, znode.ephemeralOwner, value)
@@ -141,7 +170,7 @@ class ZooKeeper(AbstractDCS):
     def load_members(self, sync_standby):
         members = []
         for member in self.get_children(self.members_path, self.cluster_watcher):
-            watch = member == sync_standby and self.cluster_watcher or None
+            watch = member in sync_standby and self.cluster_watcher or None
             data = self.get_node(self.members_path + member, watch)
             if data is not None:
                 members.append(self.member(member, *data))
@@ -165,16 +194,12 @@ class ZooKeeper(AbstractDCS):
         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
-
         # get synchronization state
         sync = self.get_node(self.sync_path, watch=self.cluster_watcher) if self._SYNC in nodes else None
         sync = SyncState.from_node(sync and sync[1].version, sync and sync[0])
 
         # get list of members
-        sync_standby = sync.leader == self._name and sync.sync_standby or None
+        sync_standby = sync.leader == self._name and sync.members or []
         members = self.load_members(sync_standby) if self._MEMBERS[:-1] in nodes else []
 
         # get leader
@@ -193,66 +218,93 @@ class ZooKeeper(AbstractDCS):
                 leader = Leader(leader[1].version, leader[1].ephemeralOwner, member)
                 self._fetch_cluster = member.index == -1
 
+        # get last leader operation
+        last_leader_operation = self._OPTIME in nodes and self.get_leader_optime(leader)
+
         # failover key
         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, history)
+        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')
+        # Optime ZNode was updated or doesn't exist and we are not leader
+        elif (self._fetch_optime and not self._fetch_cluster or not cluster.last_leader_operation) and\
+                not (cluster.leader and cluster.leader.name == self._name):
+            try:
+                optime = self.get_leader_optime(cluster.leader)
+                cluster = Cluster(cluster.initialize, cluster.config, cluster.leader, optime,
+                                  cluster.members, cluster.failover, cluster.sync, cluster.history)
+            except Exception:
+                pass
+        return cluster
 
-    def _create(self, path, value, **kwargs):
+    def _bypass_caches(self):
+        self._fetch_cluster = True
+
+    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 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_or_sync_state_value(self, key, 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, key, 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(key, value))
+            if do_not_create_empty and not value:
+                return True
+            elif index is None:
+                return self._create(key, value, retry)
+            else:
+                return False
         except Exception:
-            logging.exception('set_failover_value')
-            return False
+            logger.exception('Failed to update %s', key)
+        return False
 
     def set_failover_value(self, value, index=None):
-        return self.__set_failover_or_sync_state_value(self.failover_path, value, index)
+        return self._set_or_create(self.failover_path, value, index)
 
     def set_config_value(self, value, index=None):
-        try:
-            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
+        return self._set_or_create(self.config_path, value, index, retry=True)
 
     def initialize(self, create_new=True, sysid=""):
-        return self._create(self.initialize_path, sysid, makepath=True) if create_new \
-            else self._client.retry(self._client.set, self.initialize_path, sysid.encode("utf-8"))
+        sysid = 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)
         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:
@@ -262,13 +314,12 @@ class ZooKeeper(AbstractDCS):
             member = None
 
         if member:
-            if deep_compare(data, self._my_member_data):
+            if deep_compare(data, member.data):
                 return True
         else:
             try:
                 self._client.create_async(self.member_path, encoded_data, makepath=True,
                                           ephemeral=not permanent).get(timeout=1)
-                self._my_member_data = data
                 return True
             except Exception as e:
                 if not isinstance(e, NodeExistsError):
@@ -276,7 +327,6 @@ class ZooKeeper(AbstractDCS):
                     return False
         try:
             self._client.set_async(self.member_path, encoded_data).get(timeout=1)
-            self._my_member_data = data
             return True
         except Exception:
             logger.exception('touch_member')
@@ -286,30 +336,14 @@ class ZooKeeper(AbstractDCS):
     def take_leader(self):
         return self.attempt_to_acquire_leader()
 
-    def __write_leader_optime_or_history_value(self, key, value):
-        value = value.encode('utf-8')
-        try:
-            self._client.set_async(key, value).get(timeout=1)
-            return True
-        except NoNodeError:
-            try:
-                self._client.create_async(key, value, makepath=True).get(timeout=1)
-                return True
-            except Exception:
-                logger.exception('Failed to create %s', key)
-        except Exception:
-            logger.exception('Failed to update %s', key)
-        return False
-
     def _write_leader_optime(self, last_operation):
-        return self.__write_leader_optime_or_history_value(self.leader_optime_path, last_operation)
+        return self._set_or_create(self.leader_optime_path, last_operation)
 
     def _update_leader(self):
         return True
 
-    def delete_leader(self):
+    def _delete_leader(self):
         self._client.restart()
-        self._my_member_data = None
         return True
 
     def _cancel_initialization(self):
@@ -330,15 +364,16 @@ class ZooKeeper(AbstractDCS):
             return True
 
     def set_history_value(self, value):
-        return self.__write_leader_optime_or_history_value(self.history_path, value)
+        return self._set_or_create(self.history_path, value)
 
     def set_sync_state_value(self, value, index=None):
-        return self.__set_failover_or_sync_state_value(self.sync_path, value, index)
+        return self._set_or_create(self.sync_path, value, index, retry=True, do_not_create_empty=True)
 
     def delete_sync_state(self, index=None):
         return self.set_sync_state_value("{}", index)
 
     def watch(self, leader_index, timeout):
-        if super(ZooKeeper, self).watch(leader_index, timeout):
+        ret = super(ZooKeeper, self).watch(leader_index, timeout)
+        if ret and not self._fetch_optime:
             self._fetch_cluster = True
-        return self._fetch_cluster
+        return ret or self._fetch_cluster

patroni/exceptions.py

@@ -13,6 +13,10 @@ class PatroniException(Exception):
         return repr(self.value)
 
 
+class PatroniFatalException(PatroniException):
+    pass
+
+
 class PostgresException(PatroniException):
     pass
 
@@ -27,3 +31,7 @@ class PostgresConnectionException(PostgresException):
 
 class WatchdogError(PatroniException):
     pass
+
+
+class ConfigParseError(PatroniException):
+    pass

patroni/log.py

@@ -0,0 +1,194 @@
+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__)
+
+
+def debug_exception(logger_obj, msg, *args, **kwargs):
+    kwargs.pop("exc_info", False)
+    if logger_obj.isEnabledFor(logging.DEBUG):
+        logger_obj.debug(msg, *args, exc_info=True, **kwargs)
+    else:
+        msg = "{0}, DETAIL: '{1}'".format(msg, sys.exc_info()[1])
+        logger_obj.error(msg, *args, exc_info=False, **kwargs)
+
+
+def error_exception(logger_obj, msg, *args, **kwargs):
+    exc_info = kwargs.pop("exc_info", True)
+    logger_obj.error(msg, *args, exc_info=exc_info, **kwargs)
+
+
+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 ProxyHandler(logging.Handler):
+
+    def __init__(self, patroni_logger):
+        logging.Handler.__init__(self)
+        self.patroni_logger = patroni_logger
+
+    def emit(self, record):
+        self.patroni_logger.log_handler.handle(record)
+
+
+class PatroniLogger(Thread):
+
+    DEFAULT_LEVEL = 'INFO'
+    DEFAULT_TRACEBACK_LEVEL = 'ERROR'
+    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._queue_handler = QueueHandler()
+        self._root_logger = logging.getLogger()
+        self._config = None
+        self.log_handler = None
+        self.log_handler_lock = Lock()
+        self._old_handlers = []
+        self.reload_config({'level': 'DEBUG'})
+        # We will switch to the QueueHandler only when thread was started.
+        # This is necessary to protect from the cases when Patroni constructor
+        # failed and PatroniLogger thread remain running and prevent shutdown.
+        self._proxy_handler = ProxyHandler(self)
+        self._root_logger.addHandler(self._proxy_handler)
+
+    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))
+            if config.get('traceback_level', PatroniLogger.DEFAULT_TRACEBACK_LEVEL).lower() == 'debug':
+                logging.Logger.exception = debug_exception
+            else:
+                logging.Logger.exception = error_exception
+
+            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):
+        # switch to QueueHandler only when the thread was started
+        with self.log_handler_lock:
+            self._root_logger.addHandler(self._queue_handler)
+            self._root_logger.removeHandler(self._proxy_handler)
+
+        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/bootstrap.py

@@ -0,0 +1,381 @@
+import logging
+import os
+import shlex
+import tempfile
+import time
+
+from patroni.dcs import RemoteMember
+from patroni.utils import deep_compare
+from six import string_types
+
+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 = [] if config.get('no_params') else ['--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
+            connstring = self._postgresql.config.format_dsn(r, True)
+            if 'host' not in r:
+                # 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
+
+            env = self._postgresql.config.write_pgpass(r)
+            env['PGOPTIONS'] = '-c synchronous_commit=local'
+
+            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)
+            # add the credentials to connect to the replica origin to pgpass.
+            env = self._postgresql.config.write_pgpass(r)
+            connstring = self._postgresql.config.format_dsn(r, True)
+        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('SET log_statement TO none')
+        self._postgresql.query('SET log_min_duration_statement TO -1')
+        self._postgresql.query("SET log_min_error_statement TO 'log'")
+        try:
+            self._postgresql.query(sql, *params)
+        finally:
+            self._postgresql.query('RESET log_min_error_statement')
+            self._postgresql.query('RESET log_min_duration_statement')
+            self._postgresql.query('RESET log_statement')
+
+    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'):
+                        if os.path.exists(postgresql.config.pg_hba_conf):
+                            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:
+                        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

@@ -0,0 +1,36 @@
+import logging
+
+from patroni.postgresql.cancellable import CancellableExecutor
+from threading import Condition, Thread
+
+logger = logging.getLogger(__name__)
+
+
+class CallbackExecutor(CancellableExecutor, Thread):
+
+    def __init__(self):
+        CancellableExecutor.__init__(self)
+        Thread.__init__(self)
+        self.daemon = True
+        self._cmd = None
+        self._condition = Condition()
+        self.start()
+
+    def call(self, cmd):
+        self._kill_process()
+        with self._condition:
+            self._cmd = cmd
+            self._condition.notify()
+
+    def run(self):
+        while True:
+            with self._condition:
+                if self._cmd is None:
+                    self._condition.wait()
+                cmd, self._cmd = self._cmd, None
+
+            with self._lock:
+                if not self._start_process(cmd, close_fds=True):
+                    continue
+            self._process.wait()
+            self._kill_children()

