Bugfix: raise ConfigParseError (#2463)

instead of returning error string

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1 @@
+blank_issues_enabled: false

.github/workflows/install_deps.py

@@ -30,6 +30,7 @@ def install_requirements(what):
                 requirements.append(r)
 
     subprocess.call([sys.executable, '-m', 'pip', 'install', '--upgrade', 'pip'])
+    subprocess.call([sys.executable, '-m', 'pip', 'install', '--upgrade', 'wheel'])
     r = subprocess.call([sys.executable, '-m', 'pip', 'install'] + requirements)
     s = subprocess.call([sys.executable, '-m', 'pip', 'install', '--upgrade', 'setuptools'])
     return s | r
@@ -45,10 +46,14 @@ def install_packages(what):
     packages['exhibitor'] = packages['zookeeper']
     packages = packages.get(what, [])
     ver = versions.get(what)
+    subprocess.call(['sudo', 'apt-get', 'update', '-y'])
+    subprocess.call(['sudo', 'apt-get', 'install', '-y', 'wget', 'ca-certificates', 'gnupg', 'expect-dev'])
+    subprocess.call(['sudo', 'sh', '-c', "wget -qO - https://www.postgresql.org/media/keys/ACCC4CF8.asc"
+                     " | gpg --dearmor > /etc/apt/trusted.gpg.d/apt.postgresql.org.gpg"])
     subprocess.call(['sudo', 'sed', '-i', 's/pgdg main.*$/pgdg main {0}/'.format(ver),
                      '/etc/apt/sources.list.d/pgdg.list'])
     subprocess.call(['sudo', 'apt-get', 'update', '-y'])
-    return subprocess.call(['sudo', 'apt-get', 'install', '-y', 'postgresql-' + ver, 'expect-dev', 'wget'] + packages)
+    return subprocess.call(['sudo', 'apt-get', 'install', '-y', 'postgresql-' + ver] + packages)
 
 
 def get_file(url, name):

.github/workflows/mapping.py

@@ -1 +1 @@
-versions = {'etcd': '9.6', 'etcd3': '14', 'consul': '13', 'exhibitor': '12', 'raft': '11', 'kubernetes': '14'}
+versions = {'etcd': '9.6', 'etcd3': '14', 'consul': '13', 'exhibitor': '12', 'raft': '11', 'kubernetes': '15'}

.github/workflows/release.yaml

@@ -0,0 +1,41 @@
+name: Publish Patroni distributions to PyPI and TestPyPI
+on:
+  push:
+    tags:
+      - 'v[0-9]+.[0-9]+.[0-9]+'
+  release:
+    types:
+      - published
+jobs:
+  build-n-publish:
+    name: Build and publish Patroni distributions to PyPI and TestPyPI
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@master
+
+    - name: Set up Python 3.9
+      uses: actions/setup-python@v4
+      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: Build a binary wheel and a source tarball
+      run: python setup.py sdist bdist_wheel
+
+    - name: Publish distribution to Test PyPI
+      if: github.event_name == 'push'
+      uses: pypa/[email protected]
+      with:
+        password: ${{ secrets.TEST_PYPI_API_TOKEN }}
+        repository_url: https://test.pypi.org/legacy/
+
+    - name: Publish distribution to PyPI
+      if: github.event_name == 'release'
+      uses: pypa/[email protected]
+      with:
+        password: ${{ secrets.PYPI_API_TOKEN }}

.github/workflows/run_tests.py

@@ -28,16 +28,15 @@ def main():
         version = versions.get(what)
         path = '/usr/lib/postgresql/{0}/bin:.'.format(version)
         unbuffer = ['timeout', '900', 'unbuffer']
-        args = ['--tags=-skip'] if what == 'etcd' else []
     else:
         path = os.path.abspath(os.path.join('pgsql', 'bin'))
         if sys.platform == 'darwin':
             path += ':.'
-        args = unbuffer = []
+        unbuffer = []
     env['PATH'] = path + os.pathsep + env['PATH']
     env['DCS'] = what
 
-    ret = subprocess.call(unbuffer + [sys.executable, '-m', 'behave'] + args, env=env)
+    ret = subprocess.call(unbuffer + [sys.executable, '-m', 'behave'], env=env)
 
     if ret != 0:
         if subprocess.call('grep . features/output/*_failed/*postgres?.*', shell=True) != 0:

.github/workflows/tests.yaml

@@ -17,9 +17,9 @@ jobs:
         os: [ubuntu, windows, macos]
 
     steps:
-    - uses: actions/checkout@v1
+    - uses: actions/checkout@v3
     - name: Set up Python 2.7
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: 2.7
       if: matrix.os != 'windows'
@@ -31,7 +31,7 @@ jobs:
       if: matrix.os != 'windows'
 
     - name: Set up Python 3.6
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: 3.6
     - name: Install dependencies
@@ -40,7 +40,7 @@ jobs:
       run: python .github/workflows/run_tests.py
 
     - name: Set up Python 3.7
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: 3.7
     - name: Install dependencies
@@ -49,7 +49,7 @@ jobs:
       run: python .github/workflows/run_tests.py
 
     - name: Set up Python 3.8
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: 3.8
     - name: Install dependencies
@@ -58,7 +58,7 @@ jobs:
       run: python .github/workflows/run_tests.py
 
     - name: Set up Python 3.9
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: 3.9
     - name: Install dependencies
@@ -67,7 +67,7 @@ jobs:
       run: python .github/workflows/run_tests.py
 
     - name: Set up Python 3.10
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: '3.10'
     - name: Install dependencies
@@ -115,9 +115,9 @@ jobs:
           dcs: etcd3
 
     steps:
-    - uses: actions/checkout@v1
+    - uses: actions/checkout@v3
     - name: Set up Python
-      uses: actions/setup-python@v2
+      uses: actions/setup-python@v4
       with:
         python-version: ${{ matrix.python-version }}
     - name: Add postgresql apt repo
@@ -127,7 +127,7 @@ jobs:
       run: python .github/workflows/install_deps.py
     - name: Run behave tests
       run: python .github/workflows/run_tests.py
-    - uses: actions/setup-python@v2
+    - uses: actions/setup-python@v4
       with:
         python-version: '3.10'
     - name: Install coveralls
@@ -144,7 +144,7 @@ jobs:
     needs: [unit, behave]
     runs-on: ubuntu-latest
     steps:
-    - uses: actions/setup-python@v2
+    - uses: actions/setup-python@v4
     - run: python -m pip install coveralls
     - run: python -m coveralls --service=github --finish
       env:

CODEOWNERS

@@ -0,0 +1,2 @@
+# global owners
+*        @CyberDem0n @hughcapet

Dockerfile

@@ -1,6 +1,6 @@
 ## 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
-ARG PG_MAJOR=10
+ARG PG_MAJOR=15
 ARG COMPRESS=false
 ARG PGHOME=/home/postgres
 ARG PGDATA=$PGHOME/data
@@ -117,7 +117,7 @@ RUN if [ "$COMPRESS" = "true" ]; then \
 FROM scratch
 COPY --from=builder / /
 
-LABEL maintainer="Alexander Kukushkin <alexander.kukushkin@zalando.de>"
+LABEL maintainer="Alexander Kukushkin <akukushkin@microsoft.com>"
 
 ARG PG_MAJOR
 ARG COMPRESS

MAINTAINERS

@@ -1,3 +1,2 @@
-Alexander Kukushkin <alexander.kukushkin@zalando.de>
-Feike Steenbergen <feike.steenbergen@zalando.de>
-Oleksii Kliukin <[email protected]>
+Alexander Kukushkin <akukushkin@microsoft.com>
+Polina Bungina <polina.bungina@zalando.de>

README.rst

@@ -12,7 +12,7 @@ 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 14.
+Currently supported PostgreSQL versions: 9.3 to 15.
 
 **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.
 

docs/ENVIRONMENT.rst

@@ -49,6 +49,7 @@ Consul
 -  **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
+-  **PATRONI\_CONSUL\_SERVICE\_CHECK\_TLS\_SERVER\_NAME**: (optional) overide SNI host when connecting via TLS, see also `consul agent check API reference <https://www.consul.io/api-docs/agent/check#tlsservername>`__.
 
 Etcd
 ----
@@ -122,11 +123,12 @@ PostgreSQL
 ----------
 -  **PATRONI\_POSTGRESQL\_LISTEN**: IP address + port that Postgres listens to. Multiple comma-separated addresses are permitted, as long as the port component is appended after to the last one with a colon, i.e. ``listen: 127.0.0.1,127.0.0.2:5432``. Patroni will use the first address from this list to establish local connections to the PostgreSQL node.
 -  **PATRONI\_POSTGRESQL\_CONNECT\_ADDRESS**: IP address + port through which Postgres is accessible from other nodes and applications.
+-  **PATRONI\_POSTGRESQL\_PROXY\_ADDRESS**: IP address + port through which a connection pool (e.g. pgbouncer) running next to Postgres is accessible. The value is written to the member key in DCS as ``proxy_url`` and could be used/useful for service discovery.
 -  **PATRONI\_POSTGRESQL\_DATA\_DIR**: The location of the Postgres data directory, either existing or to be initialized by Patroni.
 -  **PATRONI\_POSTGRESQL\_CONFIG\_DIR**: The location of the Postgres configuration directory, defaults to the data directory. Must be writable by Patroni.
 -  **PATRONI\_POSTGRESQL\_BIN_DIR**: Path to PostgreSQL binaries. (pg_ctl, pg_rewind, pg_basebackup, postgres) The  default value is an empty string meaning that PATH environment variable will be used to find the executables.
 -  **PATRONI\_POSTGRESQL\_PGPASS**: path to the `.pgpass <https://www.postgresql.org/docs/current/static/libpq-pgpass.html>`__ password file. Patroni creates this file before executing pg\_basebackup and under some other circumstances. The location must be writable by Patroni.
--  **PATRONI\_REPLICATION\_USERNAME**: replication username; the user will be created during initialization. Replicas will use this user to access master via streaming replication
+-  **PATRONI\_REPLICATION\_USERNAME**: replication username; the user will be created during initialization. Replicas will use this user to access the replication source 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.

docs/README.rst

@@ -79,7 +79,7 @@ Patroni can be installed with pip:
 where dependencies can be either empty, or consist of one or more of the following:
 
 etcd or etcd3
-    `python-etcd` module in order to use Etcd as DCS
+    `python-etcd` module in order to use Etcd as Distributed Configuration Store (DCS)
 consul
     `python-consul` module in order to use Consul as DCS
 zookeeper
@@ -109,7 +109,7 @@ 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 
+there is no requirement on the minimal number of nodes. Running a cluster consisting of one primary and one standby is 
 perfectly fine. You can add more standby nodes later.
 
 Running and Configuring

docs/SETTINGS.rst

@@ -17,21 +17,21 @@ Dynamic configuration is stored in the DCS (Distributed Configuration Store) and
 -  **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\_start\_timeout**: the amount of time a primary 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 primary 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.
+-  **synchronous\_mode\_strict**: prevents disabling synchronous replication if no synchronous replicas are available, blocking all client writes to the primary. 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`
+    -  **host**: an address of remote node
+    -  **port**: a port of remote node
+    -  **primary\_slot\_name**: which slot on the remote node 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 primary, can be different from the list defined in :ref:`postgresql_settings`
+    -  **restore\_command**: command to restore WAL records from the remote primary to nodes in a standby cluster, 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. The logical slots are copied from the primary to a standby with restart, and after that their position advanced every **loop_wait** seconds (if necessary). Copying logical slot files performed via ``libpq`` connection and using either rewind or superuser credentials (see **postgresql.authentication** section). There is always a chance that the logical slot position on the replica is a bit behind the former primary, therefore application should be prepared that some messages could be received the second time after the failover. The easiest way of doing so - tracking ``confirmed_flush_lsn``. Enabling permanent logical replication slots requires **postgresql.use_slots** to be set and will also automatically enable the ``hot_standby_feedback``. Since the failover of logical replication slots is unsafe on PostgreSQL 9.6 and older and PostgreSQL version 10 is missing some important functions, the feature only works with PostgreSQL 11+.
@@ -131,7 +131,8 @@ Most of the parameters are optional, but you have to specify one of the **host**
 -  **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.
+-  **service\_check\_interval**: (optional) how often to perform health check against registered url. Defaults to '5s'.
+-  **service\_check\_tls\_server\_name**: (optional) overide SNI host when connecting via TLS, see also `consul agent check API reference <https://www.consul.io/api-docs/agent/check#tlsservername>`__.
 
 The ``token`` needs to have the following ACL permissions:
 
@@ -223,7 +224,7 @@ Raft
 
   - 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
+    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?
 
@@ -261,7 +262,7 @@ PostgreSQL
             -  **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
+            -  **username**: replication username; the user will be created during initialization. Replicas will use this user to access the replication source 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.
@@ -291,6 +292,7 @@ PostgreSQL
             -  **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.
+    -  **proxy\_address**: IP address + port through which a connection pool (e.g. pgbouncer) running next to Postgres is accessible. The value is written to the member key in DCS as ``proxy_url`` and could be used/useful for service discovery.
     -  **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.
@@ -313,14 +315,14 @@ PostgreSQL
     -  **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**.
+    -  **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 primary can not start streaming from the new primary. This option is useful when ``pg_rewind`` can not be used. While performing timelines divergence check on PostgreSQL v10 and older Patroni will try to connect with replication credential to the "postgres" database. Hence, such access should be allowed in the  pg_hba.conf. 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
 --------
 -  **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.
+        -  **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 leader 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).
 

docs/dynamic_configuration.rst

@@ -22,7 +22,7 @@ Patroni configuration is stored in the DCS (Distributed Configuration Store). Th
 
 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:
+Some of the PostgreSQL parameters must hold the same values on the primary 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
 - max_locks_per_transaction: 64
@@ -32,7 +32,7 @@ Some of the PostgreSQL parameters must hold the same values on the master and th
 - wal_log_hints: on
 - track_commit_timestamp: off
 
-For the parameters below, PostgreSQL does not require equal values among the master and all the replicas. However, considering the possibility of a replica to become the master at any time, it doesn't really make sense to set them differently; therefore, Patroni restricts setting their values to the Dynamic configuration
+For the parameters below, PostgreSQL does not require equal values among the primary and all the replicas. However, considering the possibility of a replica to become the primary at any time, it doesn't really make sense to set them differently; therefore, Patroni restricts setting their values to the Dynamic configuration
 
 - max_wal_senders: 5
 - max_replication_slots: 5
@@ -86,4 +86,4 @@ Also, the following Patroni configuration options can be changed only dynamicall
 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.
+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 leader is allowed to restore these options from the on-disk dump if these are completely absent from the DCS or if they are invalid.

docs/existing_data.rst

@@ -29,11 +29,11 @@ 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
+2. Upgrade PostgreSQL binaries and perform `pg_upgrade <https://www.postgresql.org/docs/current/pgupgrade.html>`_ on the primary 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.
+6. Start Patroni on the primary 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.
 

docs/index.rst

@@ -10,7 +10,7 @@ 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 14.
+Currently supported PostgreSQL versions: 9.3 to 15.
 
 **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.
 

docs/kubernetes.rst

@@ -23,7 +23,7 @@ Use ConfigMaps
 In this mode, Patroni will create ConfigMaps instead of Endpoints and store keys inside meta-data of those ConfigMaps.
 Changing the leader takes at least two updates, one to the leader ConfigMap and another to the respective Endpoint.
 
-There are two ways to direct the traffic to the Postgres master:
+There are two ways to direct the traffic to the Postgres leader:
 
 - use the `callback script <https://github.com/zalando/patroni/blob/master/kubernetes/callback.py>`_ provided by Patroni
 - configure the Kubernetes Postgres service to use the label selector with the `role_label` (configured in patroni configuration).

docs/pause.rst

@@ -6,7 +6,7 @@ Pause/Resume mode for the cluster
 The goal
 --------
 
-Under certain circumstances Patroni needs to temporary step down from managing the cluster, while still retaining the cluster state in DCS. Possible use cases are uncommon activities on the cluster, such as major version upgrades or corruption recovery. During those activities nodes are often started and stopped for the reason unknown to Patroni, some nodes can be even temporary promoted, violating the assumption of running only one master. Therefore, Patroni needs to be able to "detach" from the running cluster, implementing an equivalent of the maintenance mode in Pacemaker.
+Under certain circumstances Patroni needs to temporarily step down from managing the cluster, while still retaining the cluster state in DCS. Possible use cases are uncommon activities on the cluster, such as major version upgrades or corruption recovery. During those activities nodes are often started and stopped for reasons unknown to Patroni, some nodes can be even temporarily promoted, violating the assumption of running only one primary. Therefore, Patroni needs to be able to "detach" from the running cluster, implementing an equivalent of the maintenance mode in Pacemaker.
 
 
 
@@ -17,16 +17,18 @@ When Patroni runs in a paused mode, it does not change the state of PostgreSQL,
 
 - For each node, the member key in DCS is updated with the current information about the cluster. This causes Patroni to run read-only queries on a member node if the member is running.
 
-- For the Postgres master with the leader lock Patroni updates the lock. If the node with the leader lock stops being the master (i.e. is demoted manually), Patroni will release the lock instead of promoting the node back.
+- For the Postgres primary with the leader lock Patroni updates the lock. If the node with the leader lock stops being the primary (i.e. is demoted manually), Patroni will release the lock instead of promoting the node back.
 
-- Manual unscheduled restart, reinitialize and manual failover are allowed. Manual failover is only allowed if the node to failover to is specified. In the paused mode, manual failover does not require a running master node.
+- Manual unscheduled restart, reinitialize and manual failover are allowed. Manual failover is only allowed if the node to failover to is specified. In the paused mode, manual failover does not require a running primary node.
 
-- If 'parallel' masters are detected by Patroni, it emits a warning, but does not demote the masters without the leader lock.
+- If 'parallel' primaries are detected by Patroni, it emits a warning, but does not demote the primary without the leader lock.
 
-- If there is no leader lock in the cluster, the running master acquires the lock. If there is more than one master node, then the first master to acquire the lock wins. If there are no masters altogether, Patroni does not try to promote any replicas. There is an exception in this rule: if there is no leader lock because the old master has demoted itself due to the manual promotion, then only the candidate node mentioned in the promotion request may take the leader lock. When the new leader lock is granted (i.e. after promoting a replica manually), Patroni makes sure the replicas that were streaming from the previous leader will switch to the new one.
+- If there is no leader lock in the cluster, the running primary acquires the lock. If there is more than one primary node, then the first primary to acquire the lock wins. If there are no primary altogether, Patroni does not try to promote any replicas. There is an exception in this rule: if there is no leader lock because the old primary has demoted itself due to the manual promotion, then only the candidate node mentioned in the promotion request may take the leader lock. When the new leader lock is granted (i.e. after promoting a replica manually), Patroni makes sure the replicas that were streaming from the previous leader will switch to the new one.
 
 - When Postgres is stopped, Patroni does not try to start it. When Patroni is stopped, it does not try to stop the Postgres instance it is managing.
 