patroni/postgresql/cancellable.py

@@ -0,0 +1,133 @@
+import logging
+import psutil
+import subprocess
+
+from patroni.exceptions import PostgresException
+from patroni.utils import polling_loop
+from threading import Lock
+
+logger = logging.getLogger(__name__)
+
+
+class CancellableExecutor(object):
+
+    """
+    There must be only one such process so that AsyncExecutor can easily cancel it.
+    """
+
+    def __init__(self):
+        self._process = None
+        self._process_cmd = None
+        self._process_children = []
+        self._lock = Lock()
+
+    def _start_process(self, cmd, *args, **kwargs):
+        """This method must be executed only when the `_lock` is acquired"""
+
+        try:
+            self._process_children = []
+            self._process_cmd = cmd
+            self._process = psutil.Popen(cmd, *args, **kwargs)
+        except Exception:
+            return logger.exception('Failed to execute %s',  cmd)
+        return True
+
+    def _kill_process(self):
+        with self._lock:
+            if self._process is not None and self._process.is_running() and not self._process_children:
+                try:
+                    self._process.suspend()  # Suspend the process before getting list of childrens
+                except psutil.Error as e:
+                    logger.info('Failed to suspend the process: %s', e.msg)
+
+                try:
+                    self._process_children = self._process.children(recursive=True)
+                except psutil.Error:
+                    pass
+
+                try:
+                    self._process.kill()
+                    logger.warning('Killed %s because it was still running', self._process_cmd)
+                except psutil.NoSuchProcess:
+                    pass
+                except psutil.AccessDenied as e:
+                    logger.warning('Failed to kill the process: %s', e.msg)
+
+    def _kill_children(self):
+        waitlist = []
+        with self._lock:
+            for child in self._process_children:
+                try:
+                    child.kill()
+                except psutil.NoSuchProcess:
+                    continue
+                except psutil.AccessDenied as e:
+                    logger.info('Failed to kill child process: %s', e.msg)
+                waitlist.append(child)
+        psutil.wait_procs(waitlist)
+
+
+class CancellableSubprocess(CancellableExecutor):
+
+    def __init__(self):
+        super(CancellableSubprocess, self).__init__()
+        self._is_cancelled = False
+
+    def call(self, *args, **kwargs):
+        for s in ('stdin', 'stdout', 'stderr'):
+            kwargs.pop(s, None)
+
+        communicate = kwargs.pop('communicate', None)
+        if isinstance(communicate, dict):
+            input_data = communicate.get('input')
+            if input_data:
+                if input_data[-1] != '\n':
+                    input_data += '\n'
+                input_data = input_data.encode('utf-8')
+            kwargs['stdin'] = subprocess.PIPE
+            kwargs['stdout'] = subprocess.PIPE
+            kwargs['stderr'] = subprocess.PIPE
+
+        try:
+            with self._lock:
+                if self._is_cancelled:
+                    raise PostgresException('cancelled')
+
+                self._is_cancelled = False
+                started = self._start_process(*args, **kwargs)
+
+            if started:
+                if isinstance(communicate, dict):
+                    communicate['stdout'], communicate['stderr'] = self._process.communicate(input_data)
+                return self._process.wait()
+        finally:
+            with self._lock:
+                self._process = None
+            self._kill_children()
+
+    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, kill=False):
+        with self._lock:
+            self._is_cancelled = True
+            if self._process is None or not self._process.is_running():
+                return
+
+            logger.info('Terminating %s', self._process_cmd)
+            self._process.terminate()
+
+        for _ in polling_loop(10):
+            with self._lock:
+                if self._process is None or not self._process.is_running():
+                    return
+            if kill:
+                break
+
+        self._kill_process()

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

@@ -1,21 +1,43 @@
 import logging
+import multiprocessing
 import os
 import psutil
 import re
 import signal
 import subprocess
+import sys
 
-from patroni import call_self
+from patroni import PATRONI_ENV_PREFIX, KUBERNETES_ENV_PREFIX
+
+# avoid spawning the resource tracker process
+if sys.version_info >= (3, 8):  # pragma: no cover
+    import multiprocessing.resource_tracker
+    multiprocessing.resource_tracker.getfd = lambda: 0
+elif sys.version_info >= (3, 4):  # pragma: no cover
+    import multiprocessing.semaphore_tracker
+    multiprocessing.semaphore_tracker.getfd = lambda: 0
 
 logger = logging.getLogger(__name__)
 
 STOP_SIGNALS = {
-    'smart': signal.SIGTERM,
-    'fast': signal.SIGINT,
-    'immediate': signal.SIGQUIT,
+    'smart': 'TERM',
+    'fast': 'INT',
+    'immediate': 'QUIT',
 }
 
 
+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):
@@ -83,7 +105,43 @@ class PostmasterProcess(psutil.Process):
         except psutil.NoSuchProcess:
             return None
 
-    def signal_stop(self, mode):
+    def signal_kill(self):
+        """to suspend and kill postmaster and all children
+
+        :returns True if postmaster and children are killed, False if error
+        """
+        try:
+            self.suspend()
+        except psutil.NoSuchProcess:
+            return True
+        except psutil.Error as e:
+            logger.warning('Failed to suspend postmaster: %s', e)
+
+        try:
+            children = self.children(recursive=True)
+        except psutil.NoSuchProcess:
+            return True
+        except psutil.Error as e:
+            logger.warning('Failed to get a list of postmaster children: %s', e)
+            children = []
+
+        try:
+            self.kill()
+        except psutil.NoSuchProcess:
+            return True
+        except psutil.Error as e:
+            logger.warning('Could not kill postmaster: %s', e)
+            return False
+
+        for child in children:
+            try:
+                child.kill()
+            except psutil.Error:
+                pass
+        psutil.wait_procs(children + [self])
+        return True
+
+    def signal_stop(self, mode, pg_ctl='pg_ctl'):
         """Signal postmaster process to stop
 
         :returns None if signaled, True if process is already gone, False if error
@@ -91,8 +149,10 @@ class PostmasterProcess(psutil.Process):
         if self.is_single_user:
             logger.warning("Cannot stop server; single-user server is running (PID: {0})".format(self.pid))
             return False
+        if os.name != 'posix':
+            return self.pg_ctl_kill(mode, pg_ctl)
         try:
-            self.send_signal(STOP_SIGNALS[mode])
+            self.send_signal(getattr(signal, 'SIG' + STOP_SIGNALS[mode]))
         except psutil.NoSuchProcess:
             return True
         except psutil.AccessDenied as e:
@@ -101,29 +161,42 @@ class PostmasterProcess(psutil.Process):
 
         return None
 
+    def pg_ctl_kill(self, mode, pg_ctl):
+        try:
+            status = subprocess.call([pg_ctl, "kill", STOP_SIGNALS[mode], str(self.pid)])
+        except OSError:
+            return False
+        if status == 0:
+            return None
+        else:
+            return not self.is_running()
+
     def wait_for_user_backends_to_close(self):
-        # These regexps are cross checked against versions PostgreSQL 9.1 .. 9.6
-        aux_proc_re = re.compile("(?:postgres:)( .*:)? (?:""(?:startup|logger|checkpointer|writer|wal writer|"
-                                 "autovacuum launcher|autovacuum worker|stats collector|wal receiver|archiver|"
-                                 "wal sender) process|bgworker: )")
+        # 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:
-            user_backends = []
-            user_backends_cmdlines = []
-            for child in self.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")
+            children = self.children()
         except psutil.Error:
-            logger.exception('wait_for_user_backends_to_close')
+            return logger.debug('Failed to get list of postmaster children')
+
+        user_backends = []
+        user_backends_cmdlines = []
+        for child in children:
+            try:
+                cmdline = child.cmdline()
+                if cmdline and not aux_proc_re.match(cmdline[0]):
+                    user_backends.append(child)
+                    user_backends_cmdlines.append(cmdline[0])
+            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):
@@ -138,7 +211,9 @@ class PostmasterProcess(psutil.Process):
         # 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.
-        env = {p: os.environ[p] for p in ('PATH', 'LD_LIBRARY_PATH', 'LC_ALL', 'LANG') if p in os.environ}
+        # 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 os.environ if not p.startswith(
+            PATRONI_ENV_PREFIX) and not p.startswith(KUBERNETES_ENV_PREFIX)}
         try:
             proc = PostmasterProcess._from_pidfile(data_dir)
             if proc and not proc._is_postmaster_process():
@@ -153,12 +228,16 @@ class PostmasterProcess(psutil.Process):
                 env['PG_GRANDPARENT_PID'] = str(proc.pid)
         except psutil.NoSuchProcess:
             pass
-
-        proc = call_self(['pg_ctl_start', pgcommand, '-D', data_dir,
-                          '--config-file={}'.format(conf)] + options, close_fds=True,
-                         preexec_fn=os.setsid, stdout=subprocess.PIPE, env=env)
-        pid = int(proc.stdout.readline().strip())
-        proc.wait()
+        cmdline = [pgcommand, '-D', data_dir, '--config-file={}'.format(conf)] + options
+        logger.debug("Starting postgres: %s", " ".join(cmdline))
+        ctx = multiprocessing.get_context('spawn') if sys.version_info >= (3, 4) else multiprocessing
+        parent_conn, child_conn = ctx.Pipe(False)
+        proc = ctx.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