+- Patroni will not try to remove replication slots that don't represent the other cluster member or are not listed in the configuration of the permanent slots.
+
 User guide
 ----------
 

docs/releases.rst

@@ -3,6 +3,176 @@
 Release notes
 =============
 
+Version 2.1.5
+-------------
+
+This version enhances compatibility with PostgreSQL 15 and declares Etcd v3 support as production ready. The Patroni on Raft remains in Beta.
+
+**New features**
+
+- Improve ``patroni --validate-config`` (Denis Laxalde)
+
+  Exit with code 1 if config is invalid and print errors to stderr.
+
+- Don't drop replication slots in pause (Alexander Kukushkin)
+
+  Patroni is automatically creating/removing physical replication slots when members are joining/leaving the cluster. In pause slots will no longer be removed.
+
+- Support the ``HEAD`` request method for monitoring endpoints (Robert Cutajar)
+
+  If used instead of ``GET`` Patroni will return only the HTTP Status Code.
+
+- Support behave tests on Windows (Alexander)
+
+  Emulate graceful Patroni shutdown (``SIGTERM``) on Windows by introduce the new REST API endpoint ``POST /sigterm``.
+
+- Introduce ``postgresql.proxy_address`` (Alexander)
+
+  It will be written to the member key in DCS as the ``proxy_url`` and could be used/useful for service discovery.
+
+
+**Stability improvements**
+
+- Call ``pg_replication_slot_advance()`` from a thread (Alexander)
+
+  On busy clusters with many logical replication slots the ``pg_replication_slot_advance()`` call was affecting the main HA loop and could result in the member key expiration.
+
+- Archive possibly missing WALs before calling ``pg_rewind`` on the old primary (Polina Bungina)
+
+  If the primary crashed and was down during considerable time, some WAL files could be missing from archive and from the new primary. There is a chance that ``pg_rewind`` could remove these WAL files from the old primary making it impossible to start it as a standby. By archiving ``ready`` WAL files we not only mitigate this problem but in general improving continues archiving experience.
+
+- Ignore ``403`` errors when trying to create Kubernetes Service (Nick Hudson, Polina)
+
+  Patroni was spamming logs by unsuccessful attempts to create the service, which in fact could already exist.
+
+- Improve liveness probe (Alexander)
+
+  The liveness problem will start failing if the heartbeat loop is running longer than `ttl` on the primary or `2*ttl` on the replica. That will allow us to use it as an alternative for :ref:`watchdog <watchdog>` on Kubernetes.
+
+- Make sure only sync node tries to grab the lock when switchover (Alexander, Polina)
+
+  Previously there was a slim chance that up-to-date async member could become the leader if the manual switchover was performed without specifying the target.
+
+- Avoid cloning while bootstrap is running (Ants Aasma)
+
+  Do not allow a create replica method that does not require a leader to be triggered while the cluster bootstrap is running.
+
+- Compatibility with kazoo-2.9.0 (Alexander)
+
+  Depending on python version the ``SequentialThreadingHandler.select()`` method may raise ``TypeError`` and ``IOError`` exceptions if ``select()`` is called on the closed socket.
+
+- Explicitly shut down SSL connection before socket shutdown (Alexander)
+
+  Not doing it resulted in ``unexpected eof while reading`` errors with OpenSSL 3.0.
+
+- Compatibility with `prettytable>=2.2.0` (Alexander)
+
+  Due to the internal API changes the cluster name header was shown on the incorrect line.
+
+
+**Bugfixes**
+
+- Handle expired token for Etcd lease_grant (monsterxx03)
+
+  In case of error get the new token and retry request.
+
+- Fix bug in the ``GET /read-only-sync`` endpoint (Alexander)
+
+  It was introduced in previous release and effectively never worked.
+
+- Handle the case when data dir storage disappeared (Alexander)
+
+  Patroni is periodically checking that the PGDATA is there and not empty, but in case of issues with storage the ``os.listdir()`` is raising the ``OSError`` exception, breaking the heart-beat loop.
+
+- Apply ``master_stop_timeout`` when waiting for user backends to close (Alexander)
+
+  Something that looks like user backend could be in fact a background worker (e.g., Citus Maintenance Daemon) that is failing to stop.
+
+- Accept ``*:<port>`` for ``postgresql.listen`` (Denis)
+
+  The ``patroni --validate-config`` was complaining about it being invalid.
+
+- Timeouts fixes in Raft (Alexander)
+
+  When Patroni or patronictl are starting they try to get Raft cluster topology from known members. These calls were made without proper timeouts.
+
+- Forcefully update consul service if token was changed (John A. Lotoski)
+
+  Not doing so results in errors "rpc error making call: rpc error making call: ACL not found".
+
+
+Version 2.1.4
+-------------
+
+**New features**
+
+- Improve ``pg_rewind`` behavior on typical Debian/Ubuntu systems (Gunnar "Nick" Bluth)
+
+  On Postgres setups that keep `postgresql.conf` outside of the data directory (e.g. Ubuntu/Debian packages), ``pg_rewind --restore-target-wal``  fails to figure out the value of the ``restore_command``.
+
+- Allow setting ``TLSServerName`` on Consul service checks (Michael Gmelin)
+
+  Useful when checks are performed by IP and the Consul ``node_name`` is not a FQDN.
+
+- Added ``ppc64le`` support in watchdog (Jean-Michel Scheiwiler)
+
+  And fixed watchdog support on some non-x86 platforms.
+
+- Switched aws.py callback from ``boto`` to ``boto3`` (Alexander Kukushkin)
+
+ ``boto``  2.x is abandoned since 2018 and fails with python 3.9.
+
+- Periodically refresh service account token on K8s (Haitao Li)
+
+  Since Kubernetes v1.21 service account tokens expire in 1 hour.
+
+- Added ``/read-only-sync`` monitoring endpoint (Dennis4b)
+
+  It is similar to the ``/read-only`` but includes only synchronous replicas.
+
+
+**Stability improvements**
+
+- Don't copy the logical replication slot to a replica if there is a configuration mismatch in the logical decoding setup with the primary (Alexander)
+
+  A replica won't copy a logical replication slot from the primary anymore if the slot doesn't match the ``plugin`` or ``database`` configuration options. Previously, the check for whether the slot matches those configuration options was not performed until after the replica copied the slot and started with it, resulting in unnecessary and repeated restarts.
+
+- Special handling of recovery configuration parameters for PostgreSQL v12+ (Alexander)
+
+  While starting as replica Patroni should be able to update ``postgresql.conf`` and restart/reload if the leader address has changed by caching current parameters values instead of querying them from ``pg_settings``.
+
+- Better handling of IPv6 addresses in the ``postgresql.listen`` parameters (Alexander)
+
+  Since the ``listen`` parameter has a port, people try to put IPv6 addresses into square brackets, which were not correctly stripped when there is more than one IP in the list.
+
+- Use ``replication`` credentials when performing divergence check only on PostgreSQL v10 and older (Alexander)
+
+  If ``rewind`` is enabled, Patroni will again use either ``superuser`` or ``rewind`` credentials on newer Postgres versions.
+
+
+**Bugfixes**
+
+- Fixed missing import of ``dateutil.parser`` (Wesley Mendes)
+
+  Tests weren't failing only because it was also imported from other modules.
+
+- Ensure that ``optime`` annotation is a string (Sebastian Hasler)
+
+  In certain cases Patroni was trying to pass it as numeric.
+
+- Better handling of failed ``pg_rewind`` attempt (Alexander)
+
+  If the primary becomes unavailable during ``pg_rewind``, ``$PGDATA`` will be left in a broken state. Following that,  Patroni will remove the data directory even if this is not allowed by the configuration.
+
+- Don't remove ``slots`` annotations from the leader ``ConfigMap``/``Endpoint`` when PostgreSQL isn't ready (Alexander)
+
+  If ``slots`` value isn't passed the annotation will keep the current value.
+
+- Handle concurrency problem with K8s API watchers (Alexander)
+
+  Under certain (unknown) conditions watchers might become stale; as a result, ``attempt_to_acquire_leader()`` method could fail due to the HTTP status code 409. In that case we reset watchers connections and restart from scratch.
+
+
 Version 2.1.3
 -------------
 

docs/replica_bootstrap.rst

@@ -63,7 +63,7 @@ Building replicas
 -----------------
 
 Patroni uses tried and proven ``pg_basebackup`` in order to create new replicas. One downside of it is that it requires
-a running master node. Another one is the lack of 'on-the-fly' compression for the backup data and no built-in cleanup
+a running leader node. Another one is the lack of 'on-the-fly' compression for the backup data and no built-in cleanup
 for outdated backup files. Some people prefer other backup solutions, such as ``WAL-E``, ``pgBackRest``, ``Barman`` and
 others, or simply roll their own scripts. In order to accommodate all those use-cases Patroni supports running custom
 scripts to clone a new replica. Those are configured in the ``postgresql`` configuration block:
@@ -123,11 +123,11 @@ to execute and any custom parameters that should be passed to that command. All
 --role
     Always 'replica'
 --connstring
-    Connection string to connect to the cluster member to clone from (master or other replica). The user in the
+    Connection string to connect to the cluster member to clone from (primary or other replica). The user in the
     connection string can execute SQL and replication protocol commands.
 
 A special ``no_master`` parameter, if defined, allows Patroni to call the replica creation method even if there is no
-running master or replicas. In that case, an empty string will be passed in a connection string. This is useful for
+running leader 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.
@@ -137,7 +137,7 @@ A special ``no_params`` parameter, if defined, restricts passing parameters to c
 A ``basebackup`` method is a special case: it will be used if
 ``create_replica_methods`` is empty, although it is possible
 to list it explicitly among the ``create_replica_methods`` methods. This method initializes a new replica with the
-``pg_basebackup``, the base backup is taken from the master unless there are replicas with ``clonefrom`` tag, in which case one
+``pg_basebackup``, the base backup is taken from the leader unless there are replicas with ``clonefrom`` tag, in which case one
 of such replicas will be used as the origin for pg_basebackup. It works without any configuration; however, it is
 possible to specify a ``basebackup`` configuration section. Same rules as with the other method configuration apply,
 namely, only long (with --) options should be specified there. Not all parameters make sense, if you override a connection
@@ -176,10 +176,10 @@ 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 nodes replicating from some remote node. This type of clusters has:
 
 * "standby leader", that behaves pretty much like a regular cluster leader,
-  except it replicates from a remote master.
+  except it replicates from a remote node.
 
 * cascade replicas, that are replicating from standby leader.
 
@@ -187,6 +187,13 @@ 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.
 
+There is no further relationship between the standby cluster and the primary
+cluster it replicates from, in particular, they must not share the same DCS
+scope if they use the same DCS. They do not know anything else from each other
+apart from replication information. Also, the standby cluster is not being
+displayed in ``patronictl list`` or ``patronictl topology`` output on the
+primary cluster.
+
 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
@@ -212,4 +219,9 @@ in a patroni configuration:
 Note, that these options will be applied only once during cluster bootstrap,
 and the only way to change them afterwards is through DCS.
 
+Patroni expects to find `postgresql.conf` or `postgresql.conf.backup` in PGDATA
+of the remote primary and will not start if it does not find it after a
+basebackup. If the remote primary keeps its `postgresql.conf` elsewhere, it is
+your responsibility to copy it to PGDATA.
+
 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

@@ -13,7 +13,7 @@ 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``.
+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 primary become the new leader by changing the value of ``check_timeline`` parameter to ``true``.
 
 PostgreSQL synchronous replication
 ----------------------------------

docs/rest_api.rst

@@ -7,7 +7,7 @@ Patroni has a rich REST API, which is used by Patroni itself during the leader r
 
 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``.
+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 ``HEAD`` or ``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 primary with leader lock:
 
@@ -44,8 +44,11 @@ For all health check ``GET`` requests Patroni returns a JSON document with the s
 
 - ``GET /synchronous`` or ``GET /sync``: returns HTTP status code **200** only when the Patroni node is running as a synchronous standby.
 
+- ``GET /read-only-sync``: like the above endpoint, but also includes the primary.
+
 - ``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``
@@ -55,7 +58,7 @@ For all health check ``GET`` requests Patroni returns a JSON document with the s
 
 - ``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 /liveness``: returns HTTP status code **200** if Patroni heartbeat loop is properly running and **503** if the last run was more than ``ttl`` seconds ago on the primary or ``2*ttl`` on the replica. 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 Kubernetes endpoints for leader elections (OpenShift).
 

docs/watchdog.rst

@@ -3,7 +3,7 @@
 Watchdog support
 ================
 
-Having multiple PostgreSQL servers running as master can result in transactions lost due to diverging timelines. This situation is also called a split-brain problem. To avoid split-brain Patroni needs to ensure PostgreSQL will not accept any transaction commits after leader key expires in the DCS. Under normal circumstances Patroni will try to achieve this by stopping PostgreSQL when leader lock update fails for any reason. However, this may fail to happen due to various reasons:
+Having multiple PostgreSQL servers running as primary can result in transactions lost due to diverging timelines. This situation is also called a split-brain problem. To avoid split-brain Patroni needs to ensure PostgreSQL will not accept any transaction commits after leader key expires in the DCS. Under normal circumstances Patroni will try to achieve this by stopping PostgreSQL when leader lock update fails for any reason. However, this may fail to happen due to various reasons:
 
 - Patroni has crashed due to a bug, out-of-memory condition or by being accidentally killed by a system administrator.
 
@@ -13,7 +13,7 @@ Having multiple PostgreSQL servers running as master can result in transactions
 
 To guarantee correct behavior under these conditions Patroni supports watchdog devices. Watchdog devices are software or hardware mechanisms that will reset the whole system when they do not get a keepalive heartbeat within a specified timeframe. This adds an additional layer of fail safe in case usual Patroni split-brain protection mechanisms fail.
 
-Patroni will try to activate the watchdog before promoting PostgreSQL to master. If watchdog activation fails and watchdog mode is ``required`` then the node will refuse to become master. When deciding to participate in leader election Patroni will also check that watchdog configuration will allow it to become leader at all. After demoting PostgreSQL (for example due to a manual failover) Patroni will disable the watchdog again. Watchdog will also be disabled while Patroni is in paused state.
+Patroni will try to activate the watchdog before promoting PostgreSQL to primary. If watchdog activation fails and watchdog mode is ``required`` then the node will refuse to become leader. When deciding to participate in leader election Patroni will also check that watchdog configuration will allow it to become leader at all. After demoting PostgreSQL (for example due to a manual failover) Patroni will disable the watchdog again. Watchdog will also be disabled while Patroni is in paused state.
 
 By default Patroni will set up the watchdog to expire 5 seconds before TTL expires. With the default setup of ``loop_wait=10`` and ``ttl=30`` this gives HA loop at least 15 seconds (``ttl`` - ``safety_margin`` - ``loop_wait``) to complete before the system gets forcefully reset. By default accessing DCS is configured to time out after 10 seconds. This means that when DCS is unavailable, for example due to network issues, Patroni and PostgreSQL will have at least 5 seconds (``ttl`` - ``safety_margin`` - ``loop_wait`` - ``retry_timeout``) to come to a state where all client connections are terminated.
 

extras/confd/templates/haproxy.tmpl

@@ -18,14 +18,14 @@ listen stats
 
 listen master
 	bind *:5000
-	option httpchk OPTIONS /master
+	option httpchk HEAD /master
 	http-check expect status 200
 	default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
 {{range gets "/members/*"}}	server {{base .Key}} {{$data := json .Value}}{{base (replace (index (split $data.conn_url "/") 2) "@" "/" -1)}} maxconn 100 check port {{index (split (index (split $data.api_url "/") 2) ":") 1}}
 {{end}}
 listen replicas
 	bind *:5001
-	option httpchk OPTIONS /replica
+	option httpchk HEAD /replica
 	http-check expect status 200
 	default-server inter 3s fall 3 rise 2 on-marked-down shutdown-sessions
 {{range gets "/members/*"}}	server {{base .Key}} {{$data := json .Value}}{{base (replace (index (split $data.conn_url "/") 2) "@" "/" -1)}} maxconn 100 check port {{index (split (index (split $data.api_url "/") 2) ":") 1}}

extras/startup-scripts/patroni.service

@@ -14,7 +14,7 @@ Group=postgres
 # Read in configuration file if it exists, otherwise proceed
 EnvironmentFile=-/etc/patroni_env.conf
 
-# the default is the user's home directory, and if you want to change it, you must provide an absolute path.
+# 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
@@ -32,14 +32,14 @@ ExecStart=/bin/patroni /etc/patroni.yml
 # Send HUP to reload from patroni.yml
 ExecReload=/bin/kill -s HUP $MAINPID
 
-# only kill the patroni process, not it's children, so it will gracefully stop postgres
+# Only kill the patroni process, not it's children, so it will gracefully stop postgres
 KillMode=process
 
 # Give a reasonable amount of time for the server to start up/shut down
 TimeoutSec=30
 
-# Do not restart the service if it crashes, we want to manually inspect database on failure
-Restart=no
+# Restart the service if it crashed
+Restart=on-failure
 
 [Install]
 WantedBy=multi-user.target

features/basic_replication.feature

@@ -5,7 +5,7 @@ Feature: basic replication
     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}
+    When I issue a PATCH request to http://127.0.0.1:8008/config with {"ttl": 20, "synchronous_mode": true}
     Then I receive a response code 200
     When I start postgres1
     And I configure and start postgres2 with a tag replicatefrom postgres0

features/environment.py

@@ -14,6 +14,7 @@ import yaml
 
 import patroni.psycopg as psycopg
 
+from patroni.request import PatroniRequest
 from six.moves.BaseHTTPServer import BaseHTTPRequestHandler, HTTPServer
 
 
@@ -138,12 +139,24 @@ class PatroniController(AbstractController):
     def _start(self):
         if self.watchdog:
             self.watchdog.start()
+        env = os.environ.copy()
         if isinstance(self._context.dcs_ctl, KubernetesController):
             self._context.dcs_ctl.create_pod(self._name[8:], self._scope)
-            os.environ['PATRONI_KUBERNETES_POD_IP'] = '10.0.0.' + self._name[-1]
-        return subprocess.Popen([sys.executable, '-m', 'coverage', 'run',
-                                '--source=patroni', '-p', 'patroni.py', self._config],
-                                stdout=self._log, stderr=subprocess.STDOUT, cwd=self._work_directory)
+            env['PATRONI_KUBERNETES_POD_IP'] = '10.0.0.' + self._name[-1]
+        if os.name == 'nt':
+            env['BEHAVE_DEBUG'] = 'true'
+        patroni = subprocess.Popen([sys.executable, '-m', 'coverage', 'run',
+                                   '--source=patroni', '-p', 'patroni.py', self._config], env=env,
+                                   stdout=self._log, stderr=subprocess.STDOUT, cwd=self._work_directory)
+        if os.name == 'nt':
+            patroni.terminate = self.terminate
+        return patroni
+
+    def terminate(self):
+        try:
+            self._context.request_executor.request('POST', self._restapi_url + '/sigterm')
+        except Exception:
+            pass
 
     def stop(self, kill=False, timeout=15, postgres=False):
         if postgres:
@@ -178,15 +191,16 @@ class PatroniController(AbstractController):
         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']['data_dir'] = self._data_dir.replace('\\', '/')
         config['postgresql']['basebackup'] = [{'checkpoint': 'fast'}]
         config['postgresql']['use_unix_socket'] = os.name != 'nt'  # windows doesn't yet support unix-domain sockets
-        config['postgresql']['use_unix_socket_repl'] = os.name != 'nt'  # windows doesn't yet support unix-domain sockets
-        config['postgresql']['pgpass'] = os.path.join(tempfile.gettempdir(), 'pgpass_' + name)
+        config['postgresql']['use_unix_socket_repl'] = os.name != 'nt'
+        config['postgresql']['pgpass'] = os.path.join(tempfile.gettempdir(), 'pgpass_' + name).replace('\\', '/')
         config['postgresql']['parameters'].update({
-            'logging_collector': 'on', 'log_destination': 'csvlog', 'log_directory': self._output_dir,
+            'logging_collector': 'on', 'log_destination': 'csvlog',
+            'log_directory': self._output_dir.replace('\\', '/'),
             'log_filename': name + '.log', 'log_statement': 'all', 'log_min_messages': 'debug1',
-            'unix_socket_directories': self._data_dir})
+            'unix_socket_directories': tempfile.gettempdir().replace('\\', '/')})
 
         if 'bootstrap' in config:
             config['bootstrap']['post_bootstrap'] = 'psql -w -c "SELECT 1"'
@@ -197,7 +211,7 @@ class PatroniController(AbstractController):
             self.recursive_update(config, custom_config)
 
         self.recursive_update(config, {
-            'bootstrap': {'dcs': {'postgresql': {'parameters': {'wal_keep_segments': 100}}}}})
+            'bootstrap': {'dcs': {'loop_wait': 2, 'postgresql': {'parameters': {'wal_keep_segments': 100}}}}})
         if config['postgresql'].get('callbacks', {}).get('on_role_change'):
             config['postgresql']['callbacks']['on_role_change'] += ' ' + str(self.__PORT)
 
@@ -210,6 +224,7 @@ class PatroniController(AbstractController):
 
         self._replication = config['postgresql'].get('authentication', config['postgresql']).get('replication', {})
         self._replication.update({'host': host, 'port': self.__PORT, 'dbname': 'postgres'})
+        self._restapi_url = 'http://{0}'.format(config['restapi']['connect_address'])
 
         return patroni_config_path
 
@@ -394,7 +409,7 @@ class AbstractEtcdController(AbstractDcsController):
         self._client_cls = client_cls
 
     def _start(self):
-        return subprocess.Popen(["etcd", "--debug", "--data-dir", self._work_directory],
+        return subprocess.Popen(["etcd", "--data-dir", self._work_directory],
                                 stdout=self._log, stderr=subprocess.STDOUT)
 
     def _is_running(self):
@@ -626,15 +641,17 @@ class RaftController(AbstractDcsController):
             self.start()
 
         ready_event = threading.Event()
-        self._raft = KVStoreTTL(ready_event.set, None, None, partner_addrs=[self.CONTROLLER_ADDR], password=self.PASSWORD)
+        self._raft = KVStoreTTL(ready_event.set, None, None,
+                                partner_addrs=[self.CONTROLLER_ADDR], password=self.PASSWORD)
         self._raft.startAutoTick()
         ready_event.wait()
 
 
 class PatroniPoolController(object):
 
-    BACKUP_SCRIPT = [sys.executable, 'features/backup_create.py']
-    ARCHIVE_RESTORE_SCRIPT = ' '.join((sys.executable, os.path.abspath('features/archive-restore.py')))
+    PYTHON = sys.executable.replace('\\', '/')
+    BACKUP_SCRIPT = [PYTHON, 'features/backup_create.py']
+    ARCHIVE_RESTORE_SCRIPT = ' '.join((PYTHON, os.path.abspath('features/archive-restore.py')))
 
     def __init__(self, context):
         self._context = context
@@ -710,7 +727,7 @@ class PatroniPoolController(object):
                     'archive_mode': 'on',
                     'archive_command': (self.ARCHIVE_RESTORE_SCRIPT + ' --mode archive ' +
                                         '--dirname {} --filename %f --pathname %p').format(
-                                        os.path.join(self.patroni_path, 'data', 'wal_archive'))
+                                        os.path.join(self.patroni_path, 'data', 'wal_archive').replace('\\', '/'))
                 },
                 'authentication': {
                     'superuser': {'password': 'zalando1'},
@@ -726,14 +743,14 @@ class PatroniPoolController(object):
             'bootstrap': {
                 'method': 'backup_restore',
                 'backup_restore': {
-                    'command': (sys.executable + ' features/backup_restore.py --sourcedir=' +
-                                os.path.join(self.patroni_path, 'data', 'basebackup')),
+                    'command': (self.PYTHON + ' features/backup_restore.py --sourcedir=' +
+                                os.path.join(self.patroni_path, 'data', 'basebackup').replace('\\', '/')),
                     'recovery_conf': {
                         'recovery_target_action': 'promote',
                         'recovery_target_timeline': 'latest',
                         'restore_command': (self.ARCHIVE_RESTORE_SCRIPT + ' --mode restore ' +
                                             '--dirname {} --filename %f --pathname %p').format(
-                                            os.path.join(self.patroni_path, 'data', 'wal_archive'))
+                                            os.path.join(self.patroni_path, 'data', 'wal_archive').replace('\\', '/'))
                     }
                 }
             },
@@ -873,7 +890,10 @@ class WatchdogMonitor(object):
 # actions to execute on start/stop of the tests and before running individual features
 def before_all(context):
     os.environ.update({'PATRONI_RESTAPI_USERNAME': 'username', 'PATRONI_RESTAPI_PASSWORD': 'password'})
-    context.ci = any(a in os.environ for a in ('TRAVIS_BUILD_NUMBER', 'BUILD_NUMBER', 'GITHUB_ACTIONS'))
+    context.request_executor = PatroniRequest({'ctl': {'auth': os.environ['PATRONI_RESTAPI_USERNAME'] +
+                                               ':' + os.environ['PATRONI_RESTAPI_PASSWORD']}})
+    context.ci = os.name == 'nt' or\
+        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)
@@ -893,13 +913,26 @@ def after_all(context):
 
 def before_feature(context, feature):
     """ create per-feature output directory to collect Patroni and PostgreSQL logs """
-    context.pctl.create_and_set_output_directory(feature.name)
+    if feature.name == 'watchdog' and os.name == 'nt':
+        feature.skip("Watchdog isn't supported on Windows")
+    else:
+        context.pctl.create_and_set_output_directory(feature.name)
 
 
 def after_feature(context, feature):
     """ stop all Patronis, remove their data directory and cleanup the keys in etcd """
     context.pctl.stop_all()
-    shutil.rmtree(os.path.join(context.pctl.patroni_path, 'data'))
+    data = os.path.join(context.pctl.patroni_path, 'data')
+    if os.path.exists(data):
+        shutil.rmtree(data)
     context.dcs_ctl.cleanup_service_tree()
     if feature.status == 'failed':
         shutil.copytree(context.pctl.output_dir, context.pctl.output_dir + '_failed')
+
+
+def before_scenario(context, scenario):
+    if 'slot-advance' in scenario.effective_tags:
+        for p in context.pctl._processes.values():
+            if p._conn and p._conn.server_version < 110000:
+                scenario.skip('pg_replication_slot_advance() is not supported on {0}'.format(p._conn.server_version))
+                break

features/ignored_slots.feature

@@ -3,7 +3,7 @@ Feature: ignored slots
     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"}}}
+    When I issue a PATCH request to http://127.0.0.1:8009/config with {"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.

features/patroni_api.feature

@@ -35,13 +35,13 @@ Scenario: check local configuration reload
 	Then I receive a response code 202
 
 Scenario: check dynamic configuration change via DCS
-	Given I run patronictl.py edit-config -s 'ttl=10' -s 'loop_wait=2' -p 'max_connections=101' --force batman
+	Given I run patronictl.py edit-config -s 'ttl=10' -p 'max_connections=101' --force batman
 	Then I receive a response returncode 0
-	And I receive a response output "+loop_wait: 2"
+	And I receive a response output "+ttl: 10"
 	And Response on GET http://127.0.0.1:8008/patroni contains pending_restart after 11 seconds
 	When I issue a GET request to http://127.0.0.1:8008/config
 	Then I receive a response code 200
-	And I receive a response loop_wait 2
+	And I receive a response ttl 10
 	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 {'new_tag': 'new_value'}
@@ -109,7 +109,7 @@ Scenario: check the scheduled switchover
 	And I receive a response output "Can't schedule switchover in the paused state"
 	When I run patronictl.py resume batman
 	Then I receive a response returncode 0
-	Given I issue a scheduled switchover from postgres1 to postgres0 in 5 seconds
+	Given I issue a scheduled switchover from postgres1 to postgres0 in 10 seconds
 	Then I receive a response returncode 0
 	And postgres0 is a leader after 20 seconds
 	And postgres0 role is the primary after 10 seconds

features/standby_cluster.feature

@@ -3,7 +3,7 @@ Feature: standby cluster
     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"}}}
+    When I issue a PATCH request to http://127.0.0.1:8009/config with {"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
@@ -14,7 +14,7 @@ Feature: standby cluster
     Then "members/postgres0" key in DCS has state=running after 10 seconds
     And replication works from postgres1 to postgres0 after 15 seconds
 
-  @skip
+  @slot-advance
   Scenario: check permanent logical slots are synced to the replica
     Given I run patronictl.py restart batman postgres1 --force
     Then Logical slot test_logical is in sync between postgres0 and postgres1 after 10 seconds

features/steps/patroni_api.py

@@ -10,10 +10,8 @@ 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|\.|:|/)+')
@@ -75,9 +73,9 @@ def do_post_empty(context, url):
 def do_request(context, request_method, url, data):
     data = data and json.loads(data)
     try:
-        r = request_executor.request(request_method, url, data)
+        r = context.request_executor.request(request_method, url, data)
         if request_method == 'PATCH' and r.status == 409:
-            r = request_executor.request(request_method, url, data)
+            r = context.request_executor.request(request_method, url, data)
     except Exception:
         context.status_code = context.response = None
     else:
@@ -139,7 +137,7 @@ 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 = request_executor.request('GET', url)
+        r = context.request_executor.request('GET', url)
         if (value in r.data.decode('utf-8')) != negate:
             break
         time.sleep(1)

features/steps/standby_cluster.py

@@ -1,17 +1,12 @@
 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 "
+def callbacks(context, name):
+    return {c: '{0} features/callback2.py {1}'.format(context.pctl.PYTHON, name)
+            for c in ('on_start', 'on_stop', 'on_restart', 'on_role_change')}
 
 
 @step('I start {name:w} in a cluster {cluster_name:w}')
@@ -19,10 +14,10 @@ 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')},
+            "callbacks": callbacks(context, name),
             "backup_restore": {
-                "command": (executable + " features/backup_restore.py --sourcedir=" +
-                            os.path.join(context.pctl.patroni_path, 'data', 'basebackup'))}
+                "command": (context.pctl.PYTHON + " features/backup_restore.py --sourcedir=" +
+                            os.path.join(context.pctl.patroni_path, 'data', 'basebackup').replace('\\', '/'))}
         }
     })
 
@@ -49,7 +44,7 @@ def start_patroni_standby_cluster(context, name, cluster_name, name2):
             }
         },
         "postgresql": {
-            "callbacks": {c: callback + name for c in ('on_start', 'on_stop', 'on_restart', 'on_role_change')}
+            "callbacks": callbacks(context, name)
         }
     })
     return context.pctl.start(name)
@@ -62,7 +57,7 @@ def check_replication_status(context, pg_name1, pg_name2, timeout):
     while time.time() < bound_time:
         cur = context.pctl.query(
             pg_name2,
-            select_replication_query.format(pg_name1),
+            "SELECT * FROM pg_catalog.pg_stat_replication WHERE application_name = '{0}'".format(pg_name1),
             fail_ok=True
         )
 

kubernetes/Dockerfile

@@ -1,5 +1,5 @@
-FROM postgres:11
-MAINTAINER Alexander Kukushkin <alexander.kukushkin@zalando.de>
+FROM postgres:15
+LABEL maintainer="Alexander Kukushkin <akukushkin@microsoft.com>"
 
 RUN export DEBIAN_FRONTEND=noninteractive \
     && echo 'APT::Install-Recommends "0";\nAPT::Install-Suggests "0";' > /etc/apt/apt.conf.d/01norecommend \

patroni/api.py

@@ -38,6 +38,7 @@ class RestApiHandler(BaseHTTPRequestHandler):
         self.log_request(status_code)
 
     def _write_response(self, status_code, body, content_type='text/html', headers=None):
+        # TODO: try-catch ConnectionResetError: [Errno 104] Connection reset by peer and log it in DEBUG level
         self.send_response(status_code)
         headers = headers or {}
         if content_type:
@@ -141,7 +142,7 @@ class RestApiHandler(BaseHTTPRequestHandler):
             ignore_tags = True
         elif 'replica' in path:
             status_code = replica_status_code
-        elif 'read-only' in path:
+        elif 'read-only' in path and 'sync' not 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
@@ -152,6 +153,11 @@ class RestApiHandler(BaseHTTPRequestHandler):
                 status_code = replica_status_code
             elif path in ('/async', '/asynchronous') and not is_synchronous:
                 status_code = replica_status_code
+            elif path in ('/read-only-sync', '/read-only-synchronous'):
+                if 200 in (primary_status_code, standby_leader_status_code):
+                    status_code = 200
+                elif is_synchronous:
+                    status_code = replica_status_code
 
         # check for user defined tags in query params
         if not ignore_tags and status_code == 200:
@@ -180,8 +186,20 @@ class RestApiHandler(BaseHTTPRequestHandler):
     def do_OPTIONS(self):
         self.do_GET(write_status_code_only=True)
 
+    def do_HEAD(self):
+        self.do_GET(write_status_code_only=True)
+
     def do_GET_liveness(self):
-        self._write_status_code_only(200)
+        patroni = self.server.patroni
+        is_primary = patroni.postgresql.role == 'master' and patroni.postgresql.is_running()
+        # We can tolerate Patroni problems longer on the replica.
+        # On the primary the liveness probe most likely will start failing only after the leader key expired.
+        # It should not be a big problem because replicas will see that the primary is still alive via REST API call.
+        liveness_threshold = patroni.dcs.ttl * (1 if is_primary else 2)
+
+        # In maintenance mode (pause) we are fine if heartbeat loop stuck.
+        status_code = 200 if patroni.ha.is_paused() or patroni.next_run + liveness_threshold > time.time() else 503
+        self._write_status_code_only(status_code)
 
     def do_GET_readiness(self):
         patroni = self.server.patroni
@@ -350,6 +368,14 @@ class RestApiHandler(BaseHTTPRequestHandler):
         self.server.patroni.sighup_handler()
         self._write_response(202, 'reload scheduled')
 
+    @check_access
+    def do_POST_sigterm(self):
+        """Only for behave testing on windows"""
+
+        if os.name == 'nt' and os.getenv('BEHAVE_DEBUG'):
+            self.server.patroni.api_sigterm()
+        self._write_response(202, 'shutdown scheduled')
+
     @staticmethod
     def parse_schedule(schedule, action):
         """ parses the given schedule and validates at """
@@ -807,32 +833,20 @@ class RestApiServer(ThreadingMixIn, HTTPServer, Thread):
                 else:
                     logger.error('Bad value in the "restapi.verify_client": %s', verify_client)
             self.__ssl_serial_number = self.get_certificate_serial_number()
-            self.socket = ctx.wrap_socket(self.socket, server_side=True)
+            self.socket = ctx.wrap_socket(self.socket, server_side=True, do_handshake_on_connect=False)
         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
+        enable_keepalive(request, 10, 3)
+        if hasattr(request, 'context'):  # SSLSocket
+            request.do_handshake()
         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)
+        if hasattr(request, 'context'):  # SSLSocket
+            request.unwrap()
+        super(RestApiServer, self).shutdown_request(request)
 
     def get_certificate_serial_number(self):
         if self.__ssl_options.get('certfile'):

patroni/config.py

@@ -32,7 +32,7 @@ _AUTH_ALLOWED_PARAMETERS = (
 
 def default_validator(conf):
     if not conf:
-        return "Config is empty."
+        raise ConfigParseError("Config is empty.")
 
 
 class Config(object):
@@ -102,9 +102,9 @@ class Config(object):
             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 validator:
-            error = validator(self._local_configuration)
-            if error:
-                raise ConfigParseError(error)
+            errors = validator(self._local_configuration)
+            if errors:
+                raise ConfigParseError("\n".join(errors))
 
         self.__effective_configuration = self._build_effective_configuration({}, self._local_configuration)
         self._data_dir = self.__effective_configuration.get('postgresql', {}).get('data_dir', "")
@@ -227,7 +227,8 @@ class Config(object):
                 for name, value in (value or {}).items():
                     if name == 'parameters':
                         config['postgresql'][name].update(self._process_postgresql_parameters(value))
-                    elif name not in ('connect_address', 'listen', 'data_dir', 'pgpass', 'authentication'):
+                    elif name not in ('connect_address', 'proxy_address', 'listen',
+                                      'config_dir', 'data_dir', 'pgpass', 'authentication'):
                         config['postgresql'][name] = deepcopy(value)
             elif name == 'standby_cluster':
                 for name, value in (value or {}).items():
@@ -271,7 +272,8 @@ class Config(object):
                                         'cafile', 'ciphers', 'verify_client', 'http_extra_headers',
                                         'https_extra_headers', 'allowlist', 'allowlist_include_members'])
         _set_section_values('ctl', ['insecure', 'cacert', 'certfile', 'keyfile', 'keyfile_password'])
-        _set_section_values('postgresql', ['listen', 'connect_address', 'config_dir', 'data_dir', 'pgpass', 'bin_dir'])
+        _set_section_values('postgresql', ['listen', 'connect_address', 'proxy_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'])
         _set_section_values('raft', ['data_dir', 'self_addr', 'partner_addrs', 'password', 'bind_addr'])
@@ -350,9 +352,9 @@ class Config(object):
                 name, suffix = (param[8:].split('_', 1) + [''])[:2]
                 if suffix in ('HOST', 'HOSTS', 'PORT', 'USE_PROXIES', 'PROTOCOL', 'SRV', 'SRV_SUFFIX', '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', 'SET_ACLS') and name:
+                              'REGISTER_SERVICE', 'SERVICE_CHECK_INTERVAL', 'SERVICE_CHECK_TLS_SERVER_NAME',
+                              'NAMESPACE', 'CONTEXT', 'USE_ENDPOINTS', 'SCOPE_LABEL', 'ROLE_LABEL', 'POD_IP',
+                              'PORTS', 'LABELS', 'BYPASS_API_SERVICE', 'KEY_PASSWORD', 'USE_SSL', 'SET_ACLS') and name:
                     value = os.environ.pop(param)
                     if suffix == 'PORT':
                         value = value and parse_int(value)

patroni/ctl.py

@@ -60,6 +60,18 @@ class PatronictlPrettyTable(PrettyTable):
         self.__hline_num = 0
         self.__hline = None
 
+    def __build_header(self, line):
+        header = self.__table_header[:len(line) - 2]
+        return "".join([line[0], header, line[1 + len(header):]])
+
+    def _stringify_hrule(self, *args, **kwargs):
+        ret = super(PatronictlPrettyTable, self)._stringify_hrule(*args, **kwargs)
+        where = args[1] if len(args) > 1 else kwargs.get('where')
+        if where == 'top_' and self.__table_header:
+            ret = self.__build_header(ret)
+            self.__hline_num += 1
+        return ret
+
     def _is_first_hline(self):
         return self.__hline_num == 0
 
@@ -71,8 +83,7 @@ class PatronictlPrettyTable(PrettyTable):
 
         # 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):]])
+            ret = self.__build_header(ret)
 
         self.__hline_num += 1
         return ret

patroni/daemon.py

@@ -1,3 +1,5 @@
+from __future__ import print_function
+
 import abc
 import os
 import signal
@@ -22,11 +24,15 @@ class AbstractPatroniDaemon(object):
     def sighup_handler(self, *args):
         self._received_sighup = True
 
-    def sigterm_handler(self, *args):
+    def api_sigterm(self):
         with self._sigterm_lock:
             if not self._received_sigterm:
                 self._received_sigterm = True
-                sys.exit()
+                return True
+
+    def sigterm_handler(self, *args):
+        if self.api_sigterm():
+            sys.exit()
 
     def setup_signal_handlers(self):
         self._received_sighup = False
@@ -83,16 +89,18 @@ def abstract_main(cls, validator=None):
                         help='Patroni may also read the configuration from the {0} environment variable'
                         .format(Config.PATRONI_CONFIG_VARIABLE))
     args = parser.parse_args()
+    validate_config = validator and args.validate_config
     try:
-        if validator and args.validate_config:
+        if 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()
+            print(e.value, file=sys.stderr)
+        if not validate_config:
+            parser.print_help()
         sys.exit(1)
 
     controller = cls(config)

patroni/dcs/__init__.py

@@ -1,5 +1,5 @@
 import abc
-import dateutil
+import dateutil.parser
 import importlib
 import inspect
 import json

patroni/dcs/consul.py

@@ -233,8 +233,10 @@ class Consul(AbstractDCS):
         if self._register_service:
             self._set_service_name()
         self._service_check_interval = config.get('service_check_interval', '5s')
+        self._service_check_tls_server_name = config.get('service_check_tls_server_name', None)
         if not self._ctl:
             self.create_session()
+        self._previous_loop_token = self._client.token
 
     def retry(self, *args, **kwargs):
         return self._retry.copy()(*args, **kwargs)
@@ -458,9 +460,12 @@ class Consul(AbstractDCS):
         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))
+        if self._service_check_tls_server_name is not None:
+            check['TLSServerName'] = self._service_check_tls_server_name
         tags = self._service_tags[:]
         tags.append(role)
         self._previous_loop_service_tags = self._service_tags
+        self._previous_loop_token = self._client.token
 
         params = {
             'service_id': '{0}/{1}'.format(self._scope, self._name),
@@ -497,6 +502,7 @@ class Consul(AbstractDCS):
         if (
             force or update or self._register_service != self._previous_loop_register_service
             or self._service_tags != self._previous_loop_service_tags
+            or self._client.token != self._previous_loop_token
         ):
             return self._update_service(new_data)
 

patroni/dcs/etcd.py

@@ -216,10 +216,13 @@ class AbstractEtcdClientWithFailover(etcd.Client):
                 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 not retry:
+                    if len(machines_cache) == 1:
+                        self.set_base_uri(self._base_uri)  # trigger Etcd3 watcher restart
+                    # 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.
+                    elif 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.")

patroni/dcs/etcd3.py

@@ -11,6 +11,7 @@ import time
 import urllib3
 
 from threading import Condition, Lock, Thread
+from urllib3.exceptions import ReadTimeoutError, ProtocolError
 
 from . import ClusterConfig, Cluster, Failover, Leader, Member, SyncState, TimelineHistory
 from .etcd import AbstractEtcdClientWithFailover, AbstractEtcd, catch_etcd_errors
@@ -315,6 +316,7 @@ class Etcd3Client(AbstractEtcdClientWithFailover):
     def prefix(self, key, retry=None):
         return self.range(key, prefix_range_end(key), retry)
 
+    @_handle_auth_errors
     def lease_grant(self, ttl, retry=None):
         return self.call_rpc('/lease/grant', {'TTL': ttl}, retry)['ID']
 