patroni/postgresql/rewind.py

@@ -0,0 +1,442 @@
+import logging
+import os
+import shlex
+import six
+import subprocess
+
+from threading import Lock, Thread
+
+from .connection import get_connection_cursor
+from .misc import parse_history, parse_lsn
+from ..async_executor import CriticalTask
+from ..dcs import Leader
+
+logger = logging.getLogger(__name__)
+
+REWIND_STATUS = type('Enum', (), {'INITIAL': 0, 'CHECKPOINT': 1, 'CHECK': 2, 'NEED': 3,
+                                  'NOT_NEED': 4, 'SUCCESS': 5, 'FAILED': 6})
+
+
+def format_lsn(lsn, full=False):
+    template = '{0:X}/{1:08X}' if full else '{0:X}/{1:X}'
+    return template.format(lsn >> 32, lsn & 0xFFFFFFFF)
+
+
+class Rewind(object):
+
+    def __init__(self, postgresql):
+        self._postgresql = postgresql
+        self._checkpoint_task_lock = Lock()
+        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
+
+    @staticmethod
+    def check_leader_is_not_in_recovery(conn_kwargs):
+        try:
+            with get_connection_cursor(connect_timeout=3, options='-c statement_timeout=2000', **conn_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_checkpoint_end(self, timeline, lsn):
+        """The checkpoint record size in WAL depends on postgres major version and platform (memory alignment).
+        Hence, the only reliable way to figure out where it ends, read the record from file with the help of pg_waldump
+        and parse the output. We are trying to read two records, and expect that it wil fail to read the second one:
+        `pg_waldump: fatal: error in WAL record at 0/182E220: invalid record length at 0/182E298: wanted 24, got 0`
+        The error message contains information about LSN of the next record, which is exactly where checkpoint ends."""
+
+        cmd = self._postgresql.pgcommand('pg_{0}dump'.format(self._postgresql.wal_name))
+        lsn8 = format_lsn(lsn, True)
+        lsn = format_lsn(lsn)
+        env = os.environ.copy()
+        env.update(LANG='C', LC_ALL='C', PGDATA=self._postgresql.data_dir)
+        try:
+            waldump = subprocess.Popen([cmd, '-t', str(timeline), '-s', lsn, '-n', '2'],
+                                       stdout=subprocess.PIPE, stderr=subprocess.PIPE, env=env)
+            out, err = waldump.communicate()
+            waldump.wait()
+        except Exception as e:
+            logger.error('Failed to execute `%s -t %s -s %s -n 2`: %r', cmd, timeline, lsn, e)
+        else:
+            out = out.decode('utf-8').rstrip().split('\n')
+            err = err.decode('utf-8').rstrip().split('\n')
+            pattern = 'error in WAL record at {0}: invalid record length at '.format(lsn)
+
+            if len(out) == 1 and len(err) == 1 and ', lsn: {0}, prev '.format(lsn8) in out[0] and pattern in err[0]:
+                i = err[0].find(pattern) + len(pattern)
+                j = err[0].find(": wanted ", i)
+                if j > -1:
+                    try:
+                        return parse_lsn(err[0][i:j])
+                    except Exception as e:
+                        logger.error('Failed to parse lsn %s: %r', err[0][i:j], e)
+            logger.error('Failed to parse `%s -t %s -s %s -n 2` output', cmd, timeline, lsn)
+            logger.error(' stdout=%s', '\n'.join(out))
+            logger.error(' stderr=%s', '\n'.join(err))
+
+        return 0
+
+    def _get_local_timeline_lsn_from_controldata(self):
+        in_recovery = timeline = lsn = None
+        data = self._postgresql.controldata()
+        try:
+            if data.get('Database cluster state') == 'shut down in recovery':
+                in_recovery = True
+                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':
+                in_recovery = False
+                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')
+
+        if lsn is not None:
+            try:
+                lsn = parse_lsn(lsn)
+            except (IndexError, ValueError) as e:
+                logger.error('Exception when parsing lsn %s: %r', lsn, e)
+                lsn = None
+
+        return in_recovery, timeline, lsn
+
+    def _get_local_timeline_lsn(self):
+        if self._postgresql.is_running():  # if postgres is running - get timeline from replication connection
+            in_recovery = True
+            timeline = self._postgresql.received_timeline() or self._postgresql.get_replica_timeline()
+            lsn = self._postgresql.replayed_location()
+        else:  # otherwise analyze pg_controldata output
+            in_recovery, timeline, lsn = self._get_local_timeline_lsn_from_controldata()
+
+        log_lsn = format_lsn(lsn) if isinstance(lsn, six.integer_types) else lsn
+        logger.info('Local timeline=%s lsn=%s', timeline, log_lsn)
+        return in_recovery, timeline, lsn
+
+    @staticmethod
+    def _log_master_history(history, i):
+        start = max(0, i - 3)
+        end = None if i + 4 >= len(history) else i + 2
+        history_show = []
+
+        def format_history_line(line):
+            return '{0}\t{1}\t{2}'.format(line[0], format_lsn(line[1]), line[2])
+
+        for line in history[start:end]:
+            history_show.append(format_history_line(line))
+
+        if line != history[-1]:
+            history_show.append('...')
+            history_show.append(format_history_line(history[-1]))
+
+        logger.info('master: history=%s', '\n'.join(history_show))
+
+    def _conn_kwargs(self, member, auth):
+        ret = member.conn_kwargs(auth)
+        if not ret.get('database'):
+            ret['database'] = self._postgresql.database
+        return ret
+
+    def _check_timeline_and_lsn(self, leader):
+        in_recovery, 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(self._conn_kwargs(leader, 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 local_timeline == master_timeline:
+                    need_rewind = False
+                elif master_timeline > 1:
+                    cur.execute('TIMELINE_HISTORY %s', (master_timeline,))
+                    history = bytes(cur.fetchone()[1]).decode('utf-8')
+                    logger.debug('master: history=%s', history)
+        except Exception:
+            return logger.exception('Exception when working with master via replication connection')
+
+        if history is not None:
+            history = list(parse_history(history))
+            for i, (parent_timeline, switchpoint, _) in enumerate(history):
+                if parent_timeline == local_timeline:
+                    # We don't need to rewind when:
+                    # 1. for replica: replayed location is not ahead of switchpoint
+                    # 2. for the former primary: end of checkpoint record is the same as switchpoint
+                    if in_recovery:
+                        need_rewind = local_lsn > switchpoint
+                    elif local_lsn >= switchpoint:
+                        need_rewind = True
+                    else:
+                        need_rewind = switchpoint != self._get_checkpoint_end(local_timeline, local_lsn)
+                    break
+                elif parent_timeline > local_timeline:
+                    break
+            self._log_master_history(history, i)
+
+        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 __checkpoint(self, task, wakeup):
+        try:
+            result = self._postgresql.checkpoint()
+        except Exception as e:
+            result = 'Exception: ' + str(e)
+        with task:
+            task.complete(not bool(result))
+            if task.result:
+                wakeup()
+
+    def ensure_checkpoint_after_promote(self, wakeup):
+        """After promote issue a CHECKPOINT from a new thread and asynchronously check the result.
+        In case if CHECKPOINT failed, just check that timeline in pg_control was updated."""
+
+        if self._state == REWIND_STATUS.INITIAL and self._postgresql.is_leader():
+            with self._checkpoint_task_lock:
+                if self._checkpoint_task:
+                    with self._checkpoint_task:
+                        if self._checkpoint_task.result:
+                            self._state = REWIND_STATUS.CHECKPOINT
+                        if self._checkpoint_task.result is not False:
+                            return
+                else:
+                    self._checkpoint_task = CriticalTask()
+                    return Thread(target=self.__checkpoint, args=(self._checkpoint_task, wakeup)).start()
+
+            if self._postgresql.get_master_timeline() == self._postgresql.pg_control_timeline():
+                self._state = REWIND_STATUS.CHECKPOINT
+
+    def checkpoint_after_promote(self):
+        return self._state == REWIND_STATUS.CHECKPOINT
+
+    def _fetch_missing_wal(self, restore_command, wal_filename):
+        cmd = ''
+        length = len(restore_command)
+        i = 0
+        while i < length:
+            if restore_command[i] == '%' and i + 1 < length:
+                i += 1
+                if restore_command[i] == 'p':
+                    cmd += os.path.join(self._postgresql.wal_dir, wal_filename)
+                elif restore_command[i] == 'f':
+                    cmd += wal_filename
+                elif restore_command[i] == 'r':
+                    cmd += '000000010000000000000001'
+                elif restore_command[i] == '%':
+                    cmd += '%'
+                else:
+                    cmd += '%'
+                    i -= 1
+            else:
+                cmd += restore_command[i]
+            i += 1
+
+        logger.info('Trying to fetch the missing wal: %s', cmd)
+        return self._postgresql.cancellable.call(shlex.split(cmd)) == 0
+
+    def _find_missing_wal(self, data):
+        # could not open file "$PGDATA/pg_wal/0000000A00006AA100000068": No such file or directory
+        pattern = 'could not open file "'
+        for line in data.decode('utf-8').split('\n'):
+            b = line.find(pattern)
+            if b > -1:
+                b += len(pattern)
+                e = line.find('": ', b)
+                if e > -1 and '/' in line[b:e]:
+                    waldir, wal_filename = line[b:e].rsplit('/', 1)
+                    if waldir.endswith('/pg_' + self._postgresql.wal_name) and len(wal_filename) == 24:
+                        return wal_filename
+
+    def pg_rewind(self, r):
+        # prepare pg_rewind connection
+        env = self._postgresql.config.write_pgpass(r)
+        env.update(LANG='C', LC_ALL='C', PGOPTIONS='-c statement_timeout=0')
+        dsn = self._postgresql.config.format_dsn(r, True)
+        logger.info('running pg_rewind from %s', dsn)
+
+        restore_command = self._postgresql.config.get('recovery_conf', {}).get('restore_command') \
+            if self._postgresql.major_version < 120000 else self._postgresql.get_guc_value('restore_command')
+
+        cmd = [self._postgresql.pgcommand('pg_rewind')]
+        if self._postgresql.major_version >= 130000 and restore_command:
+            cmd.append('--restore-target-wal')
+        cmd.extend(['-D', self._postgresql.data_dir, '--source-server', dsn])
+
+        while True:
+            results = {}
+            ret = self._postgresql.cancellable.call(cmd, env=env, communicate=results)
+
+            logger.info('pg_rewind exit code=%s', ret)
+            if ret is None:
+                return False
+
+            logger.info(' stdout=%s', results['stdout'].decode('utf-8'))
+            logger.info(' stderr=%s', results['stderr'].decode('utf-8'))
+            if ret == 0:
+                return True
+
+            if not restore_command or self._postgresql.major_version >= 130000:
+                return False
+
+            missing_wal = self._find_missing_wal(results['stderr']) or self._find_missing_wal(results['stdout'])
+            if not missing_wal:
+                return False
+
+            if not self._fetch_missing_wal(restore_command, missing_wal):
+                logger.info('Failed to fetch WAL segment %s required for pg_rewind', missing_wal)
+                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 = self._conn_kwargs(leader, 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(self._conn_kwargs(leader, 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
+        with self._checkpoint_task_lock:
+            self._checkpoint_task = None
+
+    @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
+
+    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._postgresql.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, communicate=None, options=None):
+        """run a given command in a single-user mode. If the command is empty - then just start and stop"""
+        cmd = [self._postgresql.pgcommand('postgres'), '--single', '-D', self._postgresql.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('template1')
+        return self._postgresql.cancellable.call(cmd, communicate=communicate)
+
+    def cleanup_archive_status(self):
+        status_dir = os.path.join(self._postgresql.wal_dir, '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 ensure_clean_shutdown(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._postgresql.config.remove_recovery_conf()
+        output = {}
+        ret = self.single_user_mode(communicate=output, options=opts)
+        if ret != 0:
+            logger.error('Crash recovery finished with code=%s', ret)
+            logger.info(' stdout=%s', output['stdout'].decode('utf-8'))
+            logger.info(' stderr=%s', output['stderr'].decode('utf-8'))
+        return ret == 0 or None

patroni/postgresql/slots.py

@@ -0,0 +1,111 @@
+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._replication_slots = {}  # already existing replication slots
+        self.schedule()
+
+    def _query(self, sql, *params):
+        return self._postgresql.query(sql, *params, retry=False)
+
+    def load_replication_slots(self):
+        if self._postgresql.major_version >= 90400 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 ignore_replication_slot(self, cluster, name):
+        slot = self._replication_slots[name]
+        for matcher in cluster.config.ignore_slots_matchers:
+            if ((matcher.get("name") is None or matcher["name"] == name)
+               and all(not matcher.get(a) or matcher[a] == slot.get(a) for a in ('database', 'plugin', 'type'))):
+                return True
+        return 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._postgresql.major_version >= 90400 and cluster.config:
+            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.ignore_replication_slot(cluster, name) and 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._postgresql.major_version >= 90400
+        self._schedule_load_slots = value

patroni/postgresql/validator.py

@@ -0,0 +1,518 @@
+import abc
+import logging
+import six
+
+from collections import namedtuple
+from urllib3.response import HTTPHeaderDict
+
+from ..utils import parse_bool, parse_int, parse_real
+
+logger = logging.getLogger(__name__)
+
+
+class CaseInsensitiveDict(HTTPHeaderDict):
+
+    def add(self, key, val):
+        self[key] = val
+
+    def __getitem__(self, key):
+        return self._container[key.lower()][1]
+
+    def __repr__(self):
+        return str(dict(self.items()))
+
+    def copy(self):
+        return CaseInsensitiveDict(self._container.values())
+
+
+class Bool(namedtuple('Bool', 'version_from,version_till')):
+
+    @staticmethod
+    def transform(name, value):
+        if parse_bool(value) is not None:
+            return value
+        logger.warning('Removing bool parameter=%s from the config due to the invalid value=%s', name, value)
+
+
+@six.add_metaclass(abc.ABCMeta)
+class Number(namedtuple('Number', 'version_from,version_till,min_val,max_val,unit')):
+
+    @staticmethod
+    @abc.abstractmethod
+    def parse(value, unit):
+        """parse value"""
+
+    def transform(self, name, value):
+        num_value = self.parse(value, self.unit)
+        if num_value is not None:
+            if num_value < self.min_val:
+                logger.warning('Value=%s of parameter=%s is too low, increasing to %s%s',
+                               value, name, self.min_val, self.unit or '')
+                return self.min_val
+            if num_value > self.max_val:
+                logger.warning('Value=%s of parameter=%s is too big, decreasing to %s%s',
+                               value, name, self.max_val, self.unit or '')
+                return self.max_val
+            return value
+        logger.warning('Removing %s parameter=%s from the config due to the invalid value=%s',
+                       self.__class__.__name__.lower(), name, value)
+
+
+class Integer(Number):
+
+    @staticmethod
+    def parse(value, unit):
+        return parse_int(value, unit)
+
+
+class Real(Number):
+
+    @staticmethod
+    def parse(value, unit):
+        return parse_real(value, unit)
+
+
+class Enum(namedtuple('Enum', 'version_from,version_till,possible_values')):
+
+    def transform(self, name, value):
+        if str(value).lower() in self.possible_values:
+            return value
+        logger.warning('Removing enum parameter=%s from the config due to the invalid value=%s', name, value)
+
+
+class EnumBool(Enum):
+
+    def transform(self, name, value):
+        if parse_bool(value) is not None:
+            return value
+        return super(EnumBool, self).transform(name, value)
+
+
+class String(namedtuple('String', 'version_from,version_till')):
+
+    @staticmethod
+    def transform(name, value):
+        return value
+
+
+# Format:
+#  key - parameter name
+#  value - tuple or multiple tuples if something was changing in GUC across postgres versions
+parameters = CaseInsensitiveDict({
+    'allow_system_table_mods': Bool(90300, None),
+    'application_name': String(90300, None),
+    'archive_command': String(90300, None),
+    'archive_mode': (
+        Bool(90300, 90500),
+        EnumBool(90500, None, ('always',))
+    ),
+    'archive_timeout': Integer(90300, None, 0, 1073741823, 's'),
+    'array_nulls': Bool(90300, None),
+    'authentication_timeout': Integer(90300, None,  1, 600, 's'),
+    'autovacuum': Bool(90300, None),
+    'autovacuum_analyze_scale_factor': Real(90300, None, 0, 100, None),
+    'autovacuum_analyze_threshold': Integer(90300, None, 0, 2147483647, None),
+    'autovacuum_freeze_max_age': Integer(90300, None, 100000, 2000000000, None),
+    'autovacuum_max_workers': (
+        Integer(90300, 90600, 1, 8388607, None),
+        Integer(90600, None, 1, 262143, None)
+    ),
+    'autovacuum_multixact_freeze_max_age': Integer(90300, None, 10000, 2000000000, None),
+    'autovacuum_naptime': Integer(90300, None, 1, 2147483, 's'),
+    'autovacuum_vacuum_cost_delay': (
+        Integer(90300, 120000, -1, 100, 'ms'),
+        Real(120000, None, -1, 100, 'ms')
+    ),
+    'autovacuum_vacuum_cost_limit': Integer(90300, None, -1, 10000, None),
+    'autovacuum_vacuum_insert_scale_factor': Real(130000, None, 0, 100, None),
+    'autovacuum_vacuum_insert_threshold': Integer(130000, None, -1, 2147483647, None),
+    'autovacuum_vacuum_scale_factor': Real(90300, None, 0, 100, None),
+    'autovacuum_vacuum_threshold': Integer(90300, None, 0, 2147483647, None),
+    'autovacuum_work_mem': Integer(90400, None, -1, 2147483647, 'kB'),
+    'backend_flush_after': Integer(90600, None, 0, 256, '8kB'),
+    'backslash_quote': EnumBool(90300, None, ('safe_encoding',)),
+    'backtrace_functions': String(130000, None),
+    'bgwriter_delay': Integer(90300, None, 10, 10000, 'ms'),
+    'bgwriter_flush_after': Integer(90600, None, 0, 256, '8kB'),
+    'bgwriter_lru_maxpages': (
+        Integer(90300, 100000, 0, 1000, None),
+        Integer(100000, None, 0, 1073741823, None)
+    ),
+    'bgwriter_lru_multiplier': Real(90300, None, 0, 10, None),
+    'bonjour': Bool(90300, None),
+    'bonjour_name': String(90300, None),
+    'bytea_output': Enum(90300, None, ('escape', 'hex')),
+    'check_function_bodies': Bool(90300, None),
+    'checkpoint_completion_target': Real(90300, None, 0, 1, None),
+    'checkpoint_flush_after': Integer(90600, None, 0, 256, '8kB'),
+    'checkpoint_segments': Integer(90300, 90500, 1, 2147483647, None),
+    'checkpoint_timeout': (
+        Integer(90300, 90600, 30, 3600, 's'),
+        Integer(90600, None, 30, 86400, 's')
+    ),
+    'checkpoint_warning': Integer(90300, None, 0, 2147483647, 's'),
+    'client_encoding': String(90300, None),
+    'client_min_messages': Enum(90300, None, ('debug5', 'debug4', 'debug3', 'debug2',
+                                              'debug1', 'log', 'notice', 'warning', 'error')),
+    'cluster_name': String(90500, None),
+    'commit_delay': Integer(90300, None, 0, 100000, None),
+    'commit_siblings': Integer(90300, None, 0, 1000, None),
+    'config_file': String(90300, None),
+    'constraint_exclusion': EnumBool(90300, None, ('partition',)),
+    'cpu_index_tuple_cost': Real(90300, None, 0, 1.79769e+308, None),
+    'cpu_operator_cost': Real(90300, None, 0, 1.79769e+308, None),
+    'cpu_tuple_cost': Real(90300, None, 0, 1.79769e+308, None),
+    'cursor_tuple_fraction': Real(90300, None, 0, 1, None),
+    'data_directory': String(90300, None),
+    'data_sync_retry': Bool(90400, None),
+    'DateStyle': String(90300, None),
+    'db_user_namespace': Bool(90300, None),
+    'deadlock_timeout': Integer(90300, None, 1, 2147483647, 'ms'),
+    'debug_pretty_print': Bool(90300, None),
+    'debug_print_parse': Bool(90300, None),
+    'debug_print_plan': Bool(90300, None),
+    'debug_print_rewritten': Bool(90300, None),
+    'default_statistics_target': Integer(90300, None, 1, 10000, None),
+    'default_table_access_method': String(120000, None),
+    'default_tablespace': String(90300, None),
+    'default_text_search_config': String(90300, None),
+    'default_transaction_deferrable': Bool(90300, None),
+    'default_transaction_isolation': Enum(90300, None, ('serializable', 'repeatable read',
+                                                        'read committed', 'read uncommitted')),
+    'default_transaction_read_only': Bool(90300, None),
+    'default_with_oids': Bool(90300, 120000),
+    'dynamic_library_path': String(90300, None),
+    'dynamic_shared_memory_type': (
+        Enum(90400, 120000, ('posix', 'sysv', 'mmap', 'none')),
+        Enum(120000, None, ('posix', 'sysv', 'mmap'))
+    ),
+    'effective_cache_size': Integer(90300, None, 1, 2147483647, '8kB'),
+    'effective_io_concurrency': Integer(90300, None, 0, 1000, None),
+    'enable_bitmapscan': Bool(90300, None),
+    'enable_gathermerge': Bool(100000, None),
+    'enable_hashagg': Bool(90300, None),
+    'enable_hashjoin': Bool(90300, None),
+    'enable_incremental_sort': Bool(130000, None),
+    'enable_indexonlyscan': Bool(90300, None),
+    'enable_indexscan': Bool(90300, None),
+    'enable_material': Bool(90300, None),
+    'enable_mergejoin': Bool(90300, None),
+    'enable_nestloop': Bool(90300, None),
+    'enable_parallel_append': Bool(110000, None),
+    'enable_parallel_hash': Bool(110000, None),
+    'enable_partition_pruning': Bool(110000, None),
+    'enable_partitionwise_aggregate': Bool(110000, None),
+    'enable_partitionwise_join': Bool(110000, None),
+    'enable_seqscan': Bool(90300, None),
+    'enable_sort': Bool(90300, None),
+    'enable_tidscan': Bool(90300, None),
+    'escape_string_warning': Bool(90300, None),
+    'event_source': String(90300, None),
+    'exit_on_error': Bool(90300, None),
+    'external_pid_file': String(90300, None),
+    'extra_float_digits': Integer(90300, None, -15, 3, None),
+    'force_parallel_mode': EnumBool(90600, None, ('regress',)),
+    'from_collapse_limit': Integer(90300, None, 1, 2147483647, None),
+    'fsync': Bool(90300, None),
+    'full_page_writes': Bool(90300, None),
+    'geqo': Bool(90300, None),
+    'geqo_effort': Integer(90300, None, 1, 10, None),
+    'geqo_generations': Integer(90300, None, 0, 2147483647, None),
+    'geqo_pool_size': Integer(90300, None, 0, 2147483647, None),
+    'geqo_seed': Real(90300, None, 0, 1, None),
+    'geqo_selection_bias': Real(90300, None, 1.5, 2, None),
+    'geqo_threshold': Integer(90300, None, 2, 2147483647, None),
+    'gin_fuzzy_search_limit': Integer(90300, None, 0, 2147483647, None),
+    'gin_pending_list_limit': Integer(90500, None, 64, 2147483647, 'kB'),
+    'hash_mem_multiplier': Real(130000, None, 1, 1000, None),
+    'hba_file': String(90300, None),
+    'hot_standby': Bool(90300, None),
+    'hot_standby_feedback': Bool(90300, None),
+    'huge_pages': EnumBool(90400, None, ('try',)),
+    'ident_file': String(90300, None),
+    'idle_in_transaction_session_timeout': Integer(90600, None, 0, 2147483647, 'ms'),
+    'ignore_checksum_failure': Bool(90300, None),
+    'ignore_invalid_pages': Bool(130000, None),
+    'ignore_system_indexes': Bool(90300, None),
+    'IntervalStyle': Enum(90300, None, ('postgres', 'postgres_verbose', 'sql_standard', 'iso_8601')),
+    'jit': Bool(110000, None),
+    'jit_above_cost': Real(110000, None, -1, 1.79769e+308, None),
+    'jit_debugging_support': Bool(110000, None),
+    'jit_dump_bitcode': Bool(110000, None),
+    'jit_expressions': Bool(110000, None),
+    'jit_inline_above_cost': Real(110000, None, -1, 1.79769e+308, None),
+    'jit_optimize_above_cost': Real(110000, None, -1, 1.79769e+308, None),
+    'jit_profiling_support': Bool(110000, None),
+    'jit_provider': String(110000, None),
+    'jit_tuple_deforming': Bool(110000, None),
+    'join_collapse_limit': Integer(90300, None, 1, 2147483647, None),
+    'krb_caseins_users': Bool(90300, None),
+    'krb_server_keyfile': String(90300, None),
+    'krb_srvname': String(90300, 90400),
+    'lc_messages': String(90300, None),
+    'lc_monetary': String(90300, None),
+    'lc_numeric': String(90300, None),
+    'lc_time': String(90300, None),
+    'listen_addresses': String(90300, None),
+    'local_preload_libraries': String(90300, None),
+    'lock_timeout': Integer(90300, None, 0, 2147483647, 'ms'),
+    'lo_compat_privileges': Bool(90300, None),
+    'log_autovacuum_min_duration': Integer(90300, None, -1, 2147483647, 'ms'),
+    'log_checkpoints': Bool(90300, None),
+    'log_connections': Bool(90300, None),
+    'log_destination': String(90300, None),
+    'log_directory': String(90300, None),
+    'log_disconnections': Bool(90300, None),
+    'log_duration': Bool(90300, None),
+    'log_error_verbosity': Enum(90300, None, ('terse', 'default', 'verbose')),
+    'log_executor_stats': Bool(90300, None),
+    'log_file_mode': Integer(90300, None, 0, 511, None),
+    'log_filename': String(90300, None),
+    'logging_collector': Bool(90300, None),
+    'log_hostname': Bool(90300, None),
+    'logical_decoding_work_mem': Integer(130000, None, 64, 2147483647, 'kB'),
+    'log_line_prefix': String(90300, None),
+    'log_lock_waits': Bool(90300, None),
+    'log_min_duration_sample': Integer(130000, None, -1, 2147483647, 'ms'),
+    'log_min_duration_statement': Integer(90300, None, -1, 2147483647, 'ms'),
+    'log_min_error_statement': Enum(90300, None, ('debug5', 'debug4', 'debug3', 'debug2', 'debug1', 'info',
+                                                  'notice', 'warning', 'error', 'log', 'fatal', 'panic')),
+    'log_min_messages': Enum(90300, None, ('debug5', 'debug4', 'debug3', 'debug2', 'debug1', 'info',