@@ -349,7 +351,7 @@ class Etcd3Client(AbstractEtcdClientWithFailover):
     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):
+    def watchrange(self, key, range_end=None, start_revision=None, filters=None, read_timeout=None):
         """returns: response object"""
         params = build_range_request(key, range_end)
         if start_revision is not None:
@@ -357,11 +359,11 @@ class Etcd3Client(AbstractEtcdClientWithFailover):
         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)
+        kwargs.update(timeout=urllib3.Timeout(connect=kwargs['timeout'], read=read_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)
+    def watchprefix(self, key, start_revision=None, filters=None, read_timeout=None):
+        return self.watchrange(key, prefix_range_end(key), start_revision, filters, read_timeout)
 
 
 class KVCache(Thread):
@@ -450,7 +452,14 @@ class KVCache(Thread):
     def _do_watch(self, revision):
         with self._response_lock:
             self._response = None
-        response = self._client.watchprefix(self._dcs.cluster_prefix, revision)
+        # We do most of requests with timeouts. The only exception /watch requests to Etcd v3.
+        # In order to interrupt the /watch request we do socket.shutdown() from the main thread,
+        # which doesn't work on Windows. Therefore we want to use the last resort, `read_timeout`.
+        # Setting it to TTL will help to partially mitigate the problem.
+        # Setting it to lower value is not nice because for idling clusters it will increase
+        # the numbers of interrupts and reconnects.
+        read_timeout = self._dcs.ttl if os.name == 'nt' else None
+        response = self._client.watchprefix(self._dcs.cluster_prefix, revision, read_timeout=read_timeout)
         with self._response_lock:
             if self._response is None:
                 self._response = response
@@ -472,7 +481,9 @@ class KVCache(Thread):
         try:
             self._do_watch(result['header']['revision'])
         except Exception as e:
-            logger.error('watchprefix failed: %r', e)
+            # Following exceptions are expected on Windows because the /watch request  is done with `read_timeout`
+            if not (os.name == 'nt' and isinstance(e, (ReadTimeoutError, ProtocolError))):
+                logger.error('watchprefix failed: %r', e)
         finally:
             with self.condition:
                 self._is_ready = False

patroni/dcs/kubernetes.py

@@ -48,14 +48,29 @@ class K8sConfig(object):
 
     def __init__(self):
         self.pool_config = {'maxsize': 10, 'num_pools': 10}  # configuration for urllib3.PoolManager
+        self._token_expires_at = datetime.datetime.max
         self._make_headers()
 
+    def _set_token(self, token):
+        self._headers['authorization'] = 'Bearer ' + token
+
     def _make_headers(self, token=None, **kwargs):
         self._headers = urllib3.make_headers(user_agent=USER_AGENT, **kwargs)
         if token:
-            self._headers['authorization'] = 'Bearer ' + token
+            self._set_token(token)
 
-    def load_incluster_config(self, ca_certs=SERVICE_CERT_FILENAME):
+    def _read_token_file(self):
+        if not os.path.isfile(SERVICE_TOKEN_FILENAME):
+            raise self.ConfigException('Service token file does not exists.')
+        with open(SERVICE_TOKEN_FILENAME) as f:
+            token = f.read()
+            if not token:
+                raise self.ConfigException('Token file exists but empty.')
+            self._token_expires_at = datetime.datetime.now() + self._token_refresh_interval
+            return token
+
+    def load_incluster_config(self, ca_certs=SERVICE_CERT_FILENAME,
+                              token_refresh_interval=datetime.timedelta(minutes=1)):
         if SERVICE_HOST_ENV_NAME not in os.environ or SERVICE_PORT_ENV_NAME not in os.environ:
             raise self.ConfigException('Service host/port is not set.')
         if not os.environ[SERVICE_HOST_ENV_NAME] or not os.environ[SERVICE_PORT_ENV_NAME]:
@@ -67,14 +82,9 @@ class K8sConfig(object):
             if not f.read():
                 raise self.ConfigException('Cert file exists but empty.')
         self.pool_config['ca_certs'] = ca_certs
-
-        if not os.path.isfile(SERVICE_TOKEN_FILENAME):
-            raise self.ConfigException('Service token file does not exists.')
-        with open(SERVICE_TOKEN_FILENAME) as f:
-            token = f.read()
-            if not token:
-                raise self.ConfigException('Token file exists but empty.')
-            self._make_headers(token=token)
+        self._token_refresh_interval = token_refresh_interval
+        token = self._read_token_file()
+        self._make_headers(token=token)
         self._server = uri('https', (os.environ[SERVICE_HOST_ENV_NAME], os.environ[SERVICE_PORT_ENV_NAME]))
 
     @staticmethod
@@ -109,6 +119,11 @@ class K8sConfig(object):
 
     @property
     def headers(self):
+        if self._token_expires_at <= datetime.datetime.now():
+            try:
+                self._set_token(self._read_token_file())
+            except Exception as e:
+                logger.error('Failed to refresh service account token: %r', e)
         return self._headers.copy()
 
 
@@ -503,6 +518,8 @@ class ObjectCache(Thread):
         self._condition = condition
         self._name = name  # name of this pod
         self._is_ready = False
+        self._response = None  # needs to be accessible from the `kill_stream()` method
+        self._response_lock = Lock()  # protect the `self._response` from concurrent access
         self._object_cache = {}
         self._object_cache_lock = Lock()
         self._annotations_map = {self._dcs.leader_path: self._dcs._LEADER, self._dcs.config_path: self._dcs._CONFIG}
@@ -543,63 +560,99 @@ class ObjectCache(Thread):
         with self._object_cache_lock:
             return self._object_cache.get(name)
 
+    def _process_event(self, event):
+        ev_type = event['type']
+        obj = event['object']
+        name = obj['metadata']['name']
+
+        if ev_type in ('ADDED', 'MODIFIED'):
+            obj = K8sObject(obj)
+            success, old_value = self.set(name, obj)
+            if success:
+                new_value = (obj.metadata.annotations or {}).get(self._annotations_map.get(name))
+        elif ev_type == 'DELETED':
+            success, old_value = self.delete(name, obj['metadata']['resourceVersion'])
+            new_value = None
+        else:
+            return logger.warning('Unexpected event type: %s', ev_type)
+
+        if success and obj.get('kind') != 'Pod':
+            if old_value:
+                old_value = (old_value.metadata.annotations or {}).get(self._annotations_map.get(name))
+
+            value_changed = old_value != new_value and \
+                (name != self._dcs.config_path or old_value is not None and new_value is not None)
+
+            if value_changed:
+                logger.debug('%s changed from %s to %s', name, old_value, new_value)
+
+            # Do not wake up HA loop if we run as leader and received leader object update event
+            if value_changed or name == self._dcs.leader_path and self._name != new_value:
+                self._dcs.event.set()
+
+    @staticmethod
+    def _finish_response(response):
+        try:
+            response.close()
+        finally:
+            response.release_conn()
+
+    def _do_watch(self, resource_version):
+        with self._response_lock:
+            self._response = None
+        response = self._watch(resource_version)
+        with self._response_lock:
+            if self._response is None:
+                self._response = response
+
+        if not self._response:
+            return self._finish_response(response)
+
+        for event in iter_response_objects(response):
+            if event['object'].get('code') == 410:
+                break
+            self._process_event(event)
+
     def _build_cache(self):
         objects = self._list()
-        return_type = 'V1' + objects.kind[:-4]
         with self._object_cache_lock:
             self._object_cache = {item.metadata.name: item for item in objects.items}
         with self._condition:
             self._is_ready = True
             self._condition.notify()
 
-        response = self._watch(objects.metadata.resource_version)
         try:
-            for event in iter_response_objects(response):
-                obj = event['object']
-                if obj.get('code') == 410:
-                    break
-
-                ev_type = event['type']
-                name = obj['metadata']['name']
-
-                if ev_type in ('ADDED', 'MODIFIED'):
-                    obj = K8sObject(obj)
-                    success, old_value = self.set(name, obj)
-                    if success:
-                        new_value = (obj.metadata.annotations or {}).get(self._annotations_map.get(name))
-                elif ev_type == 'DELETED':
-                    success, old_value = self.delete(name, obj['metadata']['resourceVersion'])
-                    new_value = None
-                else:
-                    logger.warning('Unexpected event type: %s', ev_type)
-                    continue
-
-                if success and return_type != 'V1Pod':
-                    if old_value:
-                        old_value = (old_value.metadata.annotations or {}).get(self._annotations_map.get(name))
-
-                    value_changed = old_value != new_value and \
-                        (name != self._dcs.config_path or old_value is not None and new_value is not None)
-
-                    if value_changed:
-                        logger.debug('%s changed from %s to %s', name, old_value, new_value)
-
-                    # Do not wake up HA loop if we run as leader and received leader object update event
-                    if value_changed or name == self._dcs.leader_path and self._name != new_value:
-                        self._dcs.event.set()
+            self._do_watch(objects.metadata.resource_version)
         finally:
             with self._condition:
                 self._is_ready = False
-            response.close()
-            response.release_conn()
+            with self._response_lock:
+                response, self._response = self._response, None
+            if response:
+                self._finish_response(response)
+
+    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 run(self):
         while True:
             try:
                 self._build_cache()
             except Exception as e:
-                with self._condition:
-                    self._is_ready = False
                 logger.error('ObjectCache.run %r', e)
 
     def is_ready(self):
@@ -874,7 +927,14 @@ class Kubernetes(AbstractDCS):
     def patch_or_create(self, name, annotations, resource_version=None, patch=False, retry=True, ips=None):
         if retry is True:
             retry = self.retry
-        return self._patch_or_create(name, annotations, resource_version, patch, retry, ips)
+        try:
+            return self._patch_or_create(name, annotations, resource_version, patch, retry, ips)
+        except k8s_client.rest.ApiException as e:
+            if e.status == 409 and resource_version:  # Conflict in resource_version
+                # Terminate watchers, it could be a sign that K8s API is in a failed state
+                self._kinds.kill_stream()
+                self._pods.kill_stream()
+            raise e
 
     def patch_or_create_config(self, annotations, resource_version=None, patch=False, retry=True):
         # SCOPE-config endpoint requires corresponding service otherwise it might be "cleaned" by k8s master
@@ -890,7 +950,8 @@ class Kubernetes(AbstractDCS):
             if not self._api.create_namespaced_service(self._namespace, body):
                 return
         except Exception as e:
-            if not isinstance(e, k8s_client.rest.ApiException) or e.status != 409:  # Service already exists
+            # 409 - service already exists, 403 - creation forbidden
+            if not isinstance(e, k8s_client.rest.ApiException) or e.status not in (409, 403):
                 return logger.exception('create_config_service failed')
         self._should_create_config_service = False
 
@@ -961,7 +1022,7 @@ class Kubernetes(AbstractDCS):
                        'transitions': leader_observed_record.get('transitions') or '0'}
         if last_lsn:
             annotations[self._OPTIME] = str(last_lsn)
-        annotations['slots'] = json.dumps(slots) if slots else None
+            annotations['slots'] = json.dumps(slots) if slots else None
 
         resource_version = kind and kind.metadata.resource_version
         return self._update_leader_with_retry(annotations, resource_version, self.__ips)
@@ -1045,7 +1106,7 @@ class Kubernetes(AbstractDCS):
         if kind and (kind.metadata.annotations or {}).get(self._LEADER) == self._name:
             annotations = {self._LEADER: None}
             if last_lsn:
-                annotations[self._OPTIME] = last_lsn
+                annotations[self._OPTIME] = str(last_lsn)
             self.patch_or_create(self.leader_path, annotations, kind.metadata.resource_version, True, False, [])
             self.reset_cluster()
 

patroni/dcs/raft.py

@@ -39,9 +39,9 @@ setattr(TCPNode, 'ip', property(resolve_host))
 
 class SyncObjUtility(object):
 
-    def __init__(self, otherNodes, conf):
+    def __init__(self, otherNodes, conf, retry_timeout=10):
         self._nodes = otherNodes
-        self._utility = TcpUtility(conf.password)
+        self._utility = TcpUtility(conf.password, retry_timeout/max(1, len(otherNodes)))
 
     def executeCommand(self, command):
         try:
@@ -58,11 +58,11 @@ class SyncObjUtility(object):
 
 class DynMemberSyncObj(SyncObj):
 
-    def __init__(self, selfAddress, partnerAddrs, conf):
+    def __init__(self, selfAddress, partnerAddrs, conf, retry_timeout=10):
         self.__early_apply_local_log = selfAddress is not None
         self.applied_local_log = False
 
-        utility = SyncObjUtility(partnerAddrs, conf)
+        utility = SyncObjUtility(partnerAddrs, conf, retry_timeout)
         members = utility.getMembers()
         add_self = members and selfAddress not in members
 
@@ -97,7 +97,7 @@ class KVStoreTTL(DynMemberSyncObj):
         self.__on_set = on_set
         self.__on_delete = on_delete
         self.__limb = {}
-        self.__retry_timeout = None
+        self.set_retry_timeout(int(config.get('retry_timeout') or 10))
 
         self_addr = config.get('self_addr')
         partner_addrs = set(config.get('partner_addrs', []))
@@ -121,7 +121,7 @@ class KVStoreTTL(DynMemberSyncObj):
                            journalFile=(file_template + '.journal' if self_addr else None),
                            onReady=on_ready, dynamicMembershipChange=True)
 
-        super(KVStoreTTL, self).__init__(self_addr, partner_addrs, conf)
+        super(KVStoreTTL, self).__init__(self_addr, partner_addrs, conf, self.__retry_timeout)
         self.__data = {}
 
     @staticmethod
@@ -275,7 +275,6 @@ class Raft(AbstractDCS):
                 break
             else:
                 logger.info('waiting on raft')
-        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')

patroni/dcs/zookeeper.py

@@ -1,6 +1,7 @@
 import json
 import logging
 import select
+import six
 import time
 
 from kazoo.client import KazooClient, KazooState, KazooRetry
@@ -50,11 +51,21 @@ class PatroniSequentialThreadingHandler(SequentialThreadingHandler):
         return super(PatroniSequentialThreadingHandler, self).create_connection(*args, **kwargs)
 
     def select(self, *args, **kwargs):
-        """Python3 raises `ValueError` if socket is closed, because fd == -1"""
+        """
+        Python 3.XY may raise following exceptions if select/poll are called with an invalid socket:
+        - `ValueError`: because fd == -1
+        - `TypeError`: Invalid file descriptor: -1 (starting from kazoo 2.9)
+        Python 2.7 may raise the `IOError` instead of `socket.error` (starting from kazoo 2.9)
+
+        When it is appropriate we map these exceptions to `socket.error`.
+        """
+
         try:
             return super(PatroniSequentialThreadingHandler, self).select(*args, **kwargs)
-        except ValueError as e:
-            raise select.error(9, str(e))
+        except IOError as e:
+            raise (select.error(e.errno, e.strerror) if six.PY2 else e)
+        except (TypeError, ValueError) as e:
+            raise (e if six.PY2 and isinstance(e, TypeError) else select.error(9, str(e)))
 
 
 class PatroniKazooClient(KazooClient):

patroni/ha.py

@@ -192,6 +192,10 @@ class Ha(object):
                 'version': self.patroni.version
             }
 
+            proxy_url = self.state_handler.proxy_url
+            if proxy_url:
+                data['proxy_url'] = proxy_url
+
             if self.is_leader() and not self._rewind.checkpoint_after_promote():
                 data['checkpoint_after_promote'] = False
             tags = self.get_effective_tags()
@@ -273,7 +277,9 @@ class Ha(object):
         else:
             create_replica_methods = self.get_standby_cluster_config().get('create_replica_methods', []) \
                                      if self.is_standby_cluster() else None
-            if self.state_handler.can_create_replica_without_replication_connection(create_replica_methods):
+            can_bootstrap = self.state_handler.can_create_replica_without_replication_connection(create_replica_methods)
+            concurrent_bootstrap = self.cluster.initialize == ""
+            if can_bootstrap and not concurrent_bootstrap:
                 msg = 'bootstrap (without leader)'
                 return self._async_executor.try_run_async(msg, self.clone) or 'trying to ' + msg
             return 'waiting for {0}leader to bootstrap'.format('standby_' if self.is_standby_cluster() else '')
@@ -293,17 +299,30 @@ class Ha(object):
 
         return result
 
+    def _handle_crash_recovery(self):
+        if not self._crash_recovery_executed and (self.cluster.is_unlocked() or self._rewind.can_rewind):
+            self._crash_recovery_executed = True
+            self._crash_recovery_started = time.time()
+            msg = 'doing crash recovery in a single user mode'
+            return self._async_executor.try_run_async(msg, self._rewind.ensure_clean_shutdown) or msg
+
     def _handle_rewind_or_reinitialize(self):
         leader = self.get_remote_master() if self.is_standby_cluster() else self.cluster.leader
         if not self._rewind.rewind_or_reinitialize_needed_and_possible(leader):
             return None
 
         if self._rewind.can_rewind:
+            # rewind is required, but postgres wasn't shut down cleanly.
+            if not self.state_handler.is_running() and \
+                    self.state_handler.controldata().get('Database cluster state') == 'in archive recovery':
+                msg = self._handle_crash_recovery()
+                if msg:
+                    return msg
+
             msg = 'running pg_rewind from ' + leader.name
             return self._async_executor.try_run_async(msg, self._rewind.execute, args=(leader,)) or msg
 
-        # remove_data_directory_on_diverged_timelines is set
-        if not self.is_standby_cluster():
+        if self._rewind.should_remove_data_directory_on_diverged_timelines and not self.is_standby_cluster():
             msg = 'reinitializing due to diverged timelines'
             return self._async_executor.try_run_async(msg, self._do_reinitialize, args=(self.cluster,)) or msg
 
@@ -325,13 +344,10 @@ class Ha(object):
 
         data = self.state_handler.controldata()
         logger.info('pg_controldata:\n%s\n', '\n'.join('  {0}: {1}'.format(k, v) for k, v in data.items()))
-        if data.get('Database cluster state') in ('in production', 'shutting down', 'in crash recovery') \
-                and not self._crash_recovery_executed and \
-                (self.cluster.is_unlocked() or self._rewind.can_rewind):
-            self._crash_recovery_executed = True
-            self._crash_recovery_started = time.time()
-            msg = 'doing crash recovery in a single user mode'
-            return self._async_executor.try_run_async(msg, self._rewind.ensure_clean_shutdown) or msg
+        if data.get('Database cluster state') in ('in production', 'shutting down', 'in crash recovery'):
+            msg = self._handle_crash_recovery()
+            if msg:
+                return msg
 
         self.load_cluster_from_dcs()
 
@@ -413,9 +429,10 @@ class Ha(object):
                 self.state_handler.get_history(self._leader_timeline + 1):
             self._rewind.trigger_check_diverged_lsn()
 
-        msg = self._handle_rewind_or_reinitialize()
-        if msg:
-            return msg
+        if not self.state_handler.is_starting():
+            msg = self._handle_rewind_or_reinitialize()
+            if msg:
+                return msg
 
         if not self.is_paused():
             self.state_handler.handle_parameter_change()
@@ -726,6 +743,11 @@ class Ha(object):
                     return None
                 return False
 
+            # in synchronous mode when our name is not in the /sync key
+            # we shouldn't take any action even if the candidate is unhealthy
+            if self.is_synchronous_mode() and not self.cluster.sync.matches(self.state_handler.name):
+                return False
+
             # find specific node and check that it is healthy
             member = self.cluster.get_member(failover.candidate, fallback_to_leader=False)
             if member:
@@ -786,7 +808,7 @@ class Ha(object):
         if self.cluster.failover:
             # When doing a switchover in synchronous mode only synchronous nodes and former leader are allowed to race
             if self.is_synchronous_mode() and self.cluster.failover.leader and \
-                    self.cluster.failover.candidate and not self.cluster.sync.matches(self.state_handler.name):
+                    not self.cluster.sync.matches(self.state_handler.name):
                 return False
             return self.manual_failover_process_no_leader()
 
@@ -1396,7 +1418,14 @@ class Ha(object):
                     return 'started as a secondary'
 
             # is data directory empty?
-            if self.state_handler.data_directory_empty():
+            try:
+                data_directory_is_empty = self.state_handler.data_directory_empty()
+                data_directory_is_accessible = True
+            except OSError as e:
+                data_directory_is_accessible = False
+                data_directory_error = e
+
+            if not data_directory_is_accessible or data_directory_is_empty:
                 self.state_handler.set_role('uninitialized')
                 self.state_handler.stop('immediate', stop_timeout=self.patroni.config['retry_timeout'])
                 # In case datadir went away while we were master.
@@ -1405,8 +1434,11 @@ class Ha(object):
                 # is this instance the leader?
                 if self.has_lock():
                     self.release_leader_key_voluntarily()
-                    return 'released leader key voluntarily as data dir empty and currently leader'
+                    return 'released leader key voluntarily as data dir {0} and currently leader'.format(
+                                'empty' if data_directory_is_accessible else 'not accessible')
 
+                if not data_directory_is_accessible:
+                    return 'data directory is not accessible: {0}'.format(data_directory_error)
                 if self.is_paused():
                     return 'running with empty data directory'
                 return self.bootstrap()  # new node
@@ -1472,7 +1504,9 @@ class Ha(object):
                 # asynchronous processes are running (should be always the case for the master)
                 if not self._async_executor.busy and not self.state_handler.is_starting():
                     create_slots = self.state_handler.slots_handler.sync_replication_slots(self.cluster,
-                                                                                           self.patroni.nofailover)
+                                                                                           self.patroni.nofailover,
+                                                                                           self.patroni.replicatefrom,
+                                                                                           self.is_paused())
                     if not self.state_handler.cb_called:
                         if not self.state_handler.is_leader():
                             self._rewind.trigger_check_diverged_lsn()
@@ -1480,7 +1514,7 @@ class Ha(object):
                     if create_slots and self.cluster.leader:
                         err = self._async_executor.try_run_async('copy_logical_slots',
                                                                  self.state_handler.slots_handler.copy_logical_slots,
-                                                                 args=(self.cluster.leader, create_slots))
+                                                                 args=(self.cluster, create_slots))
                         if not err:
                             ret = 'Copying logical slots {0} from the primary'.format(create_slots)
             return ret

patroni/postgresql/__init__.py

@@ -387,7 +387,7 @@ class Postgresql(object):
         self._query('SELECT pg_catalog.pg_{0}_replay_resume()'.format(self.wal_name))
 
     def handle_parameter_change(self):
-        if self.major_version >= 140000 and self.replay_paused():
+        if self.major_version >= 140000 and not self.is_starting() and self.replay_paused():
             logger.info('Resuming paused WAL replay for PostgreSQL 14+')
             self.resume_wal_replay()
 
@@ -428,11 +428,11 @@ class Postgresql(object):
                     # If the cluster is shutdown with archive_mode=on, WAL is switched before writing the checkpoint.
                     # In this case we want to take the LSN of previous record (switch) as the last known WAL location.
                     if parse_lsn(lsn) == prev and desc.strip() in ('xlog switch', 'SWITCH'):
-                        return str(prev)
+                        return prev
             except Exception as e:
                 logger.error('Exception when parsing WAL pg_%sdump output: %r', self.wal_name, e)
             if isinstance(checkpoint_lsn, six.integer_types):
-                return str(checkpoint_lsn)
+                return checkpoint_lsn
 
     def is_running(self):
         """Returns PostmasterProcess if one is running on the data directory or None. If most recently seen process
@@ -659,7 +659,7 @@ class Postgresql(object):
         if on_safepoint:
             # Wait for our connection to terminate so we can be sure that no new connections are being initiated
             self._wait_for_connection_close(postmaster)
-            postmaster.wait_for_user_backends_to_close()
+            postmaster.wait_for_user_backends_to_close(stop_timeout)
             on_safepoint()
 
         if on_shutdown and mode in ('fast', 'smart'):
@@ -668,7 +668,7 @@ class Postgresql(object):
             while postmaster.is_running():
                 data = self.controldata()
                 if data.get('Database cluster state', '') == 'shut down':
-                    on_shutdown(int(self.latest_checkpoint_location()))
+                    on_shutdown(self.latest_checkpoint_location())
                     break
                 elif data.get('Database cluster state', '').startswith('shut down'):  # shut down in recovery
                     break
@@ -798,7 +798,8 @@ class Postgresql(object):
         return True
 
     def get_guc_value(self, name):
-        cmd = [self.pgcommand('postgres'), '-D', self._data_dir, '-C', name]
+        cmd = [self.pgcommand('postgres'), '-D', self._data_dir, '-C', name,
+               '--config-file={}'.format(self.config.postgresql_conf)]
         try:
             data = subprocess.check_output(cmd)
             if data:

patroni/postgresql/config.py

@@ -577,6 +577,9 @@ class ConfigHandler(object):
         return self._RECOVERY_PARAMETERS - skip_params
 
     def _read_recovery_params(self):
+        if self._postgresql.is_starting():
+            return None, False
+
         pg_conf_mtime = mtime(self._postgresql_conf)
         auto_conf_mtime = mtime(self._auto_conf)
         passfile_mtime = mtime(self._passfile) if self._passfile else False
@@ -650,16 +653,17 @@ class ConfigHandler(object):
         elif not primary_conninfo:
             return False
 
-        wal_receiver_primary_conninfo = self._postgresql.primary_conninfo()
-        if wal_receiver_primary_conninfo:
-            wal_receiver_primary_conninfo = parse_dsn(wal_receiver_primary_conninfo)
-            # when wal receiver is alive use primary_conninfo from pg_stat_wal_receiver for comparison
+        if not self._postgresql.is_starting():
+            wal_receiver_primary_conninfo = self._postgresql.primary_conninfo()
             if wal_receiver_primary_conninfo:
-                primary_conninfo = wal_receiver_primary_conninfo
-                # There could be no password in the primary_conninfo or it is masked.
-                # Just copy the "desired" value in order to make comparison succeed.
-                if 'password' in wanted_primary_conninfo:
-                    primary_conninfo['password'] = wanted_primary_conninfo['password']
+                wal_receiver_primary_conninfo = parse_dsn(wal_receiver_primary_conninfo)
+                # when wal receiver is alive use primary_conninfo from pg_stat_wal_receiver for comparison
+                if wal_receiver_primary_conninfo:
+                    primary_conninfo = wal_receiver_primary_conninfo
+                    # There could be no password in the primary_conninfo or it is masked.
+                    # Just copy the "desired" value in order to make comparison succeed.
+                    if 'password' in wanted_primary_conninfo:
+                        primary_conninfo['password'] = wanted_primary_conninfo['password']
 
         if 'passfile' in primary_conninfo and 'password' not in primary_conninfo \
                 and 'password' in wanted_primary_conninfo:
@@ -668,7 +672,7 @@ class ConfigHandler(object):
             else:
                 return False
 
-        return all(primary_conninfo.get(p) == str(v) for p, v in wanted_primary_conninfo.items() if v is not None)
+        return all(str(primary_conninfo.get(p)) == str(v) for p, v in wanted_primary_conninfo.items() if v is not None)
 
     def check_recovery_conf(self, member):
         """Returns a tuple. The first boolean element indicates that recovery params don't match
@@ -704,16 +708,19 @@ class ConfigHandler(object):
             else:  # empty string, primary_conninfo is not in the config
                 primary_conninfo[0] = {}
 
-        # when wal receiver is alive take primary_slot_name from pg_stat_wal_receiver
-        wal_receiver_primary_slot_name = self._postgresql.primary_slot_name()
-        if not wal_receiver_primary_slot_name and self._postgresql.primary_conninfo():
-            wal_receiver_primary_slot_name = ''
-        if wal_receiver_primary_slot_name is not None:
-            self._current_recovery_params['primary_slot_name'][0] = wal_receiver_primary_slot_name
+        if not self._postgresql.is_starting():
+            # when wal receiver is alive take primary_slot_name from pg_stat_wal_receiver
+            wal_receiver_primary_slot_name = self._postgresql.primary_slot_name()
+            if not wal_receiver_primary_slot_name and self._postgresql.primary_conninfo():
+                wal_receiver_primary_slot_name = ''
+            if wal_receiver_primary_slot_name is not None:
+                self._current_recovery_params['primary_slot_name'][0] = wal_receiver_primary_slot_name
 
         # Increment the 'reload' to enforce write of postgresql.conf when joining the running postgres
         required = {'restart': 0,
-                    'reload': int(not self._postgresql.cb_called and self._postgresql.major_version >= 120000)}
+                    'reload': int(self._postgresql.major_version >= 120000
+                                  and not self._postgresql.cb_called
+                                  and not self._postgresql.is_starting())}
 
         def record_missmatch(mtype):
             required['restart' if mtype else 'reload'] += 1
@@ -780,6 +787,15 @@ class ConfigHandler(object):
             else:
                 self._remove_file_if_exists(self._standby_signal)
                 open(self._recovery_signal, 'w').close()
+
+            def restart_required(name):
+                if self._postgresql.major_version >= 140000:
+                    return False
+                return name == 'restore_command' or (self._postgresql.major_version < 130000
+                                                     and name in ('primary_conninfo', 'primary_slot_name'))
+
+            self._current_recovery_params = {n: [v, restart_required(n), self._postgresql_conf]
+                                             for n, v in recovery_params.items()}
         else:
             with ConfigWriter(self._recovery_conf) as f:
                 os.chmod(self._recovery_conf, stat.S_IWRITE | stat.S_IREAD)
@@ -789,6 +805,7 @@ class ConfigHandler(object):
         for name in (self._recovery_conf, self._standby_signal, self._recovery_signal):
             self._remove_file_if_exists(name)
         self._recovery_params = {}
+        self._current_recovery_params = None
 
     def _sanitize_auto_conf(self):
         overwrite = False
@@ -1000,6 +1017,9 @@ class ConfigHandler(object):
         if not local_connection_address_changed:
             self.resolve_connection_addresses()
 
+        proxy_addr = config.get('proxy_address')
+        self._postgresql.proxy_url = uri('postgres', proxy_addr, self._postgresql.database) if proxy_addr else None
+
         if conf_changed:
             self.write_postgresql_conf()
 

patroni/postgresql/postmaster.py

@@ -171,8 +171,8 @@ class PostmasterProcess(psutil.Process):
         else:
             return not self.is_running()
 
-    def wait_for_user_backends_to_close(self):
-        # These regexps are cross checked against versions PostgreSQL 9.1 .. 11
+    def wait_for_user_backends_to_close(self, stop_timeout):
+        # These regexps are cross checked against versions PostgreSQL 9.1 .. 15
         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|"
@@ -184,19 +184,23 @@ class PostmasterProcess(psutil.Process):
             return logger.debug('Failed to get list of postmaster children')
 
         user_backends = []
-        user_backends_cmdlines = []
+        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])
+                    user_backends_cmdlines[child.pid] = 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")
+            logger.debug('Waiting for user backends %s to close', ', '.join(user_backends_cmdlines.values()))
+            gone, live = psutil.wait_procs(user_backends, stop_timeout)
+            if stop_timeout and live:
+                live = [user_backends_cmdlines[b.pid] for b in live]
+                logger.warning('Backends still alive after %s: %s', stop_timeout, ', '.join(live))
+            else:
+                logger.debug("Backends closed")
 
     @staticmethod
     def start(pgcommand, data_dir, conf, options):

patroni/postgresql/rewind.py

@@ -1,6 +1,8 @@
 import logging
 import os
+import re
 import shlex
+import shutil
 import six
 import subprocess
 
@@ -28,13 +30,17 @@ class Rewind(object):
     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 enabled(self):
+        return self._postgresql.config.get('use_pg_rewind')
+
     @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'):
+        if not self.enabled:
             return False
 
         cmd = [self._postgresql.pgcommand('pg_rewind'), '--help']
@@ -46,9 +52,13 @@ class Rewind(object):
             return False
         return self.configuration_allows_rewind(self._postgresql.controldata())
 
+    @property
+    def should_remove_data_directory_on_diverged_timelines(self):
+        return self._postgresql.config.get('remove_data_directory_on_diverged_timelines')
+
     @property
     def can_rewind_or_reinitialize_allowed(self):
-        return self._postgresql.config.get('remove_data_directory_on_diverged_timelines') or self.can_rewind
+        return self.should_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:
@@ -112,7 +122,7 @@ class Rewind(object):
         in_recovery = timeline = lsn = None
         data = self._postgresql.controldata()
         try:
-            if data.get('Database cluster state') == 'shut down in recovery':
+            if data.get('Database cluster state') in ('shut down in recovery', 'in archive recovery'):
                 in_recovery = True
                 lsn = data.get('Minimum recovery ending location')
                 timeline = int(data.get("Min recovery ending loc's timeline"))
@@ -182,8 +192,14 @@ class Rewind(object):
         if isinstance(leader, Leader) and leader.member.data.get('role') != 'master':
             return
 
-        if not self.check_leader_is_not_in_recovery(
-                self._conn_kwargs(leader, self._postgresql.config.replication)):
+        # We want to use replication credentials when connecting to the "postgres" database in case if
+        # `use_pg_rewind` isn't enabled and only `remove_data_directory_on_diverged_timelines` is set
+        # for Postgresql older than v11 (where Patroni can't use a dedicated user for rewind).
+        # In all other cases we will use rewind or superuser credentials.
+        check_credentials = self._postgresql.config.replication if not self.enabled and\
+            self.should_remove_data_directory_on_diverged_timelines and\
+            self._postgresql.major_version < 110000 else self._postgresql.config.rewind_credentials
+        if not self.check_leader_is_not_in_recovery(self._conn_kwargs(leader, check_credentials)):
             return
 
         history = need_rewind = None
@@ -263,28 +279,37 @@ class Rewind(object):
     def checkpoint_after_promote(self):
         return self._state == REWIND_STATUS.CHECKPOINT
 
-    def _fetch_missing_wal(self, restore_command, wal_filename):
+    def _buid_archiver_command(self, command, wal_filename):
+        """Replace placeholders in the given archiver command's template.
+        Applicable for archive_command and restore_command.
+        Can also be used for archive_cleanup_command and recovery_end_command,
+        however %r value is always set to 000000010000000000000001."""
         cmd = ''
-        length = len(restore_command)
+        length = len(command)
         i = 0
         while i < length:
-            if restore_command[i] == '%' and i + 1 < length:
+            if command[i] == '%' and i + 1 < length:
                 i += 1
-                if restore_command[i] == 'p':
+                if command[i] == 'p':
                     cmd += os.path.join(self._postgresql.wal_dir, wal_filename)
-                elif restore_command[i] == 'f':
+                elif command[i] == 'f':
                     cmd += wal_filename
-                elif restore_command[i] == 'r':
+                elif command[i] == 'r':
                     cmd += '000000010000000000000001'
-                elif restore_command[i] == '%':
+                elif command[i] == '%':
                     cmd += '%'
                 else:
                     cmd += '%'
                     i -= 1
             else:
-                cmd += restore_command[i]
+                cmd += command[i]
             i += 1
 
+        return cmd
+
+    def _fetch_missing_wal(self, restore_command, wal_filename):
+        cmd = self._buid_archiver_command(restore_command, wal_filename)
+
         logger.info('Trying to fetch the missing wal: %s', cmd)
         return self._postgresql.cancellable.call(shlex.split(cmd)) == 0
 
@@ -301,6 +326,42 @@ class Rewind(object):
                     if waldir.endswith('/pg_' + self._postgresql.wal_name) and len(wal_filename) == 24:
                         return wal_filename
 
+    def _archive_ready_wals(self):
+        """Try to archive WALs that have .ready files just in case
+        archive_mode was not set to 'always' before promote, while
+        after it the WALs were recycled on the promoted replica.
+        With this we prevent the entire loss of such WALs and the
+        consequent old leader's start failure."""
+        archive_mode = self._postgresql.get_guc_value('archive_mode')
+        archive_cmd = self._postgresql.get_guc_value('archive_command')
+        if archive_mode not in ('on', 'always') or not archive_cmd:
+            return
+
+        walseg_regex = re.compile(r'^[0-9A-F]{24}(\.partial){0,1}\.ready$')
+        status_dir = os.path.join(self._postgresql.wal_dir, 'archive_status')
+        try:
+            wals_to_archive = [f[:-6] for f in os.listdir(status_dir) if walseg_regex.match(f)]
+        except OSError as e:
+            return logger.error('Unable to list %s: %r', status_dir, e)
+
+        # skip fsync, as postgres --single or pg_rewind will anyway run it
+        for wal in sorted(wals_to_archive):
+            old_name = os.path.join(status_dir, wal + '.ready')
+            # wal file might have alredy been archived
+            if os.path.isfile(old_name) and os.path.isfile(os.path.join(self._postgresql.wal_dir, wal)):
+                cmd = self._buid_archiver_command(archive_cmd, wal)
+                # it is the author of archive_command, who is responsible
+                # for not overriding the WALs already present in archive
+                logger.info('Trying to archive %s: %s', wal, cmd)
+                if self._postgresql.cancellable.call(shlex.split(cmd)) == 0:
+                    new_name = os.path.join(status_dir, wal + '.done')
+                    try:
+                        shutil.move(old_name, new_name)
+                    except Exception as e:
+                        logger.error('Unable to rename %s to %s: %r', old_name, new_name, e)
+                else:
+                    logger.info('Failed to archive WAL segment %s', wal)
+
     def pg_rewind(self, r):
         # prepare pg_rewind connection
         env = self._postgresql.config.write_pgpass(r)
@@ -311,9 +372,18 @@ class Rewind(object):
         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')
 
+        # Until v15 pg_rewind expected postgresql.conf to be inside $PGDATA, which is not the case on e.g. Debian
+        pg_rewind_can_restore = restore_command and (self._postgresql.major_version >= 150000 or
+                                                     (self._postgresql.major_version >= 130000 and
+                                                      self._postgresql.config._config_dir == self._postgresql.data_dir))
+
         cmd = [self._postgresql.pgcommand('pg_rewind')]
-        if self._postgresql.major_version >= 130000 and restore_command:
+        if pg_rewind_can_restore:
             cmd.append('--restore-target-wal')
+            if self._postgresql.major_version >= 150000 and\
+                    self._postgresql.config._config_dir != self._postgresql.data_dir:
+                cmd.append('--config-file={0}'.format(self._postgresql.config.postgresql_conf))
+
         cmd.extend(['-D', self._postgresql.data_dir, '--source-server', dsn])
 
         while True:
@@ -329,7 +399,7 @@ class Rewind(object):
             if ret == 0:
                 return True
 
-            if not restore_command or self._postgresql.major_version >= 130000:
+            if not restore_command or pg_rewind_can_restore:
                 return False
 
             missing_wal = self._find_missing_wal(results['stderr']) or self._find_missing_wal(results['stdout'])
@@ -344,6 +414,8 @@ class Rewind(object):
         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')
 
+        self._archive_ready_wals()
+
         # prepare pg_rewind connection
         r = self._conn_kwargs(leader, self._postgresql.config.rewind_credentials)
 
@@ -368,19 +440,22 @@ class Rewind(object):
 
         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
+            if not self.check_leader_is_not_in_recovery(r):
+                logger.warning('Failed to rewind because master %s become unreachable', leader.name)
+                if not self.can_rewind:  # It is possible that the previous attempt damaged pg_control file!
+                    self._state = REWIND_STATUS.FAILED
             else:
+                logger.error('Failed to rewind from healty master: %s', leader.name)
                 self._state = REWIND_STATUS.FAILED
+
+            if self.failed:
+                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
         return False
 
     def reset_state(self):
@@ -439,6 +514,7 @@ class Rewind(object):
             logger.exception('Unable to list %s', status_dir)
 
     def ensure_clean_shutdown(self):
+        self._archive_ready_wals()
         self.cleanup_archive_status()
 
         # Start in a single user mode and stop to produce a clean shutdown

patroni/postgresql/slots.py

@@ -5,6 +5,7 @@ import shutil
 
 from collections import defaultdict
 from contextlib import contextmanager
+from threading import Condition, Thread
 
 from .connection import get_connection_cursor
 from .misc import format_lsn
@@ -31,12 +32,96 @@ def fsync_dir(path):
             os.close(fd)
 