+                                           'notice', 'warning', 'error', 'log', 'fatal', 'panic')),
+    'log_parameter_max_length': Integer(130000, None, -1, 1073741823, 'B'),
+    'log_parameter_max_length_on_error': Integer(130000, None, -1, 1073741823, 'B'),
+    'log_parser_stats': Bool(90300, None),
+    'log_planner_stats': Bool(90300, None),
+    'log_replication_commands': Bool(90500, None),
+    'log_rotation_age': Integer(90300, None, 0, 35791394, 'min'),
+    'log_rotation_size': Integer(90300, None, 0, 2097151, 'kB'),
+    'log_statement': Enum(90300, None, ('none', 'ddl', 'mod', 'all')),
+    'log_statement_sample_rate': Real(130000, None, 0, 1, None),
+    'log_statement_stats': Bool(90300, None),
+    'log_temp_files': Integer(90300, None, -1, 2147483647, 'kB'),
+    'log_timezone': String(90300, None),
+    'log_transaction_sample_rate': Real(120000, None, 0, 1, None),
+    'log_truncate_on_rotation': Bool(90300, None),
+    'maintenance_io_concurrency': Integer(130000, None, 0, 1000, None),
+    'maintenance_work_mem': Integer(90300, None, 1024, 2147483647, 'kB'),
+    'max_connections': (
+        Integer(90300, 90600, 1, 8388607, None),
+        Integer(90600, None, 1, 262143, None)
+    ),
+    'max_files_per_process': (
+        Integer(90300, 130000, 25, 2147483647, None),
+        Integer(130000, None, 64, 2147483647, None)
+    ),
+    'max_locks_per_transaction': Integer(90300, None, 10, 2147483647, None),
+    'max_logical_replication_workers': Integer(100000, None, 0, 262143, None),
+    'max_parallel_maintenance_workers': Integer(110000, None, 0, 1024, None),
+    'max_parallel_workers': Integer(100000, None, 0, 1024, None),
+    'max_parallel_workers_per_gather': Integer(90600, None, 0, 1024, None),
+    'max_pred_locks_per_page': Integer(100000, None, 0, 2147483647, None),
+    'max_pred_locks_per_relation': Integer(100000, None, -2147483648, 2147483647, None),
+    'max_pred_locks_per_transaction': Integer(90300, None, 10, 2147483647, None),
+    'max_prepared_transactions': (
+        Integer(90300, 90600, 0, 8388607, None),
+        Integer(90600, None, 0, 262143, None)
+    ),
+    'max_replication_slots': (
+        Integer(90400, 90600, 0, 8388607, None),
+        Integer(90600, None, 0, 262143, None)
+    ),
+    'max_slot_wal_keep_size': Integer(130000, None, -1, 2147483647, 'MB'),
+    'max_stack_depth': Integer(90300, None, 100, 2147483647, 'kB'),
+    'max_standby_archive_delay': Integer(90300, None, -1, 2147483647, 'ms'),
+    'max_standby_streaming_delay': Integer(90300, None, -1, 2147483647, 'ms'),
+    'max_sync_workers_per_subscription': Integer(100000, None, 0, 262143, None),
+    'max_wal_senders': (
+        Integer(90300, 90600, 0, 8388607, None),
+        Integer(90600, None, 0, 262143, None)
+    ),
+    'max_wal_size': (
+        Integer(90500, 100000, 2, 2147483647, '16MB'),
+        Integer(100000, None, 2, 2147483647, 'MB')
+    ),
+    'max_worker_processes': (
+        Integer(90400, 90600, 1, 8388607, None),
+        Integer(90600, None, 0, 262143, None)
+    ),
+    'min_parallel_index_scan_size': Integer(100000, None, 0, 715827882, '8kB'),
+    'min_parallel_relation_size': Integer(90600, 100000, 0, 715827882, '8kB'),
+    'min_parallel_table_scan_size': Integer(100000, None, 0, 715827882, '8kB'),
+    'min_wal_size': (
+        Integer(90500, 100000, 2, 2147483647, '16MB'),
+        Integer(100000, None, 2, 2147483647, 'MB')
+    ),
+    'old_snapshot_threshold': Integer(90600, None, -1, 86400, 'min'),
+    'operator_precedence_warning': Bool(90500, None),
+    'parallel_leader_participation': Bool(110000, None),
+    'parallel_setup_cost': Real(90600, None, 0, 1.79769e+308, None),
+    'parallel_tuple_cost': Real(90600, None, 0, 1.79769e+308, None),
+    'password_encryption': (
+        Bool(90300, 100000),
+        Enum(100000, None, ('md5', 'scram-sha-256'))
+    ),
+    'plan_cache_mode': Enum(120000, None, ('auto', 'force_generic_plan', 'force_custom_plan')),
+    'port': Integer(90300, None, 1, 65535, None),
+    'post_auth_delay': Integer(90300, None, 0, 2147, 's'),
+    'pre_auth_delay': Integer(90300, None, 0, 60, 's'),
+    'quote_all_identifiers': Bool(90300, None),
+    'random_page_cost': Real(90300, None, 0, 1.79769e+308, None),
+    'replacement_sort_tuples': Integer(90600, 110000, 0, 2147483647, None),
+    'restart_after_crash': Bool(90300, None),
+    'row_security': Bool(90500, None),
+    'search_path': String(90300, None),
+    'seq_page_cost': Real(90300, None, 0, 1.79769e+308, None),
+    'session_preload_libraries': String(90400, None),
+    'session_replication_role': Enum(90300, None, ('origin', 'replica', 'local')),
+    'shared_buffers': Integer(90300, None, 16, 1073741823, '8kB'),
+    'shared_memory_type': Enum(120000, None, ('sysv', 'mmap')),
+    'shared_preload_libraries': String(90300, None),
+    'sql_inheritance': Bool(90300, 100000),
+    'ssl': Bool(90300, None),
+    'ssl_ca_file': String(90300, None),
+    'ssl_cert_file': String(90300, None),
+    'ssl_ciphers': String(90300, None),
+    'ssl_crl_file': String(90300, None),
+    'ssl_dh_params_file': String(100000, None),
+    'ssl_ecdh_curve': String(90400, None),
+    'ssl_key_file': String(90300, None),
+    'ssl_max_protocol_version': Enum(120000, None, ('', 'tlsv1', 'tlsv1.1', 'tlsv1.2', 'tlsv1.3')),
+    'ssl_min_protocol_version': Enum(120000, None, ('tlsv1', 'tlsv1.1', 'tlsv1.2', 'tlsv1.3')),
+    'ssl_passphrase_command': String(110000, None),
+    'ssl_passphrase_command_supports_reload': Bool(110000, None),
+    'ssl_prefer_server_ciphers': Bool(90400, None),
+    'ssl_renegotiation_limit': Integer(90300, 90500, 0, 2147483647, 'kB'),
+    'standard_conforming_strings': Bool(90300, None),
+    'statement_timeout': Integer(90300, None, 0, 2147483647, 'ms'),
+    'stats_temp_directory': String(90300, None),
+    'superuser_reserved_connections': (
+        Integer(90300, 90600, 0, 8388607, None),
+        Integer(90600, None, 0, 262143, None),
+    ),
+    'synchronize_seqscans': Bool(90300, None),
+    'synchronous_commit': (
+        EnumBool(90300, 90600, ('local', 'remote_write')),
+        EnumBool(90600, None, ('local', 'remote_write', 'remote_apply'))
+    ),
+    'synchronous_standby_names': String(90300, None),
+    'syslog_facility': Enum(90300, None, ('local0', 'local1', 'local2', 'local3',
+                                          'local4', 'local5', 'local6', 'local7')),
+    'syslog_ident': String(90300, None),
+    'syslog_sequence_numbers': Bool(90600, None),
+    'syslog_split_messages': Bool(90600, None),
+    'tcp_keepalives_count': Integer(90300, None, 0, 2147483647, None),
+    'tcp_keepalives_idle': Integer(90300, None, 0, 2147483647, 's'),
+    'tcp_keepalives_interval': Integer(90300, None, 0, 2147483647, 's'),
+    'tcp_user_timeout': Integer(120000, None, 0, 2147483647, 'ms'),
+    'temp_buffers': Integer(90300, None, 100, 1073741823, '8kB'),
+    'temp_file_limit': Integer(90300, None, -1, 2147483647, 'kB'),
+    'temp_tablespaces': String(90300, None),
+    'TimeZone': String(90300, None),
+    'timezone_abbreviations': String(90300, None),
+    'trace_notify': Bool(90300, None),
+    'trace_recovery_messages': Enum(90300, None, ('debug5', 'debug4', 'debug3', 'debug2',
+                                                  'debug1', 'log', 'notice', 'warning', 'error')),
+    'trace_sort': Bool(90300, None),
+    'track_activities': Bool(90300, None),
+    'track_activity_query_size': (
+        Integer(90300, 110000, 100, 102400, None),
+        Integer(110000, 130000, 100, 102400, 'B'),
+        Integer(130000, None, 100, 1048576, 'B')
+    ),
+    'track_commit_timestamp': Bool(90500, None),
+    'track_counts': Bool(90300, None),
+    'track_functions': Enum(90300, None, ('none', 'pl', 'all')),
+    'track_io_timing': Bool(90300, None),
+    'transaction_deferrable': Bool(90300, None),
+    'transaction_isolation': Enum(90300, None, ('serializable', 'repeatable read',
+                                                'read committed', 'read uncommitted')),
+    'transaction_read_only': Bool(90300, None),
+    'transform_null_equals': Bool(90300, None),
+    'unix_socket_directories': String(90300, None),
+    'unix_socket_group': String(90300, None),
+    'unix_socket_permissions': Integer(90300, None, 0, 511, None),
+    'update_process_title': Bool(90300, None),
+    'vacuum_cleanup_index_scale_factor': Real(110000, None, 0, 1e+10, None),
+    'vacuum_cost_delay': (
+        Integer(90300, 120000, 0, 100, 'ms'),
+        Real(120000, None, 0, 100, 'ms')
+    ),
+    'vacuum_cost_limit': Integer(90300, None, 1, 10000, None),
+    'vacuum_cost_page_dirty': Integer(90300, None, 0, 10000, None),
+    'vacuum_cost_page_hit': Integer(90300, None, 0, 10000, None),
+    'vacuum_cost_page_miss': Integer(90300, None, 0, 10000, None),
+    'vacuum_defer_cleanup_age': Integer(90300, None, 0, 1000000, None),
+    'vacuum_freeze_min_age': Integer(90300, None, 0, 1000000000, None),
+    'vacuum_freeze_table_age': Integer(90300, None, 0, 2000000000, None),
+    'vacuum_multixact_freeze_min_age': Integer(90300, None, 0, 1000000000, None),
+    'vacuum_multixact_freeze_table_age': Integer(90300, None, 0, 2000000000, None),
+    'wal_buffers': Integer(90300, None, -1, 262143, '8kB'),
+    'wal_compression': Bool(90500, None),
+    'wal_consistency_checking': String(100000, None),
+    'wal_init_zero': Bool(120000, None),
+    'wal_keep_segments': Integer(90300, 130000, 0, 2147483647, None),
+    'wal_keep_size': Integer(130000, None, 0, 2147483647, 'MB'),
+    'wal_level': (
+        Enum(90300, 90400, ('minimal', 'archive', 'hot_standby')),
+        Enum(90400, 90600, ('minimal', 'archive', 'hot_standby', 'logical')),
+        Enum(90600, None, ('minimal', 'replica', 'logical'))
+    ),
+    'wal_log_hints': Bool(90400, None),
+    'wal_receiver_create_temp_slot': Bool(130000, None),
+    'wal_receiver_status_interval': Integer(90300, None, 0, 2147483, 's'),
+    'wal_receiver_timeout': Integer(90300, None, 0, 2147483647, 'ms'),
+    'wal_recycle': Bool(120000, None),
+    'wal_retrieve_retry_interval': Integer(90500, None, 1, 2147483647, 'ms'),
+    'wal_sender_timeout': Integer(90300, None, 0, 2147483647, 'ms'),
+    'wal_skip_threshold': Integer(130000, None, 0, 2147483647, 'kB'),
+    'wal_sync_method': Enum(90300, None, ('fsync', 'fdatasync', 'open_sync', 'open_datasync')),
+    'wal_writer_delay': Integer(90300, None, 1, 10000, 'ms'),
+    'wal_writer_flush_after': Integer(90600, None, 0, 2147483647, '8kB'),
+    'work_mem': Integer(90300, None, 64, 2147483647, 'kB'),
+    'xmlbinary': Enum(90300, None, ('base64', 'hex')),
+    'xmloption': Enum(90300, None, ('content', 'document')),
+    'zero_damaged_pages': Bool(90300, None)
+})
+
+
+recovery_parameters = CaseInsensitiveDict({
+    'archive_cleanup_command': String(90300, None),
+    'pause_at_recovery_target': Bool(90300, 90500),
+    'primary_conninfo': String(90300, None),
+    'primary_slot_name': String(90400, None),
+    'promote_trigger_file': String(120000, None),
+    'recovery_end_command': String(90300, None),
+    'recovery_min_apply_delay': Integer(90400, None, 0, 2147483647, 'ms'),
+    'recovery_target': Enum(90400, None, ('immediate', '')),
+    'recovery_target_action': Enum(90500, None, ('pause', 'promote', 'shutdown')),
+    'recovery_target_inclusive': Bool(90300, None),
+    'recovery_target_lsn': String(100000, None),
+    'recovery_target_name': String(90400, None),
+    'recovery_target_time': String(90300, None),
+    'recovery_target_timeline': String(90300, None),
+    'recovery_target_xid': String(90300, None),
+    'restore_command': String(90300, None),
+    'standby_mode': Bool(90300, 120000),
+    'trigger_file': String(90300, 120000)
+})
+
+
+def _transform_parameter_value(validators, version, name, value):
+    validators = validators.get(name)
+    if validators:
+        for validator in (validators if isinstance(validators[0], tuple) else [validators]):
+            if version >= validator.version_from and\
+                    (validator.version_till is None or version < validator.version_till):
+                return validator.transform(name, value)
+    logger.warning('Removing unexpected parameter=%s value=%s from the config', name, value)
+
+
+def transform_postgresql_parameter_value(version, name, value):
+    if '.' in name:
+        return value
+    return _transform_parameter_value(parameters, version, name, value)
+
+
+def transform_recovery_parameter_value(version, name, value):
+    return _transform_parameter_value(recovery_parameters, version, name, value)