 
+class SlotsAdvanceThread(Thread):
+
+    def __init__(self, slots_handler):
+        super(SlotsAdvanceThread, self).__init__()
+        self.daemon = True
+        self._slots_handler = slots_handler
+
+        # _copy_slots and _failed are used to asynchronously give some feedback to the main thread
+        self._copy_slots = []
+        self._failed = False
+
+        self._scheduled = defaultdict(dict)  # {'dbname1': {'slot1': 100, 'slot2': 100}, 'dbname2': {'slot3': 100}}
+        self._condition = Condition()  # protect self._scheduled from concurrent access and to wakeup the run() method
+
+        self.start()
+
+    def sync_slot(self, cur, database, slot, lsn):
+        failed = copy = False
+        try:
+            cur.execute("SELECT pg_catalog.pg_replication_slot_advance(%s, %s)", (slot, format_lsn(lsn)))
+        except Exception as e:
+            logger.error("Failed to advance logical replication slot '%s': %r", slot, e)
+            failed = True
+            copy = isinstance(e, OperationalError) and e.diag.sqlstate == '58P01'  # WAL file is gone
+        with self._condition:
+            if self._scheduled and failed:
+                if copy and slot not in self._copy_slots:
+                    self._copy_slots.append(slot)
+                self._failed = True
+
+            new_lsn = self._scheduled.get(database, {}).get(slot, 0)
+            # remove slot from the self._scheduled structure only if it wasn't changed
+            if new_lsn == lsn and database in self._scheduled:
+                self._scheduled[database].pop(slot)
+                if not self._scheduled[database]:
+                    self._scheduled.pop(database)
+
+    def sync_slots_in_database(self, database, slots):
+        with self._slots_handler.get_local_connection_cursor(dbname=database, options='-c statement_timeout=0') as cur:
+            for slot in slots:
+                with self._condition:
+                    lsn = self._scheduled.get(database, {}).get(slot, 0)
+                if lsn:
+                    self.sync_slot(cur, database, slot, lsn)
+
+    def sync_slots(self):
+        with self._condition:
+            databases = list(self._scheduled.keys())
+        for database in databases:
+            with self._condition:
+                slots = list(self._scheduled.get(database, {}).keys())
+            if slots:
+                try:
+                    self.sync_slots_in_database(database, slots)
+                except Exception as e:
+                    logger.error('Failed to advance replication slots in database %s: %r', database, e)
+
+    def run(self):
+        while True:
+            with self._condition:
+                if not self._scheduled:
+                    self._condition.wait()
+
+            self.sync_slots()
+
+    def schedule(self, advance_slots):
+        with self._condition:
+            for database, values in advance_slots.items():
+                self._scheduled[database].update(values)
+            ret = (self._failed, self._copy_slots)
+            self._copy_slots = []
+            self._failed = False
+            self._condition.notify()
+
+        return ret
+
+    def on_promote(self):
+        with self._condition:
+            self._scheduled.clear()
+            self._failed = False
+            self._copy_slots = []
+
+
 class SlotsHandler(object):
 
     def __init__(self, postgresql):
         self._postgresql = postgresql
+        self._advance = None
         self._replication_slots = {}  # already existing replication slots
-        self._unready_logical_slots = set()
+        self._unready_logical_slots = {}
         self.schedule()
 
     def _query(self, sql, *params):
@@ -95,7 +180,7 @@ class SlotsHandler(object):
             self._replication_slots = replication_slots
             self._schedule_load_slots = False
             if self._force_readiness_check:
-                self._unready_logical_slots = set(n for n, v in replication_slots.items() if v['type'] == 'logical')
+                self._unready_logical_slots = {n: None for n, v in replication_slots.items() if v['type'] == 'logical'}
                 self._force_readiness_check = False
 
     def ignore_replication_slot(self, cluster, name):
@@ -112,10 +197,10 @@ class SlotsHandler(object):
         # In normal situation rowcount should be 1, otherwise either slot doesn't exists or it is still active
         return cursor.rowcount == 1
 
-    def _drop_incorrect_slots(self, cluster, slots):
+    def _drop_incorrect_slots(self, cluster, slots, paused):
         # 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):
+            if not paused and 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
 
@@ -143,7 +228,7 @@ class SlotsHandler(object):
                 self._schedule_load_slots = True
 
     @contextmanager
-    def _get_local_connection_cursor(self, **kwargs):
+    def get_local_connection_cursor(self, **kwargs):
         conn_kwargs = self._postgresql.config.local_connect_kwargs
         conn_kwargs.update(kwargs)
         with get_connection_cursor(**conn_kwargs) as cur:
@@ -162,7 +247,7 @@ class SlotsHandler(object):
 
         # Create new logical slots
         for database, values in logical_slots.items():
-            with self._get_local_connection_cursor(dbname=database) as cur:
+            with self.get_local_connection_cursor(dbname=database) as cur:
                 for name, value in values.items():
                     try:
                         cur.execute("SELECT pg_catalog.pg_create_logical_replication_slot(%s, %s)" +
@@ -175,6 +260,11 @@ class SlotsHandler(object):
                         slots.pop(name)
                     self._schedule_load_slots = True
 
+    def schedule_advance_slots(self, slots):
+        if not self._advance:
+            self._advance = SlotsAdvanceThread(self)
+        return self._advance.schedule(slots)
+
     def _ensure_logical_slots_replica(self, cluster, slots):
         advance_slots = defaultdict(dict)  # Group logical slots to be advanced by database name
         create_slots = []  # And collect logical slots to be created on the replica
@@ -186,27 +276,18 @@ class SlotsHandler(object):
                     if name in cluster.slots:
                         try:  # Skip slots that doesn't need to be advanced
                             if value['confirmed_flush_lsn'] < int(cluster.slots[name]):
-                                advance_slots[value['database']][name] = value
+                                advance_slots[value['database']][name] = int(cluster.slots[name])
                         except Exception as e:
                             logger.error('Failed to parse "%s": %r', cluster.slots[name], e)
                 elif name in cluster.slots:  # We want to copy only slots with feedback in a DCS
                     create_slots.append(name)
 
-        # Advance logical slots
-        for database, values in advance_slots.items():
-            with self._get_local_connection_cursor(dbname=database, options='-c statement_timeout=0') as cur:
-                for name, value in values.items():
-                    try:
-                        cur.execute("SELECT pg_catalog.pg_replication_slot_advance(%s, %s)",
-                                    (name, format_lsn(int(cluster.slots[name]))))
-                    except Exception as e:
-                        logger.error("Failed to advance logical replication slot '%s': %r", name, e)
-                        if isinstance(e, OperationalError) and e.diag.sqlstate == '58P01':  # WAL file is gone
-                            create_slots.append(name)
-                        self._schedule_load_slots = True
-        return create_slots
+        error, copy_slots = self.schedule_advance_slots(advance_slots)
+        if error:
+            self._schedule_load_slots = True
+        return create_slots + copy_slots
 
-    def sync_replication_slots(self, cluster, nofailover, replicatefrom=None):
+    def sync_replication_slots(self, cluster, nofailover, replicatefrom=None, paused=False):
         ret = None
         if self._postgresql.major_version >= 90400 and cluster.config:
             try:
@@ -215,7 +296,7 @@ class SlotsHandler(object):
                 slots = cluster.get_replication_slots(self._postgresql.name, self._postgresql.role,
                                                       nofailover, self._postgresql.major_version, True)
 
-                self._drop_incorrect_slots(cluster, slots)
+                self._drop_incorrect_slots(cluster, slots, paused)
 
                 self._ensure_physical_slots(slots)
 
@@ -245,35 +326,69 @@ class SlotsHandler(object):
             slot_name = cluster.get_my_slot_name_on_primary(self._postgresql.name, replicatefrom)
             try:
                 with self._get_leader_connection_cursor(cluster.leader) as cur:
-                    cur.execute("SELECT catalog_xmin FROM pg_catalog.pg_get_replication_slots()"
-                                " WHERE NOT pg_catalog.pg_is_in_recovery() AND slot_name = %s", (slot_name,))
-                    if cur.rowcount < 1:
+                    cur.execute("SELECT slot_name, catalog_xmin FROM pg_catalog.pg_get_replication_slots()"
+                                " WHERE NOT pg_catalog.pg_is_in_recovery() AND slot_name = ANY(%s)",
+                                ([n for n, v in self._unready_logical_slots.items() if v is None] + [slot_name],))
+                    slots = {row[0]: row[1] for row in cur}
+                    if slot_name not in slots:
                         return logger.warning('Physical slot %s does not exist on the primary', slot_name)
-                    catalog_xmin = cur.fetchone()[0]
+                    catalog_xmin = slots.pop(slot_name)
             except Exception as e:
                 return logger.error("Failed to check %s physical slot on the primary: %r", slot_name, e)
+            # Remember catalog_xmin of logical slots on the primary when catalog_xmin of
+            # the physical slot became valid. Logical slots on replica will be safe to use after
+            # promote when catalog_xmin of the physical slot overtakes these values.
+            if catalog_xmin:
+                for name, value in slots.items():
+                    self._unready_logical_slots[name] = value
+            else:  # Replica isn't streaming or the hot_standby_feedback isn't enabled
+                try:
+                    cur = self._query("SELECT pg_catalog.current_setting('hot_standby_feedback')::boolean")
+                    if not cur.fetchone()[0]:
+                        logger.error('Logical slot failover requires "hot_standby_feedback".'
+                                     ' Please check postgresql.auto.conf')
+                except Exception as e:
+                    logger.error('Failed to check the hot_standby_feedback setting: %r', e)
+                return  # since `catalog_xmin` isn't valid further checks don't make any sense
+
         for name in list(self._unready_logical_slots):
             value = self._replication_slots.get(name)
-            if not value or catalog_xmin <= value['catalog_xmin']:
-                self._unready_logical_slots.remove(name)
+            # The logical slot on a replica is safe to use when the physical replica slot on the primary:
+            # 1. has a nonzero/non-null catalog_xmin
+            # 2. has a catalog_xmin that is not newer (greater) than the catalog_xmin of any slot on the standby
+            # 3. overtook the catalog_xmin of remembered values of logical slots on the primary.
+            if not value or self._unready_logical_slots[name] <= catalog_xmin <= value['catalog_xmin']:
+                del self._unready_logical_slots[name]
                 if value:
                     logger.info('Logical slot %s is safe to be used after a failover', name)
 
-    def copy_logical_slots(self, leader, slots):
+    def copy_logical_slots(self, cluster, create_slots):
+        leader = cluster.leader
+        slots = cluster.get_replication_slots(self._postgresql.name, 'replica', False, self._postgresql.major_version)
         with self._get_leader_connection_cursor(leader) as cur:
             try:
-                cur.execute("SELECT slot_name, catalog_xmin, "
+                cur.execute("SELECT slot_name, slot_type, datname, plugin, catalog_xmin, "
                             "pg_catalog.pg_wal_lsn_diff(confirmed_flush_lsn, '0/0')::bigint, "
                             "pg_catalog.pg_read_binary_file('pg_replslot/' || slot_name || '/state')"
-                            " FROM pg_catalog.pg_get_replication_slots() WHERE NOT pg_catalog.pg_is_in_recovery()"
-                            " AND slot_name = ANY(%s)", (slots,))
-                slots = {r[0]: {'catalog_xmin': r[1], 'confirmed_flush_lsn': r[2], 'data': r[3]} for r in cur}
+                            " FROM pg_catalog.pg_get_replication_slots() JOIN pg_catalog.pg_database ON datoid = oid"
+                            " WHERE NOT pg_catalog.pg_is_in_recovery() AND slot_name = ANY(%s)", (create_slots,))
+
+                create_slots = {}
+                for r in cur:
+                    if r[0] in slots:  # slot_name is defined in the global configuration
+                        slot = {'type': r[1], 'database': r[2], 'plugin': r[3],
+                                'catalog_xmin': r[4], 'confirmed_flush_lsn': r[5], 'data': r[6]}
+                        if compare_slots(slot, slots[r[0]]):
+                            create_slots[r[0]] = slot
+                        else:
+                            logger.warning('Will not copy the logical slot "%s" due to the configuration mismatch: ' +
+                                           'configuration=%s, slot on the primary=%s', r[0], slots[r[0]], slot)
             except Exception as e:
                 logger.error("Failed to copy logical slots from the %s via postgresql connection: %r", leader.name, e)
 
-        if isinstance(slots, dict) and self._postgresql.stop():
+        if isinstance(create_slots, dict) and create_slots and self._postgresql.stop():
             pg_replslot_dir = os.path.join(self._postgresql.data_dir, 'pg_replslot')
-            for name, value in slots.items():
+            for name, value in create_slots.items():
                 slot_dir = os.path.join(pg_replslot_dir, name)
                 slot_tmp_dir = slot_dir + '.tmp'
                 if os.path.exists(slot_tmp_dir):
@@ -288,7 +403,7 @@ class SlotsHandler(object):
                     shutil.rmtree(slot_dir)
                 os.rename(slot_tmp_dir, slot_dir)
                 fsync_dir(slot_dir)
-                self._unready_logical_slots.add(name)
+                self._unready_logical_slots[name] = None
             fsync_dir(pg_replslot_dir)
             self._postgresql.start()
 
@@ -298,6 +413,9 @@ class SlotsHandler(object):
         self._schedule_load_slots = self._force_readiness_check = value
 
     def on_promote(self):
+        if self._advance:
+            self._advance.on_promote()
+
         if self._unready_logical_slots:
             logger.warning('Logical replication slots that might be unsafe to use after promote: %s',
-                           self._unready_logical_slots)
+                           set(self._unready_logical_slots))

patroni/postgresql/validator.py

@@ -99,9 +99,11 @@ class String(namedtuple('String', 'version_from,version_till')):
 #  key - parameter name
 #  value - tuple or multiple tuples if something was changing in GUC across postgres versions
 parameters = CaseInsensitiveDict({
+    'allow_in_place_tablespaces': Bool(150000, None),
     'allow_system_table_mods': Bool(90300, None),
     'application_name': String(90300, None),
     'archive_command': String(90300, None),
+    'archive_library': String(150000, None),
     'archive_mode': (
         Bool(90300, 90500),
         EnumBool(90500, None, ('always',))
@@ -158,7 +160,10 @@ parameters = CaseInsensitiveDict({
     'cluster_name': String(90500, None),
     'commit_delay': Integer(90300, None, 0, 100000, None),
     'commit_siblings': Integer(90300, None, 0, 1000, None),
-    'compute_query_id': EnumBool(140000, None, ('auto',)),
+    'compute_query_id': (
+        EnumBool(140000, 150000, ('auto',)),
+        EnumBool(150000, None, ('auto', 'regress'))
+    ),
     'config_file': String(90300, None),
     'constraint_exclusion': EnumBool(90300, None, ('partition',)),
     'cpu_index_tuple_cost': Real(90300, None, 0, 1.79769e+308, None),
@@ -170,6 +175,7 @@ parameters = CaseInsensitiveDict({
     'DateStyle': String(90300, None),
     'db_user_namespace': Bool(90300, None),
     'deadlock_timeout': Integer(90300, None, 1, 2147483647, 'ms'),
+    'debug_discard_caches': Integer(150000, None, 0, 0, None),
     'debug_pretty_print': Bool(90300, None),
     'debug_print_parse': Bool(90300, None),
     'debug_print_plan': Bool(90300, None),
@@ -200,6 +206,7 @@ parameters = CaseInsensitiveDict({
     'enable_indexonlyscan': Bool(90300, None),
     'enable_indexscan': Bool(90300, None),
     'enable_material': Bool(90300, None),
+    'enable_memoize': Bool(150000, None),
     'enable_mergejoin': Bool(90300, None),
     'enable_nestloop': Bool(90300, None),
     'enable_parallel_append': Bool(110000, None),
@@ -294,6 +301,7 @@ parameters = CaseInsensitiveDict({
     '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_startup_progress_interval': Integer(150000, None, 0, 2147483647, 'ms'),
     '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),
@@ -368,6 +376,8 @@ parameters = CaseInsensitiveDict({
     'quote_all_identifiers': Bool(90300, None),
     'random_page_cost': Real(90300, None, 0, 1.79769e+308, None),
     'recovery_init_sync_method': Enum(140000, None, ('fsync', 'syncfs')),
+    'recovery_prefetch': EnumBool(150000, None, ('try',)),
+    'recursive_worktable_factor': Real(150000, None, 0.001, 1e+06, None),
     'remove_temp_files_after_crash': Bool(140000, None),
     'replacement_sort_tuples': Integer(90600, 110000, 0, 2147483647, None),
     'restart_after_crash': Bool(90300, None),
@@ -397,7 +407,8 @@ parameters = CaseInsensitiveDict({
     '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),
+    'stats_fetch_consistency': Enum(150000, None, ('none', 'cache', 'snapshot')),
+    'stats_temp_directory': String(90300, 150000),
     'superuser_reserved_connections': (
         Integer(90300, 90600, 0, 8388607, None),
         Integer(90600, None, 0, 262143, None)
@@ -463,8 +474,12 @@ parameters = CaseInsensitiveDict({
     '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_compression': (
+        Bool(90500, 150000),
+        EnumBool(150000, None, ('pglz', 'lz4', 'zstd'))
+    ),
     'wal_consistency_checking': String(100000, None),
+    'wal_decode_buffer_size': Integer(150000, None, 65536, 1073741823, 'B'),
     'wal_init_zero': Bool(120000, None),
     'wal_keep_segments': Integer(90300, 130000, 0, 2147483647, None),
     'wal_keep_size': Integer(130000, None, 0, 2147483647, 'MB'),

patroni/scripts/aws.py

@@ -3,10 +3,12 @@
 import json
 import logging
 import sys
-import boto.ec2
+import boto3
 
-from patroni.utils import Retry, RetryFailedError
-from patroni.request import get as requests_get
+from ..utils import Retry, RetryFailedError
+from ..request import get as requests_get
+
+from botocore.exceptions import ClientError
 
 logger = logging.getLogger(__name__)
 
@@ -16,7 +18,7 @@ class AWSConnection(object):
     def __init__(self, cluster_name):
         self.available = False
         self.cluster_name = cluster_name if cluster_name is not None else 'unknown'
-        self._retry = Retry(deadline=300, max_delay=30, max_tries=-1, retry_exceptions=(boto.exception.StandardError,))
+        self._retry = Retry(deadline=300, max_delay=30, max_tries=-1, retry_exceptions=(ClientError,))
         try:
             # get the instance id
             r = requests_get('http://169.254.169.254/latest/dynamic/instance-identity/document', timeout=2.1)
@@ -42,20 +44,22 @@ class AWSConnection(object):
 
     def _tag_ebs(self, conn, role):
         """ set tags, carrying the cluster name, instance role and instance id for the EBS storage """
-        tags = {'Name': 'spilo_' + self.cluster_name, 'Role': role, 'Instance': self.instance_id}
-        volumes = conn.get_all_volumes(filters={'attachment.instance-id': self.instance_id})
-        conn.create_tags([v.id for v in volumes], tags)
+        tags = [{'Key': 'Name', 'Value': 'spilo_' + self.cluster_name},
+                {'Key': 'Role', 'Value': role},
+                {'Key': 'Instance', 'Value': self.instance_id}]
+        volumes = conn.volumes.filter(Filters=[{'Name': 'attachment.instance-id', 'Values': [self.instance_id]}])
+        conn.create_tags(Resources=[v.id for v in volumes], Tags=tags)
 
     def _tag_ec2(self, conn, role):
         """ tag the current EC2 instance with a cluster role """
-        tags = {'Role': role}
-        conn.create_tags([self.instance_id], tags)
+        tags = [{'Key': 'Role', 'Value': role}]
+        conn.create_tags(Resources=[self.instance_id], Tags=tags)
 
     def on_role_change(self, new_role):
         if not self.available:
             return False
         try:
-            conn = self.retry(boto.ec2.connect_to_region, self.region)
+            conn = boto3.resource('ec2', region_name=self.region)
             self.retry(self._tag_ec2, conn, new_role)
             self.retry(self._tag_ebs, conn, new_role)
         except RetryFailedError:

patroni/utils.py

@@ -362,7 +362,7 @@ 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[0] = ','.join([h.strip().strip('[]') for h in t[0].split(',')])
     t.append(default_port)
     return t[0], int(t[1])
 

patroni/validator.py

@@ -37,6 +37,10 @@ def validate_host_port(host_port, listen=False, multiple_hosts=False):
             hosts = hosts.split(",")
         else:
             hosts = [hosts]
+        if "*" in hosts:
+            if len(hosts) != 1:
+                raise ConfigParseError("expecting '*' alone")
+            hosts = [p[-1][0] for p in socket.getaddrinfo(None, port, 0, socket.SOCK_STREAM, 0, socket.AI_PASSIVE)]
         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)
@@ -178,9 +182,11 @@ class Schema(object):
         self.validator = validator
 
     def __call__(self, data):
+        errors = []
         for i in self.validate(data):
             if not i.status:
-                print(i)
+                errors.append(str(i))
+        return errors
 
     def validate(self, data):
         self.data = data
@@ -364,6 +370,7 @@ schema = Schema({
   "postgresql": {
     "listen": validate_host_port_listen_multiple_hosts,
     "connect_address": validate_connect_address,
+    "proxy_address": validate_connect_address,
     "authentication": {
       "replication": userattributes,
       "superuser": userattributes,

patroni/version.py

@@ -1 +1 @@
-__version__ = '2.1.3'
+__version__ = '2.1.5'

patroni/watchdog/linux.py

@@ -16,10 +16,10 @@ IOC_DIRBITS = 2
 
 # Non-generic platform special cases
 machine = platform.machine()
-if machine in ['mips', 'sparc', 'powerpc', 'ppc64']:  # pragma: no cover
+if machine in ['mips', 'sparc', 'powerpc', 'ppc64', 'ppc64le']:  # pragma: no cover
     IOC_SIZEBITS = 13
     IOC_DIRBITS = 3
-    IOC_NONE, IOC_WRITE, IOC_READ = 1, 2, 4
+    IOC_NONE, IOC_WRITE, IOC_READ = 1, 4, 2
 elif machine == 'parisc':  # pragma: no cover
     IOC_WRITE, IOC_READ = 2, 1
 

postgres0.yml

@@ -99,6 +99,8 @@ bootstrap:
 postgresql:
   listen: 127.0.0.1:5432
   connect_address: 127.0.0.1:5432
+
+#  proxy_address: 127.0.0.1:5433  # The address of connection pool (e.g., pgbouncer) running next to Patroni/Postgres. Only for service discovery.
   data_dir: data/postgresql0
 #  bin_dir:
 #  config_dir:
@@ -119,7 +121,7 @@ postgresql:
     # 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: '.'
+    unix_socket_directories: '..'  # parent directory of data_dir
   # Additional fencing script executed after acquiring the leader lock but before promoting the replica
   #pre_promote: /path/to/pre_promote.sh
 

postgres1.yml

@@ -93,6 +93,7 @@ bootstrap:
 postgresql:
   listen: 127.0.0.1:5433
   connect_address: 127.0.0.1:5433
+#  proxy_address: 127.0.0.1:5434  # The address of connection pool (e.g., pgbouncer) running next to Patroni/Postgres. Only for service discovery.
   data_dir: data/postgresql1
 #  bin_dir:
 #  config_dir:
@@ -113,7 +114,7 @@ postgresql:
     # 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: '.'
+    unix_socket_directories: '..'  # parent directory of data_dir
   basebackup:
       - verbose
       - max-rate: 100M

postgres2.yml

@@ -90,6 +90,7 @@ bootstrap:
 postgresql:
   listen: 127.0.0.1:5434
   connect_address: 127.0.0.1:5434
+#  proxy_address: 127.0.0.1:5435  # The address of connection pool (e.g., pgbouncer) running next to Patroni/Postgres. Only for service discovery.
   data_dir: data/postgresql2
 #  bin_dir:
 #  config_dir:
@@ -110,7 +111,7 @@ postgresql:
     # 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: '.'
+    unix_socket_directories: '..'  # parent directory of data_dir
 tags:
     nofailover: false
     noloadbalance: false

release.sh

@@ -1,31 +1,28 @@
-#!/bin/sh
+#!/bin/bash
 
-if [ $# -ne 1 ]; then
-    >&2 echo "usage: $0 <version>"
-    exit 1
-fi
-
-readonly VERSIONFILE="patroni/version.py"
+# Release process:
+# 1. Open a PR that updates release notes and Patroni version
+# 2. Merge it
+# 3. Run release.sh
+# 4. After the new tag is pushed, the .github/workflows/release.yaml will run tests and upload the new package to test.pypi.org
+# 5. Once the release is created, the .github/workflows/release.yaml will run tests and upload the new package to pypi.org
 