patroni/raft_controller.py

@@ -0,0 +1,34 @@
+import logging
+import os
+
+from patroni.daemon import AbstractPatroniDaemon, abstract_main
+from patroni.dcs.raft import KVStoreTTL
+from pysyncobj import SyncObjConf
+
+logger = logging.getLogger(__name__)
+
+
+class RaftController(AbstractPatroniDaemon):
+
+    def __init__(self, config):
+        super(RaftController, self).__init__(config)
+
+        raft_config = self.config.get('raft')
+        self_addr = raft_config['self_addr']
+        template = os.path.join(raft_config.get('data_dir', ''), self_addr)
+        self._syncobj_config = SyncObjConf(autoTick=False, appendEntriesUseBatch=False, dynamicMembershipChange=True,
+                                           journalFile=template + '.journal', fullDumpFile=template + '.dump')
+        self._raft = KVStoreTTL(self_addr, raft_config.get('partner_addrs', []), self._syncobj_config)
+
+    def _run_cycle(self):
+        try:
+            self._raft.doTick(self._syncobj_config.autoTickPeriod)
+        except Exception:
+            logger.exception('doTick')
+
+    def _shutdown(self):
+        self._raft.destroy()
+
+
+def main():
+    abstract_main(RaftController)

patroni/request.py

@@ -0,0 +1,58 @@
+import json
+import urllib3
+import six
+
+from six.moves.urllib_parse import urlparse, urlunparse
+
+from .utils import USER_AGENT
+
+
+class PatroniRequest(object):
+
+    def __init__(self, config, insecure=False):
+        cert_reqs = 'CERT_NONE' if insecure or config.get('ctl', {}).get('insecure', False) else 'CERT_REQUIRED'
+        self._pool = urllib3.PoolManager(num_pools=10, maxsize=10, cert_reqs=cert_reqs)
+        self.reload_config(config)
+
+    @staticmethod
+    def _get_cfg_value(config, name):
+        return config.get('ctl', {}).get(name) or config.get('restapi', {}).get(name)
+
+    def _apply_pool_param(self, param, value):
+        if value:
+            self._pool.connection_pool_kw[param] = value
+        else:
+            self._pool.connection_pool_kw.pop(param, None)
+
+    def _apply_ssl_file_param(self, config, name):
+        value = self._get_cfg_value(config, name + 'file')
+        self._apply_pool_param(name + '_file', value)
+        return value
+
+    def reload_config(self, config):
+        self._pool.headers = urllib3.make_headers(basic_auth=self._get_cfg_value(config, 'auth'), user_agent=USER_AGENT)
+
+        if self._apply_ssl_file_param(config, 'cert'):
+            self._apply_ssl_file_param(config, 'key')
+        else:
+            self._pool.connection_pool_kw.pop('key_file', None)
+
+        cacert = config.get('ctl', {}).get('cacert') or config.get('restapi', {}).get('cafile')
+        self._apply_pool_param('ca_certs', cacert)
+
+    def request(self, method, url, body=None, **kwargs):
+        if body is not None and not isinstance(body, six.string_types):
+            body = json.dumps(body)
+        return self._pool.request(method.upper(), url, body=body, **kwargs)
+
+    def __call__(self, member, method='GET', endpoint=None, data=None, **kwargs):
+        url = member.api_url
+        if endpoint:
+            scheme, netloc, _, _, _, _ = urlparse(url)
+            url = urlunparse((scheme, netloc, endpoint, '', '', ''))
+        return self.request(method, url, data, **kwargs)
+
+
+def get(url, verify=True, **kwargs):
+    http = PatroniRequest({}, not verify)
+    return http.request('GET', url, **kwargs)

patroni/scripts/aws.py

@@ -1,12 +1,12 @@
 #!/usr/bin/env python
 
+import json
 import logging
-import requests
-from requests.exceptions import RequestException
 import sys
 import boto.ec2
 
 from patroni.utils import Retry, RetryFailedError
+from patroni.request import get as requests_get
 
 logger = logging.getLogger(__name__)
 
@@ -19,14 +19,14 @@ class AWSConnection(object):
         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=2.1)
-        except RequestException:
+            r = requests_get('http://169.254.169.254/latest/dynamic/instance-identity/document', timeout=2.1)
+        except Exception:
             logger.error('cannot query AWS meta-data')
             return
 
-        if r.ok:
+        if r.status < 400:
             try:
-                content = r.json()
+                content = json.loads(r.data.decode('utf-8'))
                 self.instance_id = content['instanceId']
                 self.region = content['region']
             except Exception:

patroni/scripts/wale_restore.py

@@ -224,13 +224,12 @@ class WALERestore(object):
                             lsn_name = 'location'
                         con.autocommit = True
                         with con.cursor() as cur:
-                            cur.execute("""SELECT CASE WHEN pg_is_in_recovery()
-                                                       THEN GREATEST(
-                                                                pg_{0}_{1}_diff(COALESCE(
-                                                                  pg_last_{0}_receive_{1}(), '0/0'), %s)::bigint,
-                                                                pg_{0}_{1}_diff(pg_last_{0}_replay_{1}(), %s)::bigint)
-                                                       ELSE pg_{0}_{1}_diff(pg_current_{0}_{1}(), %s)::bigint
-                                                   END""".format(wal_name, lsn_name),
+                            cur.execute(("SELECT CASE WHEN pg_catalog.pg_is_in_recovery()"
+                                         " THEN GREATEST(pg_catalog.pg_{0}_{1}_diff(COALESCE("
+                                         "pg_last_{0}_receive_{1}(), '0/0'), %s)::bigint, "
+                                         "pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_last_{0}_replay_{1}(), %s)::bigint)"
+                                         " ELSE pg_catalog.pg_{0}_{1}_diff(pg_catalog.pg_current_{0}_{1}(), %s)::bigint"
+                                         " END").format(wal_name, lsn_name),
                                         (backup_start_lsn, backup_start_lsn, backup_start_lsn))
 
                             diff_in_bytes = int(cur.fetchone()[0])

patroni/utils.py

@@ -1,11 +1,30 @@
+import errno
+import json.decoder as json_decoder
+import logging
+import os
+import platform
 import random
+import re
+import socket
+import sys
+import tempfile
 import time
 
 from dateutil import tz
-from patroni.exceptions import PatroniException
+
+from .exceptions import PatroniException
+from .version import __version__
 
 tzutc = tz.tzutc()
 
+logger = logging.getLogger(__name__)
+
+USER_AGENT = 'Patroni/{0} Python/{1} {2}'.format(__version__, platform.python_version(), platform.system())
+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):
     """
@@ -88,42 +107,77 @@ 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()
-    ln = len(value)
-    i = 0
-    # skip sign:
-    if i < ln and value[i] in ('-', '+'):
-        i += 1
+    for regex, base in ((HEX_RE, 16), (OCT_RE, 8), (DEC_RE, 10)):
+        match = regex.match(value)
+        if match:
+            end = match.end()
+            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 < ln and value[i].isdigit():
-        if value[i] == '0':
-            i += 1
-            if i < ln 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
-        while i <= ln:
-            try:  # try to find maximally long number
-                i += 1  # by giving to `int` longer and longer strings
-                ret = int(value[:i], base)
-            except ValueError:  # until we will not get an exception or end of the string
-                i -= 1
-                break
-        if ret is not None:  # yay! there is a number in the beginning of the string
-            return ret, value[i:].strip()  # return the number and the "rest"
+def strtod(value):
+    """As most as possible close equivalent of strtod(3) function used by postgres to parse parameter values.
+    >>> strtod(' A ') == (None, 'A')
+    True
+    """
+    value = str(value).strip()
+    match = DBL_RE.match(value)
+    if match:
+        end = match.end()
+        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):
@@ -136,56 +190,71 @@ 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 = {
-        'kB': {'kB': 1, 'MB': 1024, 'GB': 1024 * 1024, 'TB': 1024 * 1024 * 1024},
-        'ms': {'ms': 1, 's': 1000, 'min': 1000 * 60, 'h': 1000 * 60 * 60, 'd': 1000 * 60 * 60 * 24},
-        '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}
-    }
+    val, unit = strtol(value)
+    if val is None and unit.startswith('.') or unit and unit[0] in ('.', 'e', 'E'):
+        val, unit = strtod(value)
 
-    value, unit = strtol(value)
-    if value is not None:
+    if val is not None:
+        unit = unit.strip()
         if not unit:
-            return value
+            return int(rint(val))
 
-        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]:
-            multiplier = convert[base_unit][unit]
-            if multiplier < 0:
-                value /= -multiplier
-            else:
-                value *= multiplier
-            return int(value/base_value)
+        val = convert_to_base_unit(val, unit, base_unit)
+        if val is not None:
+            return int(rint(val))
+
+
+def parse_real(value, base_unit=None):
+    """
+    >>> parse_real(' +0.0005 ') == 0.0005
+    True
+    >>> parse_real('0.0005ms', 'ms') == 0.0
+    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
-    if vartype == 'bool':
-        old_value = parse_bool(old_value)
-        new_value = parse_bool(new_value)
-    elif vartype == 'integer':
-        old_value = parse_int(old_value)
-        new_value = parse_int(new_value, unit)
-    elif vartype == 'enum':
-        return str(old_value).lower() == str(new_value).lower()
-    else:  # ('string', 'real')
-        return str(old_value) == str(new_value)
+    converters = {
+        'bool': lambda v1, v2: parse_bool(v1),
+        'integer': parse_int,
+        'real': parse_real,
+        'enum': lambda v1, v2: str(v1).lower(),
+        'string': lambda v1, v2: str(v1)
+    }
+
+    convert = converters.get(vartype) or converters['string']
+    old_value = convert(old_value, None)
+    new_value = convert(new_value, unit)
+
     return old_value is not None and new_value is not None and old_value == new_value
 
 
@@ -237,6 +306,17 @@ class Retry(object):
                      max_jitter=self.max_jitter / 100.0, max_delay=self.max_delay, sleep_func=self.sleep_func,
                      deadline=self.deadline, retry_exceptions=self.retry_exceptions)
 
+    @property
+    def sleeptime(self):
+        return self._cur_delay + (random.randint(0, self.max_jitter) / 100.0)
+
+    def update_delay(self):
+        self._cur_delay = min(self._cur_delay * self.backoff, self.max_delay)
+
+    @property
+    def stoptime(self):
+        return self._cur_stoptime
+
     def __call__(self, func, *args, **kwargs):
         """Call a function with arguments until it completes without throwing a `retry_exceptions`
 
@@ -252,18 +332,20 @@ 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)
+                sleeptime = hasattr(e, 'sleeptime') and e.sleeptime or self.sleeptime
 
                 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:
-                    self.sleep_func(sleeptime)
-                self._cur_delay = min(self._cur_delay * self.backoff, self.max_delay)
+                logger.debug('Retry got exception: %s', e)
+                self.sleep_func(sleeptime)
+                self.update_delay()
 
 
 def polling_loop(timeout, interval=1):
@@ -279,5 +361,174 @@ def polling_loop(timeout, interval=1):
 
 def split_host_port(value, default_port):
     t = value.rsplit(':', 1)
+    if ':' in t[0]:
+        t[0] = t[0].strip('[]')
     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)
+
+
+def iter_response_objects(response):
+    prev = ''
+    decoder = json_decoder.JSONDecoder()
+    for chunk in response.read_chunked(decode_content=False):
+        if isinstance(chunk, bytes):
+            chunk = chunk.decode('utf-8')
+        chunk = prev + chunk
+
+        length = len(chunk)
+        idx = json_decoder.WHITESPACE.match(chunk, 0).end()
+        while idx < length:
+            try:
+                message, idx = decoder.raw_decode(chunk, idx)
+            except ValueError:  # malformed or incomplete JSON, unlikely to happen
+                break
+            else:
+                yield message
+                idx = json_decoder.WHITESPACE.match(chunk, idx).end()
+        prev = chunk[idx:]
+
+
+def is_standby_cluster(config):
+    # Check whether or not provided configuration describes a standby cluster
+    return isinstance(config, dict) and (config.get('host') or config.get('port') or config.get('restore_command'))
+
+
+def cluster_as_json(cluster):
+    leader_name = cluster.leader.name if cluster.leader else None
+    xlog_location_cluster = cluster.last_leader_operation or 0
+
+    ret = {'members': []}
+    for m in cluster.members:
+        if m.name == leader_name:
+            config = cluster.config.data if cluster.config and cluster.config.modify_index else {}
+            role = 'standby_leader' if is_standby_cluster(config.get('standby_cluster')) else 'leader'
+        elif m.name in cluster.sync.members:
+            role = 'sync_standby'
+        else:
+            role = 'replica'
+
+        member = {'name': m.name, 'role': role, 'state': m.data.get('state', ''), 'api_url': m.api_url}
+        conn_kwargs = m.conn_kwargs()
+        if conn_kwargs.get('host'):
+            member['host'] = conn_kwargs['host']
+            if conn_kwargs.get('port'):
+                member['port'] = int(conn_kwargs['port'])
+        optional_attributes = ('timeline', 'pending_restart', 'scheduled_restart', 'tags')
+        member.update({n: m.data[n] for n in optional_attributes if n in m.data})
+
+        if m.name != leader_name:
+            xlog_location = m.data.get('xlog_location')
+            if xlog_location is None:
+                member['lag'] = 'unknown'
+            elif xlog_location_cluster >= xlog_location:
+                member['lag'] = xlog_location_cluster - xlog_location
+            else:
+                member['lag'] = 0
+
+        ret['members'].append(member)
+
+    # sort members by name for consistency
+    ret['members'].sort(key=lambda m: m['name'])
+    if cluster.is_paused():
+        ret['pause'] = True
+    if cluster.failover and cluster.failover.scheduled_at:
+        ret['scheduled_switchover'] = {'at': cluster.failover.scheduled_at.isoformat()}
+        if cluster.failover.leader:
+            ret['scheduled_switchover']['from'] = cluster.failover.leader
+        if cluster.failover.candidate:
+            ret['scheduled_switchover']['to'] = cluster.failover.candidate
+    return ret
+
+
+def is_subpath(d1, d2):
+    real_d1 = os.path.realpath(d1) + os.path.sep
+    real_d2 = os.path.realpath(os.path.join(real_d1, d2))
+    return os.path.commonprefix([real_d1, real_d2 + os.path.sep]) == real_d1
+
+
+def validate_directory(d, msg="{} {}"):
+    if not os.path.exists(d):
+        try:
+            os.makedirs(d)
+        except OSError as e:
+            logger.error(e)
+            if e.errno != errno.EEXIST:
+                raise PatroniException(msg.format(d, "couldn't create the directory"))
+    elif os.path.isdir(d):
+        try:
+            fd, tmpfile = tempfile.mkstemp(dir=d)
+            os.close(fd)
+            os.remove(tmpfile)
+        except OSError:
+            raise PatroniException(msg.format(d, "the directory is not writable"))
+    else:
+        raise PatroniException(msg.format(d, "is not a directory"))
+
+
+def data_directory_is_empty(data_dir):
+    if not os.path.exists(data_dir):
+        return True
+    return all(os.name != 'nt' and (n.startswith('.') or n == 'lost+found') for n in os.listdir(data_dir))
+
+
+def keepalive_intvl(timeout, idle, cnt=3):
+    return max(1, int(float(timeout - idle) / cnt))
+
+
+def keepalive_socket_options(timeout, idle, cnt=3):
+    yield (socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
+
+    if sys.platform.startswith('linux'):
+        yield (socket.SOL_TCP, 18, int(timeout * 1000))  # TCP_USER_TIMEOUT
+        TCP_KEEPIDLE = getattr(socket, 'TCP_KEEPIDLE', None)
+        TCP_KEEPINTVL = getattr(socket, 'TCP_KEEPINTVL', None)
+        TCP_KEEPCNT = getattr(socket, 'TCP_KEEPCNT', None)
+    elif sys.platform.startswith('darwin'):
+        TCP_KEEPIDLE = 0x10  # (named "TCP_KEEPALIVE" in C)
+        TCP_KEEPINTVL = 0x101
+        TCP_KEEPCNT = 0x102
+    else:
+        return
+
+    intvl = keepalive_intvl(timeout, idle, cnt)
+    yield (socket.IPPROTO_TCP, TCP_KEEPIDLE, idle)
+    yield (socket.IPPROTO_TCP, TCP_KEEPINTVL, intvl)
+    yield (socket.IPPROTO_TCP, TCP_KEEPCNT, cnt)
+
+
+def enable_keepalive(sock, timeout, idle, cnt=3):
+    SIO_KEEPALIVE_VALS = getattr(socket, 'SIO_KEEPALIVE_VALS', None)
+    if SIO_KEEPALIVE_VALS is not None:  # Windows
+        intvl = keepalive_intvl(timeout, idle, cnt)
+        return sock.ioctl(SIO_KEEPALIVE_VALS, (1, idle * 1000, intvl * 1000))
+
+    for opt in keepalive_socket_options(timeout, idle, cnt):
+        sock.setsockopt(*opt)
+
+
+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