 ## Bail out on any non-zero exitcode from the called processes
 set -xe
 
-python3 --version
+if python3 --version &> /dev/null; then
+    alias python=python3
+    shopt -s expand_aliases
+fi
+
+python --version
 git --version
 
-version=$1
+version=$(python -c 'from patroni.version import __version__; print(__version__)')
 
-sed -i "s/__version__ = .*/__version__ = '${version}'/"  "${VERSIONFILE}"
-python3 setup.py clean
-python3 setup.py test
-python3 setup.py flake8
+python setup.py clean
+python setup.py test
+python setup.py flake8
 
-git add "${VERSIONFILE}"
-
-git commit -m "Bumped version to $version"
-git push
-
-python3 setup.py sdist bdist_wheel upload
-
-git tag v${version}
+git tag "v$version"
 git push --tags

requirements.dev.txt

@@ -1,7 +1,7 @@
 psycopg2-binary
 behave
 coverage
-flake8
+flake8>=3.0.0
 mock
 pytest-cov
 pytest

requirements.txt

@@ -1,6 +1,6 @@
 urllib3>=1.19.1,!=1.21
 ipaddress; python_version=="2.7"
-boto
+boto3
 PyYAML
 six >= 1.7
 kazoo>=1.3.1

setup.py

@@ -18,12 +18,12 @@ MAIN_PACKAGE = NAME
 DESCRIPTION = 'PostgreSQL High-Available orchestrator and CLI'
 LICENSE = 'The MIT License'
 URL = 'https://github.com/zalando/patroni'
-AUTHOR = 'Alexander Kukushkin, Dmitrii Dolgov, Oleksii Kliukin'
-AUTHOR_EMAIL = 'alexander.kukushkin@zalando.de, [email protected], [email protected]'
+AUTHOR = 'Alexander Kukushkin, Polina Bungina'
+AUTHOR_EMAIL = 'akukushkin@microsoft.com, [email protected]'
 KEYWORDS = 'etcd governor patroni postgresql postgres ha haproxy confd' +\
     ' zookeeper exhibitor consul streaming replication kubernetes k8s'
 
-EXTRAS_REQUIRE = {'aws': ['boto'], 'etcd': ['python-etcd'], 'etcd3': ['python-etcd'],
+EXTRAS_REQUIRE = {'aws': ['boto3'], 'etcd': ['python-etcd'], 'etcd3': ['python-etcd'],
                   'consul': ['python-consul'], 'exhibitor': ['kazoo'], 'zookeeper': ['kazoo'],
                   'kubernetes': [], 'raft': ['pysyncobj', 'cryptography']}
 COVERAGE_XML = True
@@ -93,12 +93,10 @@ class Flake8(_Command):
         return [package for package in self.package_files()] + ['tests', 'setup.py']
 
     def run(self):
-        from flake8.main import application
+        from flake8.main.cli import main
 
         logging.getLogger().setLevel(logging.ERROR)
-        flake8 = application.Application()
-        flake8.run(self.targets())
-        flake8.exit()
+        main(self.targets())
 
 
 class PyTest(_Command):

tests/__init__.py

@@ -91,10 +91,10 @@ class MockCursor(object):
             raise psycopg.OperationalError()
         elif sql.startswith('RetryFailedError'):
             raise RetryFailedError('retry')
-        elif sql.startswith('SELECT catalog_xmin'):
-            self.results = [(100, 501)]
         elif sql.startswith('SELECT slot_name, catalog_xmin'):
-            self.results = [('ls', 100, 500, b'123456')]
+            self.results = [('postgresql0', 100), ('ls', 100)]
+        elif sql.startswith('SELECT slot_name, slot_type, datname, plugin, catalog_xmin'):
+            self.results = [('ls', 'logical', 'a', 'b', 100, 500, b'123456')]
         elif sql.startswith('SELECT slot_name'):
             self.results = [('blabla', 'physical'), ('foobar', 'physical'), ('ls', 'logical', 'a', 'b', 5, 100, 500)]
         elif sql.startswith('SELECT CASE WHEN pg_catalog.pg_is_in_recovery()'):
@@ -188,7 +188,8 @@ class PostgresInit(unittest.TestCase):
         self.p = Postgresql({'name': 'postgresql0', 'scope': 'batman', 'data_dir': data_dir,
                              'config_dir': data_dir, 'retry_timeout': 10,
                              'krbsrvname': 'postgres', 'pgpass': os.path.join(data_dir, 'pgpass0'),
-                             'listen': '127.0.0.2, 127.0.0.3:5432', 'connect_address': '127.0.0.2:5432',
+                             'listen': '127.0.0.2, 127.0.0.3:5432',
+                             'connect_address': '127.0.0.2:5432', 'proxy_address': '127.0.0.2:5433',
                              'authentication': {'superuser': {'username': 'foo', 'password': 'test'},
                                                 'replication': {'username': '', 'password': 'rep-pass'},
                                                 'rewind': {'username': 'rewind', 'password': 'test'}},

tests/test_api.py

@@ -12,6 +12,7 @@ from patroni.ha import _MemberStatus
 from patroni.utils import tzutc
 from six import BytesIO as IO
 from six.moves import BaseHTTPServer
+from six.moves.socketserver import ThreadingMixIn
 from . import psycopg_connect, MockCursor
 from .test_ha import get_cluster_initialized_without_leader
 
@@ -46,6 +47,10 @@ class MockPostgresql(object):
     def replica_cached_timeline(_):
         return 2
 
+    @staticmethod
+    def is_running():
+        return True
+
 
 class MockWatchdog(object):
     is_healthy = False
@@ -129,6 +134,10 @@ class MockPatroni(object):
     def sighup_handler():
         pass
 
+    @staticmethod
+    def api_sigterm():
+        pass
+
 
 class MockRequest(object):
 
@@ -180,11 +189,13 @@ class TestRestApiHandler(unittest.TestCase):
         MockPatroni.dcs.cluster.is_synchronous_mode = Mock(return_value=True)
         with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'replica'})):
             MockRestApiServer(RestApiHandler, 'GET /synchronous')
+            MockRestApiServer(RestApiHandler, 'GET /read-only-sync')
         with patch.object(RestApiHandler, 'get_postgresql_status', Mock(return_value={'role': 'replica'})):
             MockPatroni.dcs.cluster.sync.members = []
             MockRestApiServer(RestApiHandler, 'GET /asynchronous')
         with patch.object(MockHa, 'is_leader', Mock(return_value=True)):
             MockRestApiServer(RestApiHandler, 'GET /replica')
+            MockRestApiServer(RestApiHandler, 'GET /read-only-sync')
             with patch.object(MockHa, 'is_standby_cluster', Mock(return_value=True)):
                 MockRestApiServer(RestApiHandler, 'GET /standby_leader')
         MockPatroni.dcs.cluster = None
@@ -284,7 +295,12 @@ class TestRestApiHandler(unittest.TestCase):
     def test_do_OPTIONS(self):
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'OPTIONS / HTTP/1.0'))
 
-    def test_do_GET_liveness(self):
+    def test_do_HEAD(self):
+        self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'HEAD / HTTP/1.0'))
+
+    @patch.object(MockPatroni, 'dcs')
+    def test_do_GET_liveness(self, mock_dcs):
+        mock_dcs.ttl.return_value = PropertyMock(30)
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'GET /liveness HTTP/1.0'))
 
     def test_do_GET_readiness(self):
@@ -359,6 +375,11 @@ class TestRestApiHandler(unittest.TestCase):
     def test_do_POST_reload(self):
         self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'POST /reload HTTP/1.0' + self._authorization))
 
+    @patch('os.environ', {'BEHAVE_DEBUG': 'true'})
+    @patch('os.name', 'nt')
+    def test_do_POST_sigterm(self):
+        self.assertIsNotNone(MockRestApiServer(RestApiHandler, 'POST /sigterm HTTP/1.0' + self._authorization))
+
     @patch.object(MockPatroni, 'dcs')
     def test_do_POST_restart(self, mock_dcs):
         mock_dcs.get_cluster.return_value.is_paused.return_value = False
@@ -575,32 +596,14 @@ class TestRestApiServer(unittest.TestCase):
     def test_socket_error(self):
         self.assertRaises(socket.error, MockRestApiServer, Mock(), '', {'listen': '*:8008'})
 
-    @patch.object(MockRestApiServer, 'finish_request', Mock())
+    @patch.object(ThreadingMixIn, 'process_request_thread', Mock())
     def test_process_request_thread(self):
-        mock_socket = Mock()
-        self.srv.process_request_thread((mock_socket, 1), '2')
-        mock_socket.context.wrap_socket.side_effect = socket.error
-        self.srv.process_request_thread((mock_socket, 1), '2')
-
-    @patch.object(socket.socket, 'accept')
-    def test_get_request(self, mock_accept):
-        newsock = Mock()
-        mock_accept.return_value = (newsock, '2')
-        self.srv.socket = Mock()
-        self.assertEqual(self.srv.get_request(), ((self.srv.socket, newsock), '2'))
+        self.srv.process_request_thread(Mock(), '2')
 
     @patch.object(MockRestApiServer, 'process_request', Mock(side_effect=RuntimeError))
+    @patch.object(MockRestApiServer, 'get_request', Mock(return_value=(Mock(), ('127.0.0.1', 55555))))
     def test_process_request_error(self):
-        mock_address = ('127.0.0.1', 55555)
-        mock_socket = Mock()
-        mock_ssl_socket = (Mock(), Mock())
-        for mock_request in (mock_socket, mock_ssl_socket):
-            with patch.object(
-                MockRestApiServer,
-                'get_request',
-                Mock(return_value=(mock_request, mock_address))
-            ):
-                self.srv._handle_request_noblock()
+        self.srv._handle_request_noblock()
 
     @patch('ssl._ssl._test_decode_cert', Mock())
     def test_reload_local_certificate(self):

tests/test_aws.py

@@ -1,4 +1,4 @@
-import boto.ec2
+import botocore
 import sys
 import unittest
 import urllib3
@@ -8,21 +8,27 @@ from collections import namedtuple
 from patroni.scripts.aws import AWSConnection, main as _main
 
 
-class MockEc2Connection(object):
+class MockVolumes(object):
 
     @staticmethod
-    def get_all_volumes(*args, **kwargs):
+    def filter(*args, **kwargs):
         oid = namedtuple('Volume', 'id')
         return [oid(id='a'), oid(id='b')]
 
+
+class MockEc2Connection(object):
+
+    volumes = MockVolumes()
+
     @staticmethod
-    def create_tags(objects, *args, **kwargs):
-        if len(objects) == 0:
-            raise boto.exception.BotoServerError(503, 'Service Unavailable', 'Request limit exceeded')
+    def create_tags(Resources, **kwargs):
+        if len(Resources) == 0:
+            raise botocore.exceptions.ClientError({'Error': {'Code': 503, 'Message': 'Request limit exceeded'}},
+                                                  'create_tags')
         return True
 
 
-@patch('boto.ec2.connect_to_region', Mock(return_value=MockEc2Connection()))
+@patch('boto3.resource', Mock(return_value=MockEc2Connection()))
 class TestAWSConnection(unittest.TestCase):
 
     @patch('patroni.scripts.aws.requests_get', Mock(return_value=urllib3.HTTPResponse(
@@ -32,7 +38,7 @@ class TestAWSConnection(unittest.TestCase):
 
     def test_on_role_change(self):
         self.assertTrue(self.conn.on_role_change('master'))
-        with patch.object(MockEc2Connection, 'get_all_volumes', Mock(return_value=[])):
+        with patch.object(MockVolumes, 'filter', Mock(return_value=[])):
             self.conn._retry.max_tries = 1
             self.assertFalse(self.conn.on_role_change('master'))
 
@@ -46,8 +52,7 @@ class TestAWSConnection(unittest.TestCase):
         conn = AWSConnection('test')
         self.assertFalse(conn.aws_available())
 
-    @patch('patroni.scripts.aws.requests_get', Mock(return_value=urllib3.HTTPResponse(
-        status=200, body=b'{"instanceId": "012345", "region": "eu-west-1"}')))
+    @patch('patroni.scripts.aws.requests_get', Mock(return_value=urllib3.HTTPResponse(status=503, body=b'Error')))
     @patch('sys.exit', Mock())
     def test_main(self):
         self.assertIsNone(_main())

tests/test_config.py

@@ -40,6 +40,7 @@ class TestConfig(unittest.TestCase):
             'PATRONI_RESTAPI_ALLOWLIST_INCLUDE_MEMBERS': 'on',
             'PATRONI_POSTGRESQL_LISTEN': '0.0.0.0:5432',
             'PATRONI_POSTGRESQL_CONNECT_ADDRESS': '127.0.0.1:5432',
+            'PATRONI_POSTGRESQL_PROXY_ADDRESS': '127.0.0.1:5433',
             'PATRONI_POSTGRESQL_DATA_DIR': 'data/postgres0',
             'PATRONI_POSTGRESQL_CONFIG_DIR': 'data/postgres0',
             'PATRONI_POSTGRESQL_PGPASS': '/tmp/pgpass0',

tests/test_consul.py

@@ -91,7 +91,7 @@ class TestConsul(unittest.TestCase):
         Consul({'ttl': 30, 'scope': 't_', 'name': 'p', 'url': 'https://l:1', 'retry_timeout': 10,
                 'verify': 'on', 'cert': 'bar', 'cacert': 'buz', 'register_service': True})
         self.c = Consul({'ttl': 30, 'scope': 'test', 'name': 'postgresql1', 'host': 'localhost:1', 'retry_timeout': 10,
-                         'register_service': True})
+                         'register_service': True, 'service_check_tls_server_name': True})
         self.c._base_path = '/service/good'
         self.c.get_cluster()
 
@@ -161,6 +161,7 @@ class TestConsul(unittest.TestCase):
         self.c.write_leader_optime('1')
 
     @patch.object(consul.Consul.Session, 'renew', Mock())
+    @patch.object(consul.Consul.KV, 'put', Mock(side_effect=ConsulException))
     def test_update_leader(self):
         self.c.update_leader(12345)
 
@@ -217,6 +218,7 @@ class TestConsul(unittest.TestCase):
         d['role'] = 'bla'
         self.assertIsNone(self.c.update_service({}, d))
 
+    @patch.object(consul.Consul.KV, 'put', Mock(side_effect=ConsulException))
     def test_reload_config(self):
         self.assertEqual([], self.c._service_tags)
         self.c.reload_config({'consul': {'token': 'foo', 'register_service': True, 'service_tags': ['foo']},

tests/test_ctl.py

@@ -7,10 +7,11 @@ from datetime import datetime, timedelta
 from mock import patch, Mock
 from patroni.ctl import ctl, store_config, load_config, output_members, get_dcs, parse_dcs, \
     get_all_members, get_any_member, get_cursor, query_member, configure, PatroniCtlException, apply_config_changes, \
-    format_config_for_editing, show_diff, invoke_editor, format_pg_version, CONFIG_FILE_PATH
+    format_config_for_editing, show_diff, invoke_editor, format_pg_version, CONFIG_FILE_PATH, PatronictlPrettyTable
 from patroni.dcs.etcd import AbstractEtcdClientWithFailover, Failover
 from patroni.psycopg import OperationalError
 from patroni.utils import tzutc
+from prettytable import PrettyTable, ALL
 from urllib3 import PoolManager
 
 from . import MockConnect, MockCursor, MockResponse, psycopg_connect
@@ -562,7 +563,8 @@ class TestCtl(unittest.TestCase):
 
     @patch('sys.stdout.isatty', return_value=False)
     @patch('patroni.ctl.markup_to_pager')
-    def test_show_diff(self, mock_markup_to_pager, mock_isatty):
+    @patch('patroni.ctl.find_executable', return_value=None)
+    def test_show_diff(self, mock_find_executable, mock_markup_to_pager, mock_isatty):
         show_diff("foo:\n  bar: 1\n", "foo:\n  bar: 2\n")
         mock_markup_to_pager.assert_not_called()
 
@@ -570,10 +572,10 @@ class TestCtl(unittest.TestCase):
         show_diff("foo:\n  bar: 1\n", "foo:\n  bar: 2\n")
         mock_markup_to_pager.assert_called_once()
 
-        with patch('patroni.ctl.find_executable', Mock(return_value=None)):
-            show_diff("foo:\n  bar: 1\n", "foo:\n  bar: 2\n")
+        show_diff("foo:\n  bar: 1\n", "foo:\n  bar: 2\n")
 
         # Test that unicode handling doesn't fail with an exception
+        mock_find_executable.return_value = '/usr/bin/less'
         show_diff(b"foo:\n  bar: \xc3\xb6\xc3\xb6\n".decode('utf-8'),
                   b"foo:\n  bar: \xc3\xbc\xc3\xbc\n".decode('utf-8'))
 
@@ -591,6 +593,7 @@ class TestCtl(unittest.TestCase):
         self.runner.invoke(ctl, ['show-config', 'dummy'])
 
     @patch('patroni.ctl.get_dcs')
+    @patch('subprocess.call', Mock(return_value=0))
     def test_edit_config(self, mock_get_dcs):
         mock_get_dcs.return_value = self.e
         mock_get_dcs.return_value.get_cluster = get_cluster_initialized_with_leader
@@ -646,3 +649,24 @@ class TestCtl(unittest.TestCase):
             result = self.runner.invoke(ctl, ['reinit', 'alpha', 'other', '--wait'], input='y\ny')
         self.assertIn("Waiting for reinitialize to complete on: other", result.output)
         self.assertIn("Reinitialize is completed on: other", result.output)
+
+
+class TestPatronictlPrettyTable(unittest.TestCase):
+
+    def setUp(self):
+        self.pt = PatronictlPrettyTable(' header', ['foo', 'bar'], hrules=ALL)
+
+    def test__get_hline(self):
+        expected = '+-----+-----+'
+        self.pt._hrule = expected
+        self.assertEqual(self.pt._hrule, '+ header----+')
+        self.assertFalse(self.pt._is_first_hline())
+        self.assertEqual(self.pt._hrule, expected)
+
+    @patch.object(PrettyTable, '_stringify_hrule', Mock(return_value='+-----+-----+'))
+    def test__stringify_hrule(self):
+        self.assertEqual(self.pt._stringify_hrule((), 'top_'), '+ header----+')
+        self.assertFalse(self.pt._is_first_hline())
+
+    def test_output(self):
+        self.assertEqual(str(self.pt), '+ header----+\n| foo | bar |\n+-----+-----+')

tests/test_ha.py

@@ -45,6 +45,10 @@ def get_cluster_not_initialized_without_leader(cluster_config=None):
     return get_cluster(None, None, [], None, SyncState(None, None, None), cluster_config)
 