patroni/validator.py

@@ -0,0 +1,393 @@
+#!/usr/bin/env python3
+import os
+import socket
+import re
+import subprocess
+
+from six import string_types
+
+from .utils import find_executable, split_host_port, data_directory_is_empty
+from .dcs import dcs_modules
+from .exceptions import ConfigParseError
+
+
+def data_directory_empty(data_dir):
+    if os.path.isfile(os.path.join(data_dir, "global", "pg_control")):
+        return False
+    return data_directory_is_empty(data_dir)
+
+
+def validate_connect_address(address):
+    try:
+        host, _ = split_host_port(address, 1)
+    except (AttributeError, TypeError, ValueError):
+        raise ConfigParseError("contains a wrong value")
+    if host in ["127.0.0.1", "0.0.0.0", "*", "::1", "localhost"]:
+        raise ConfigParseError('must not contain "127.0.0.1", "0.0.0.0", "*", "::1", "localhost"')
+    return True
+
+
+def validate_host_port(host_port, listen=False, multiple_hosts=False):
+    try:
+        hosts, port = split_host_port(host_port, None)
+    except (ValueError, TypeError):
+        raise ConfigParseError("contains a wrong value")
+    else:
+        if multiple_hosts:
+            hosts = hosts.split(",")
+        else:
+            hosts = [hosts]
+        for host in hosts:
+            proto = socket.getaddrinfo(host, "", 0, socket.SOCK_STREAM, 0, socket.AI_PASSIVE)
+            s = socket.socket(proto[0][0], socket.SOCK_STREAM)
+            try:
+                if s.connect_ex((host, port)) == 0:
+                    if listen:
+                        raise ConfigParseError("Port {} is already in use.".format(port))
+                elif not listen:
+                    raise ConfigParseError("{} is not reachable".format(host_port))
+            except socket.gaierror as e:
+                raise ConfigParseError(e)
+            finally:
+                s.close()
+    return True
+
+
+def validate_host_port_list(value):
+    assert all([validate_host_port(v) for v in value]), "didn't pass the validation"
+    return True
+
+
+def comma_separated_host_port(string):
+    return validate_host_port_list([s.strip() for s in string.split(",")])
+
+
+def validate_host_port_listen(host_port):
+    return validate_host_port(host_port, listen=True)
+
+
+def validate_host_port_listen_multiple_hosts(host_port):
+    return validate_host_port(host_port, listen=True, multiple_hosts=True)
+
+
+def is_ipv4_address(ip):
+    try:
+        socket.inet_aton(ip)
+    except Exception:
+        raise ConfigParseError("Is not a valid ipv4 address")
+    return True
+
+
+def is_ipv6_address(ip):
+    try:
+        socket.inet_pton(socket.AF_INET6, ip)
+    except Exception:
+        raise ConfigParseError("Is not a valid ipv6 address")
+    return True
+
+
+def get_major_version(bin_dir=None):
+    if not bin_dir:
+        binary = 'postgres'
+    else:
+        binary = os.path.join(bin_dir, 'postgres')
+    version = subprocess.check_output([binary, '--version']).decode()
+    version = re.match(r'^[^\s]+ [^\s]+ (\d+)(\.(\d+))?', version)
+    return '.'.join([version.group(1), version.group(3)]) if int(version.group(1)) < 10 else version.group(1)
+
+
+def validate_data_dir(data_dir):
+    if not data_dir:
+        raise ConfigParseError("is an empty string")
+    elif os.path.exists(data_dir) and not os.path.isdir(data_dir):
+        raise ConfigParseError("is not a directory")
+    elif not data_directory_empty(data_dir):
+        if not os.path.exists(os.path.join(data_dir, "PG_VERSION")):
+            raise ConfigParseError("doesn't look like a valid data directory")
+        else:
+            with open(os.path.join(data_dir, "PG_VERSION"), "r") as version:
+                pgversion = version.read().strip()
+            waldir = ("pg_wal" if float(pgversion) >= 10 else "pg_xlog")
+            if not os.path.isdir(os.path.join(data_dir, waldir)):
+                raise ConfigParseError("data dir for the cluster is not empty, but doesn't contain"
+                                       " \"{}\" directory".format(waldir))
+            bin_dir = schema.data.get("postgresql", {}).get("bin_dir", None)
+            major_version = get_major_version(bin_dir)
+            if pgversion != major_version:
+                raise ConfigParseError("data_dir directory postgresql version ({}) doesn't match with "
+                                       "'postgres --version' output ({})".format(pgversion, major_version))
+    return True
+
+
+class Result(object):
+    def __init__(self, status, error="didn't pass validation", level=0, path="", data=""):
+        self.status = status
+        self.path = path
+        self.data = data
+        self.level = level
+        self._error = error
+        if not self.status:
+            self.error = error
+        else:
+            self.error = None
+
+    def __repr__(self):
+        return self.path + (" " + str(self.data) + " " + self._error if self.error else "")
+
+
+class Case(object):
+    def __init__(self, schema):
+        self._schema = schema
+
+
+class Or(object):
+    def __init__(self, *args):
+        self.args = args
+
+
+class Optional(object):
+    def __init__(self, name):
+        self.name = name
+
+
+class Directory(object):
+    def __init__(self, contains=None, contains_executable=None):
+        self.contains = contains
+        self.contains_executable = contains_executable
+
+    def validate(self, name):
+        if not name:
+            yield Result(False, "is an empty string")
+        elif not os.path.exists(name):
+            yield Result(False, "Directory '{}' does not exist.".format(name))
+        elif not os.path.isdir(name):
+            yield Result(False, "'{}' is not a directory.".format(name))
+        else:
+            if self.contains:
+                for path in self.contains:
+                    if not os.path.exists(os.path.join(name, path)):
+                        yield Result(False, "'{}' does not contain '{}'".format(name, path))
+            if self.contains_executable:
+                for program in self.contains_executable:
+                    if not find_executable(program, name):
+                        yield Result(False, "'{}' does not contain '{}'".format(name, program))
+
+
+class Schema(object):
+    def __init__(self, validator):
+        self.validator = validator
+
+    def __call__(self, data):
+        for i in self.validate(data):
+            if not i.status:
+                print(i)
+
+    def validate(self, data):
+        self.data = data
+        if isinstance(self.validator, string_types):
+            yield Result(isinstance(self.data, string_types), "is not a string", level=1, data=self.data)
+        elif issubclass(type(self.validator), type):
+            validator = self.validator
+            if self.validator == str:
+                validator = string_types
+            yield Result(isinstance(self.data, validator),
+                         "is not {}".format(_get_type_name(self.validator)), level=1, data=self.data)
+        elif callable(self.validator):
+            if hasattr(self.validator, "expected_type"):
+                if not isinstance(data, self.validator.expected_type):
+                    yield Result(False, "is not {}"
+                                 .format(_get_type_name(self.validator.expected_type)), level=1, data=self.data)
+                    return
+            try:
+                self.validator(data)
+                yield Result(True, data=self.data)
+            except Exception as e:
+                yield Result(False, "didn't pass validation: {}".format(e), data=self.data)
+        elif isinstance(self.validator, dict):
+            if not len(self.validator):
+                yield Result(isinstance(self.data, dict), "is not a dictionary", level=1, data=self.data)
+        elif isinstance(self.validator, list):
+            if not isinstance(self.data, list):
+                yield Result(isinstance(self.data, list), "is not a list", level=1, data=self.data)
+                return
+        for i in self.iter():
+            yield i
+
+    def iter(self):
+        if isinstance(self.validator, dict):
+            if not isinstance(self.data, dict):
+                yield Result(False, "is not a dictionary.", level=1)
+            else:
+                for i in self.iter_dict():
+                    yield i
+        elif isinstance(self.validator, list):
+            if len(self.data) == 0:
+                yield Result(False, "is an empty list", data=self.data)
+            if len(self.validator) > 0:
+                for key, value in enumerate(self.data):
+                    for v in Schema(self.validator[0]).validate(value):
+                        yield Result(v.status, v.error,
+                                     path=(str(key) + ("." + v.path if v.path else "")), level=v.level, data=value)
+        elif isinstance(self.validator, Directory):
+            for v in self.validator.validate(self.data):
+                yield v
+        elif isinstance(self.validator, Or):
+            for i in self.iter_or():
+                yield i
+
+    def iter_dict(self):
+        for key in self.validator.keys():
+            for d in self._data_key(key):
+                if d not in self.data and not isinstance(key, Optional):
+                    yield Result(False, "is not defined.", path=d)
+                elif d not in self.data and isinstance(key, Optional):
+                    continue
+                else:
+                    validator = self.validator[key]
+                    if isinstance(key, Or) and isinstance(self.validator[key], Case):
+                        validator = self.validator[key]._schema[d]
+                    for v in Schema(validator).validate(self.data[d]):
+                        yield Result(v.status, v.error,
+                                     path=(d + ("." + v.path if v.path else "")), level=v.level, data=v.data)
+
+    def iter_or(self):
+        results = []
+        for a in self.validator.args:
+            r = []
+            for v in Schema(a).validate(self.data):
+                r.append(v)
+            if any([x.status for x in r]) and not all([x.status for x in r]):
+                results += filter(lambda x: not x.status, r)
+            else:
+                results += r
+        if not any([x.status for x in results]):
+            max_level = 3
+            for v in sorted(results, key=lambda x: x.level):
+                if v.level > max_level:
+                    break
+                max_level = v.level
+                yield Result(v.status, v.error, path=v.path, level=v.level, data=v.data)
+
+    def _data_key(self, key):
+        if isinstance(self.data, dict) and isinstance(key, str):
+            yield key
+        elif isinstance(key, Optional):
+            yield key.name
+        elif isinstance(key, Or):
+            if any([i in self.data for i in key.args]):
+                for i in key.args:
+                    if i in self.data:
+                        yield i
+            else:
+                for i in key.args:
+                    yield i
+
+
+def _get_type_name(python_type):
+    return {str: 'a string', int: 'and integer', float: 'a number', bool: 'a boolean',
+            list: 'an array', dict: 'a dictionary', string_types: "a string"}.get(
+                    python_type, getattr(python_type, __name__, "unknown type"))
+
+
+def assert_(condition, message="Wrong value"):
+    assert condition, message
+
+
+userattributes = {"username": "", Optional("password"): ""}
+available_dcs = [m.split(".")[-1] for m in dcs_modules()]
+validate_host_port_list.expected_type = list
+comma_separated_host_port.expected_type = string_types
+validate_connect_address.expected_type = string_types
+validate_host_port_listen.expected_type = string_types
+validate_host_port_listen_multiple_hosts.expected_type = string_types
+validate_data_dir.expected_type = string_types
+validate_etcd = {
+    Or("host", "hosts", "srv", "url", "proxy"): Case({
+        "host": validate_host_port,
+        "hosts": Or(comma_separated_host_port, [validate_host_port]),
+        "srv": str,
+        "url": str,
+        "proxy": str})
+}
+
+schema = Schema({
+  "name": str,
+  "scope": str,
+  "restapi": {
+    "listen": validate_host_port_listen,
+    "connect_address": validate_connect_address
+  },
+  Optional("bootstrap"): {
+    "dcs": {
+        Optional("ttl"): int,
+        Optional("loop_wait"): int,
+        Optional("retry_timeout"): int,
+        Optional("maximum_lag_on_failover"): int
+        },
+    "pg_hba": [str],
+    "initdb": [Or(str, dict)]
+  },
+  Or(*available_dcs): Case({
+      "consul": {
+          Or("host", "url"): Case({
+              "host": validate_host_port,
+              "url": str})
+          },
+      "etcd": validate_etcd,
+      "etcd3": validate_etcd,
+      "exhibitor": {
+          "hosts": [str],
+          "port": lambda i: assert_(int(i) <= 65535),
+          Optional("pool_interval"): int
+          },
+      "raft": {
+          "self_addr": validate_connect_address,
+          Optional("bind_addr"): validate_host_port_listen,
+          "partner_addrs": validate_host_port_list,
+          Optional("data_dir"): str,
+          Optional("password"): str
+          },
+      "zookeeper": {
+          "hosts": Or(comma_separated_host_port, [validate_host_port]),
+          },
+      "kubernetes": {
+          "labels": {},
+          Optional("namespace"): str,
+          Optional("scope_label"): str,
+          Optional("role_label"): str,
+          Optional("use_endpoints"): bool,
+          Optional("pod_ip"): Or(is_ipv4_address, is_ipv6_address),
+          Optional("ports"): [{"name": str, "port": int}],
+          },
+      }),
+  "postgresql": {
+    "listen": validate_host_port_listen_multiple_hosts,
+    "connect_address": validate_connect_address,
+    "authentication": {
+      "replication": userattributes,
+      "superuser": userattributes,
+      "rewind":  userattributes
+    },
+    "data_dir": validate_data_dir,
+    Optional("bin_dir"): Directory(contains_executable=["pg_ctl", "initdb", "pg_controldata", "pg_basebackup",
+                                                        "postgres", "pg_isready"]),
+    Optional("parameters"): {
+      Optional("unix_socket_directories"): lambda s: assert_(all([isinstance(s, string_types), len(s)]))
+    },
+    Optional("pg_hba"): [str],
+    Optional("pg_ident"): [str],
+    Optional("pg_ctl_timeout"): int,
+    Optional("use_pg_rewind"): bool
+  },
+  Optional("watchdog"): {
+    Optional("mode"): lambda m: assert_(m in ["off", "automatic", "required"]),
+    Optional("device"): str
+  },
+  Optional("tags"): {
+    Optional("nofailover"): bool,
+    Optional("clonefrom"): bool,
+    Optional("noloadbalance"): bool,
+    Optional("replicatefrom"): str,
+    Optional("nosync"): bool
+  }
+})

patroni/version.py

@@ -1 +1 @@
-__version__ = '1.4.5'
+__version__ = '2.0.2'

patroni/watchdog/base.py

@@ -57,7 +57,7 @@ class WatchdogConfig(object):
         return not self == other
 
     def get_impl(self):
-        if self.driver == 'testing':
+        if self.driver == 'testing':  # pragma: no cover
             from patroni.watchdog.linux import TestingWatchdogDevice
             return TestingWatchdogDevice.from_config(self.driver_config)
         elif platform.system() == 'Linux' and self.driver == 'default':

patroni/watchdog/linux.py

@@ -1,6 +1,5 @@
 import collections
 import ctypes
-import fcntl
 import os
 import platform
 from patroni.watchdog.base import WatchdogBase, WatchdogError
@@ -17,11 +16,11 @@ IOC_DIRBITS = 2
 
 # Non-generic platform special cases
 machine = platform.machine()
-if machine in ['mips', 'sparc', 'powerpc', 'ppc64']:
+if machine in ['mips', 'sparc', 'powerpc', 'ppc64']:  # pragma: no cover
     IOC_SIZEBITS = 13
     IOC_DIRBITS = 3
     IOC_NONE, IOC_WRITE, IOC_READ = 1, 2, 4
-elif machine == 'parisc':
+elif machine == 'parisc':  # pragma: no cover
     IOC_WRITE, IOC_READ = 2, 1
 
 IOC_NRSHIFT = 0
@@ -162,7 +161,9 @@ class LinuxWatchdogDevice(WatchdogBase):
         Raises OSError or IOError (Python 2) when the ioctl fails."""
         if self._fd is None:
             raise WatchdogError("Watchdog device is closed")
-        fcntl.ioctl(self._fd, func, arg, True)
+        if os.name != 'nt':
+            import fcntl
+            fcntl.ioctl(self._fd, func, arg, True)
 
     def get_support(self):
         if self._support_cache is None:
@@ -217,7 +218,7 @@ class LinuxWatchdogDevice(WatchdogBase):
         return timeout.value
 
 
-class TestingWatchdogDevice(LinuxWatchdogDevice):
+class TestingWatchdogDevice(LinuxWatchdogDevice):  # pragma: no cover
     """Converts timeout ioctls to regular writes that can be intercepted from a named pipe."""
     timeout = 60
 

patroni_raft_controller.py

@@ -0,0 +1,6 @@
+#!/usr/bin/env python
+from patroni.raft_controller import main
+
+
+if __name__ == '__main__':
+    main()

postgres0.yml

@@ -11,8 +11,31 @@ 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:
+  #Provide host to do the initial discovery of the cluster topology:
   host: 127.0.0.1:2379
+  #Or use "hosts" to provide multiple endpoints
+  #Could be a comma separated string:
+  #hosts: host1:port1,host2:port2
+  #or an actual yaml list:
+  #hosts:
+  #- host1:port1
+  #- host2:port2
+  #Once discovery is complete Patroni will use the list of advertised clientURLs
+  #It is possible to change this behavior through by setting:
+  #use_proxies: true
+
+#raft:
+#  data_dir: .
+#  self_addr: 127.0.0.1:2222
+#  partner_addrs:
+#  - 127.0.0.1:2223
+#  - 127.0.0.1:2224
 
 bootstrap:
   # this section will be written into Etcd:/<namespace>/<scope>/config after initializing new cluster
@@ -24,6 +47,10 @@ 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
@@ -46,6 +73,9 @@ 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
@@ -75,8 +105,18 @@ 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: '.'
+  # Additional fencing script executed after acquiring the leader lock but before promoting the replica
+  #pre_promote: /path/to/pre_promote.sh
 
 #watchdog:
 #  mode: automatic # Allowed values: off, automatic, required

postgres1.yml

@@ -11,8 +11,31 @@ 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:
+  #Provide host to do the initial discovery of the cluster topology:
   host: 127.0.0.1:2379
+  #Or use "hosts" to provide multiple endpoints
+  #Could be a comma separated string:
+  #hosts: host1:port1,host2:port2
+  #or an actual yaml list:
+  #hosts:
+  #- host1:port1
+  #- host2:port2
+  #Once discovery is complete Patroni will use the list of advertised clientURLs
+  #It is possible to change this behavior through by setting:
+  #use_proxies: true
+
+#raft:
+#  data_dir: .
+#  self_addr: 127.0.0.1:2223
+#  partner_addrs:
+#  - 127.0.0.1:2222
+#  - 127.0.0.1:2224
 
 bootstrap:
   # this section will be written into Etcd:/<namespace>/<scope>/config after initializing new cluster
@@ -44,6 +67,9 @@ 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
@@ -73,11 +99,23 @@ 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
+#      - waldir: /pg-wal-mount/external-waldir # only needed in case pg_wal is symlinked outside of data_dir
+  # Additional fencing script executed after acquiring the leader lock but before promoting the replica
+  #pre_promote: /path/to/pre_promote.sh
+
 tags:
     nofailover: false
     noloadbalance: false