 
+def get_cluster_bootstrapping_without_leader(cluster_config=None):
+    return get_cluster("", None, [], None, SyncState(None, None, None), cluster_config)
+
+
 def get_cluster_initialized_without_leader(leader=False, failover=None, sync=None, cluster_config=None):
     m1 = Member(0, 'leader', 28, {'conn_url': 'postgres://replicator:[email protected]:5435/postgres',
                                   'api_url': 'http://127.0.0.1:8008/patroni', 'xlog_location': 4})
@@ -284,6 +288,18 @@ class TestHa(PostgresInit):
             self.ha.patroni.config.set_dynamic_configuration({'maximum_lag_on_failover': 10})
             self.assertEqual(self.ha.run_cycle(), 'terminated crash recovery because of startup timeout')
 
+    @patch.object(Rewind, 'ensure_clean_shutdown', Mock())
+    @patch.object(Rewind, 'rewind_or_reinitialize_needed_and_possible', Mock(return_value=True))
+    @patch.object(Rewind, 'can_rewind', PropertyMock(return_value=True))
+    def test_crash_recovery_before_rewind(self):
+        self.p.is_leader = false
+        self.p.is_running = false
+        self.p.controldata = lambda: {'Database cluster state': 'in archive recovery',
+                                      'Database system identifier': SYSID}
+        self.ha._rewind.trigger_check_diverged_lsn()
+        self.ha.cluster = get_cluster_initialized_with_leader()
+        self.assertEqual(self.ha.run_cycle(), 'doing crash recovery in a single user mode')
+
     @patch.object(Rewind, 'rewind_or_reinitialize_needed_and_possible', Mock(return_value=True))
     @patch.object(Rewind, 'can_rewind', PropertyMock(return_value=True))
     def test_recover_with_rewind(self):
@@ -302,6 +318,7 @@ class TestHa(PostgresInit):
             self.assertEqual(self.ha.run_cycle(), 'fake')
 
     @patch.object(Rewind, 'rewind_or_reinitialize_needed_and_possible', Mock(return_value=True))
+    @patch.object(Rewind, 'should_remove_data_directory_on_diverged_timelines', PropertyMock(return_value=True))
     @patch.object(Bootstrap, 'create_replica', Mock(return_value=1))
     def test_recover_with_reinitialize(self):
         self.p.is_running = false
@@ -392,6 +409,7 @@ class TestHa(PostgresInit):
         self.ha.has_lock = true
         self.assertEqual(self.ha.run_cycle(), 'no action. I am (postgresql0), the leader with the lock')
 
+    @patch.object(Postgresql, '_wait_for_connection_close', Mock())
     def test_demote_because_not_having_lock(self):
         self.ha.cluster.is_unlocked = false
         with patch.object(Watchdog, 'is_running', PropertyMock(return_value=True)):
@@ -456,6 +474,11 @@ class TestHa(PostgresInit):
         self.p.can_create_replica_without_replication_connection = MagicMock(return_value=True)
         self.assertEqual(self.ha.bootstrap(), 'trying to bootstrap (without leader)')
 
+    def test_bootstrap_not_running_concurrently(self):
+        self.ha.cluster = get_cluster_bootstrapping_without_leader()
+        self.p.can_create_replica_without_replication_connection = MagicMock(return_value=True)
+        self.assertEqual(self.ha.bootstrap(), 'waiting for leader to bootstrap')
+
     def test_bootstrap_initialize_lock_failed(self):
         self.ha.cluster = get_cluster_not_initialized_without_leader()
         self.assertEqual(self.ha.bootstrap(), 'failed to acquire initialize lock')
@@ -630,12 +653,60 @@ class TestHa(PostgresInit):
         # same as previous, but set the current member to nofailover. In no case it should be elected as a leader
         self.ha.patroni.nofailover = True
         self.assertEqual(self.ha.run_cycle(), 'following a different leader because I am not allowed to promote')
-        # in sync mode only the sync node is allowed to take over
-        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, 'leader', 'other', None))
-        self.ha.patroni.nofailover = False
+
+    def test_manual_failover_process_no_leader_in_synchronous_mode(self):
         self.ha.is_synchronous_mode = true
+        self.p.is_leader = false
+
+        # switchover to a specific node, which name doesn't match our name (postgresql0)
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, 'leader', 'other', None))
         self.assertEqual(self.ha.run_cycle(), 'following a different leader because i am not the healthiest node')
 
+        # switchover to our node (postgresql0), which name is not in sync nodes list
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, 'leader', 'postgresql0', None),
+                                                                 sync=('leader1', 'blabla'))
+        self.assertEqual(self.ha.run_cycle(), 'following a different leader because i am not the healthiest node')
+
+        # switchover from a specific leader, but our name (postgresql0) is not in the sync nodes list
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, 'leader', '', None),
+                                                                 sync=('leader', 'blabla'))
+        self.assertEqual(self.ha.run_cycle(), 'following a different leader because i am not the healthiest node')
+
+        # switchover from a specific leader, but the only sync node (us, postgresql0) has nofailover tag
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, 'leader', '', None),
+                                                                 sync=('postgresql0'))
+        self.ha.patroni.nofailover = True
+        self.assertEqual(self.ha.run_cycle(), 'following a different leader because I am not allowed to promote')
+        self.ha.patroni.nofailover = False
+
+        # manual failover when our name (postgresql0) isn't in the /sync key and the `other` node is not available
+        self.ha.fetch_node_status = get_node_status(nofailover=True)  # accessible, in_recovery
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, '', 'other', None),
+                                                                 sync=('leader1', 'blabla'))
+        self.assertEqual(self.ha.run_cycle(), 'following a different leader because i am not the healthiest node')
+
+        # manual failover when the `other` node isn't available but our name is in the /sync key
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, '', 'other', None),
+                                                                 sync=('leader1', 'postgresql0'))
+        self.p.pick_synchronous_standby = Mock(return_value=([], []))
+        self.ha.dcs.write_sync_state = true
+        self.assertEqual(self.ha.run_cycle(), 'promoted self to leader by acquiring session lock')
+
+        # manual failover to our node (postgresql0),
+        # which name is not in sync nodes list (the leader and all sync nodes are not available)
+        self.p.set_role('replica')
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, '', 'postgresql0', None),
+                                                                 sync=('leader1', 'other'))
+        self.assertEqual(self.ha.run_cycle(), 'promoted self to leader by acquiring session lock')
+
+        # manual failover to our node (postgresql0),
+        # which name is not in sync nodes list (some sync nodes are available)
+        self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, '', 'postgresql0', None),
+                                                                 sync=('leader1', 'other'))
+        self.p.set_role('replica')
+        self.p.pick_synchronous_standby = Mock(return_value=(['leader1'], ['leader1']))
+        self.assertEqual(self.ha.run_cycle(), 'promoted self to leader by acquiring session lock')
+
     def test_manual_failover_process_no_leader_in_pause(self):
         self.ha.is_paused = true
         self.ha.cluster = get_cluster_initialized_without_leader(failover=Failover(0, '', 'other', None))
@@ -1120,17 +1191,19 @@ class TestHa(PostgresInit):
         self.ha.shutdown()
 
     @patch('time.sleep', Mock())
-    def test_leader_with_empty_directory(self):
+    def test_leader_with_not_accessible_data_directory(self):
         self.ha.cluster = get_cluster_initialized_with_leader()
         self.ha.has_lock = true
-        self.p.data_directory_empty = true
-        self.assertEqual(self.ha.run_cycle(), 'released leader key voluntarily as data dir empty and currently leader')
+        self.p.data_directory_empty = Mock(side_effect=OSError(5, "Input/output error: '{}'".format(self.p.data_dir)))
+        self.assertEqual(self.ha.run_cycle(),
+                         'released leader key voluntarily as data dir not accessible and currently leader')
         self.assertEqual(self.p.role, 'uninitialized')
 
         # as has_lock is mocked out, we need to fake the leader key release
         self.ha.has_lock = false
-        # will not say bootstrap from leader as replica can't self elect
-        self.assertEqual(self.ha.run_cycle(), "trying to bootstrap from replica 'other'")
+        # will not say bootstrap because data directory is not accessible
+        self.assertEqual(self.ha.run_cycle(),
+                         "data directory is not accessible: [Errno 5] Input/output error: '{}'".format(self.p.data_dir))
 
     @patch('patroni.postgresql.mtime', Mock(return_value=1588316884))
     @patch.object(builtins, 'open', mock_open(read_data=('1\t0/40159C0\tno recovery target specified\n\n'
@@ -1199,9 +1272,12 @@ class TestHa(PostgresInit):
     @patch('os.rename', Mock())
     @patch('patroni.postgresql.Postgresql.is_starting', Mock(return_value=False))
     @patch.object(builtins, 'open', mock_open())
-    @patch.object(SlotsHandler, 'sync_replication_slots', Mock(return_value=['foo']))
+    @patch.object(ConfigHandler, 'check_recovery_conf', Mock(return_value=(False, False)))
+    @patch.object(Postgresql, 'major_version', PropertyMock(return_value=130000))
+    @patch.object(SlotsHandler, 'sync_replication_slots', Mock(return_value=['ls']))
     def test_follow_copy(self):
         self.ha.cluster.is_unlocked = false
+        self.ha.cluster.config.data['slots'] = {'ls': {'database': 'a', 'plugin': 'b'}}
         self.p.is_leader = false
         self.assertTrue(self.ha.run_cycle().startswith('Copying logical slots'))
 

tests/test_kubernetes.py

@@ -1,9 +1,10 @@
+import datetime
 import json
 import socket
 import time
 import unittest
 
-from mock import Mock, mock_open, patch
+from mock import Mock, PropertyMock, mock_open, patch
 from patroni.dcs.kubernetes import k8s_client, k8s_config, K8sConfig, K8sConnectionFailed,\
         K8sException, K8sObject, Kubernetes, KubernetesError, KubernetesRetriableException,\
         Retry, RetryFailedError, SERVICE_HOST_ENV_NAME, SERVICE_PORT_ENV_NAME
@@ -79,6 +80,28 @@ class TestK8sConfig(unittest.TestCase):
                 self.assertRaises(k8s_config.ConfigException, k8s_config.load_incluster_config)
             k8s_config.load_incluster_config()
             self.assertEqual(k8s_config.server, 'https://a:1')
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer a')
+
+    def test_refresh_token(self):
+        with patch('os.environ', {SERVICE_HOST_ENV_NAME: 'a', SERVICE_PORT_ENV_NAME: '1'}),\
+                patch('os.path.isfile', Mock(side_effect=[True, True, False, True, True, True])),\
+                patch.object(builtins, 'open', Mock(side_effect=[
+                    mock_open(read_data='cert')(), mock_open(read_data='a')(),
+                    mock_open()(), mock_open(read_data='b')(), mock_open(read_data='c')()])):
+            k8s_config.load_incluster_config(token_refresh_interval=datetime.timedelta(milliseconds=100))
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer a')
+            time.sleep(0.1)
+            # token file doesn't exist
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer a')
+            # token file is empty
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer a')
+            # token refreshed
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer b')
+            time.sleep(0.1)
+            # token refreshed
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer c')
+            # no need to refresh token
+            self.assertEqual(k8s_config.headers.get('authorization'), 'Bearer c')
 
     def test_load_kube_config(self):
         config = {
@@ -212,7 +235,9 @@ class TestKubernetesConfigMaps(BaseTestKubernetes):
             self.k.manual_failover('foo', 'bar')
 
     def test_set_config_value(self):
-        self.k.set_config_value('{}')
+        with patch.object(k8s_client.CoreV1Api, 'patch_namespaced_config_map',
+                          Mock(side_effect=k8s_client.rest.ApiException(409, '')), create=True):
+            self.k.set_config_value('{}', 1)
 
     @patch.object(k8s_client.CoreV1Api, 'patch_namespaced_pod', create=True)
     def test_touch_member(self, mock_patch_namespaced_pod):
@@ -290,11 +315,32 @@ class TestKubernetesEndpoints(BaseTestKubernetes):
     @patch.object(k8s_client.CoreV1Api, 'patch_namespaced_pod', mock_namespaced_kind, create=True)
     @patch.object(k8s_client.CoreV1Api, 'create_namespaced_endpoints', mock_namespaced_kind, create=True)
     @patch.object(k8s_client.CoreV1Api, 'create_namespaced_service',
-                  Mock(side_effect=[True, False, k8s_client.rest.ApiException(500, '')]), create=True)
-    def test__create_config_service(self):
+                  Mock(side_effect=[True,
+                                    False,
+                                    k8s_client.rest.ApiException(409, ''),
+                                    k8s_client.rest.ApiException(403, ''),
+                                    k8s_client.rest.ApiException(500, ''),
+                                    Exception("Unexpected")
+                                    ]), create=True)
+    @patch('patroni.dcs.kubernetes.logger.exception')
+    def test__create_config_service(self, mock_logger_exception):
         self.assertIsNotNone(self.k.patch_or_create_config({'foo': 'bar'}))
         self.assertIsNotNone(self.k.patch_or_create_config({'foo': 'bar'}))
+
+        self.k.patch_or_create_config({'foo': 'bar'})
+        mock_logger_exception.assert_not_called()
+
+        self.k.patch_or_create_config({'foo': 'bar'})
+        mock_logger_exception.assert_not_called()
+
+        self.k.patch_or_create_config({'foo': 'bar'})
+        mock_logger_exception.assert_called_once()
+        self.assertEqual(('create_config_service failed',), mock_logger_exception.call_args[0])
+        mock_logger_exception.reset_mock()
+
         self.k.touch_member({'state': 'running', 'role': 'replica'})
+        mock_logger_exception.assert_called_once()
+        self.assertEqual(('create_config_service failed',), mock_logger_exception.call_args[0])
 
 
 class TestCacheBuilder(BaseTestKubernetes):
@@ -322,3 +368,18 @@ class TestCacheBuilder(BaseTestKubernetes):
     def test__list(self):
         self.k._pods._func = Mock(side_effect=Exception)
         self.assertRaises(Exception, self.k._pods._list)
+
+    @patch('patroni.dcs.kubernetes.ObjectCache._watch', Mock(return_value=None))
+    def test__do_watch(self):
+        self.assertRaises(AttributeError, self.k._kinds._do_watch, '1')
+
+    @patch.object(k8s_client.CoreV1Api, 'list_namespaced_config_map', mock_list_namespaced_config_map, create=True)
+    @patch('patroni.dcs.kubernetes.ObjectCache._watch')
+    def test_kill_stream(self, mock_watch):
+        self.k._kinds.kill_stream()
+        mock_watch.return_value.read_chunked.return_value = []
+        mock_watch.return_value.connection.sock.close.side_effect = Exception
+        self.k._kinds._do_watch('1')
+        self.k._kinds.kill_stream()
+        type(mock_watch.return_value).connection = PropertyMock(side_effect=Exception)
+        self.k._kinds.kill_stream()

tests/test_patroni.py

@@ -50,12 +50,16 @@ class MockFrozenImporter(object):
 @patch.object(etcd.Client, 'read', etcd_read)
 class TestPatroni(unittest.TestCase):
 
+    @patch('sys.argv', ['patroni.py'])
     def test_no_config(self):
         self.assertRaises(SystemExit, patroni_main)
 
     @patch('sys.argv', ['patroni.py', '--validate-config', 'postgres0.yml'])
+    @patch('socket.socket.connect_ex', Mock(return_value=1))
     def test_validate_config(self):
         self.assertRaises(SystemExit, patroni_main)
+        with patch.object(config.Config, '__init__', Mock(return_value=None)):
+            self.assertRaises(SystemExit, patroni_main)
 
     @patch('pkgutil.iter_importers', Mock(return_value=[MockFrozenImporter()]))
     @patch('sys.frozen', Mock(return_value=True), create=True)

tests/test_postgresql.py

@@ -95,7 +95,7 @@ class TestPostgresql(BaseTestPostgresql):
     @patch('subprocess.call', Mock(return_value=0))
     @patch('os.rename', Mock())
     @patch('patroni.postgresql.CallbackExecutor', Mock())
-    @patch.object(Postgresql, 'get_major_version', Mock(return_value=130000))
+    @patch.object(Postgresql, 'get_major_version', Mock(return_value=140000))
     @patch.object(Postgresql, 'is_running', Mock(return_value=True))
     def setUp(self):
         super(TestPostgresql, self).setUp()
@@ -287,6 +287,8 @@ class TestPostgresql(BaseTestPostgresql):
         mock_get_pg_settings.side_effect = Exception
         with patch('patroni.postgresql.config.mtime', mock_mtime):
             self.assertEqual(self.p.config.check_recovery_conf(None), (True, True))
+        with patch.object(Postgresql, 'is_starting', Mock(return_value=True)):
+            self.assertEqual(self.p.config.check_recovery_conf(None), (False, False))
 
     @patch.object(Postgresql, 'major_version', PropertyMock(return_value=100000))
     @patch.object(Postgresql, 'primary_conninfo', Mock(return_value='host=1'))
@@ -345,7 +347,7 @@ class TestPostgresql(BaseTestPostgresql):
     @patch('subprocess.Popen')
     def test_latest_checkpoint_location(self, mock_popen):
         mock_popen.return_value.communicate.return_value = (None, None)
-        self.assertEqual(self.p.latest_checkpoint_location(), '28163096')
+        self.assertEqual(self.p.latest_checkpoint_location(), 28163096)
         # 9.3 and 9.4 format
         mock_popen.return_value.communicate.side_effect = [
             (b'rmgr: XLOG        len (rec/tot):     72/   104, tx:          0, lsn: 0/01ADBC18, prev 0/01ADBBB8, ' +
@@ -353,14 +355,14 @@ class TestPostgresql(BaseTestPostgresql):
              b' 1; offset 0; oldest xid 715 in DB 1; oldest multi 1 in DB 1; oldest running xid 0; shutdown', None),
             (b'rmgr: Transaction len (rec/tot):     64/    96, tx:        726, lsn: 0/01ADBBB8, prev 0/01ADBB70, ' +
              b'bkp: 0000, desc: commit: 2021-02-26 11:19:37.900918 CET; inval msgs: catcache 11 catcache 10', None)]
-        self.assertEqual(self.p.latest_checkpoint_location(), '28163096')
+        self.assertEqual(self.p.latest_checkpoint_location(), 28163096)
         mock_popen.return_value.communicate.side_effect = [
             (b'rmgr: XLOG        len (rec/tot):     72/   104, tx:          0, lsn: 0/01ADBC18, prev 0/01ADBBB8, ' +
              b'bkp: 0000, desc: checkpoint: redo 0/1ADBC18; tli 1; prev tli 1; fpw true; xid 0/727; oid 16386; multi' +
              b' 1; offset 0; oldest xid 715 in DB 1; oldest multi 1 in DB 1; oldest running xid 0; shutdown', None),
             (b'rmgr: XLOG        len (rec/tot):      0/    32, tx:          0, lsn: 0/01ADBBB8, prev 0/01ADBBA0, ' +
              b'bkp: 0000, desc: xlog switch ', None)]
-        self.assertEqual(self.p.latest_checkpoint_location(), '28163000')
+        self.assertEqual(self.p.latest_checkpoint_location(), 28163000)
         # 9.5+ format
         mock_popen.return_value.communicate.side_effect = [
             (b'rmgr: XLOG        len (rec/tot):    114/   114, tx:          0, lsn: 0/01ADBC18, prev 0/018260F8, ' +
@@ -369,7 +371,7 @@ class TestPostgresql(BaseTestPostgresql):
              b' oldest running xid 0; shutdown', None),
             (b'rmgr: XLOG        len (rec/tot):     24/    24, tx:          0, lsn: 0/018260F8, prev 0/01826080, ' +
              b'desc: SWITCH ', None)]
-        self.assertEqual(self.p.latest_checkpoint_location(), '25321720')
+        self.assertEqual(self.p.latest_checkpoint_location(), 25321720)
 
     def test_reload(self):
         self.assertTrue(self.p.reload())
@@ -452,24 +454,20 @@ class TestPostgresql(BaseTestPostgresql):
     def test_get_postgres_role_from_data_directory(self):
         self.assertEqual(self.p.get_postgres_role_from_data_directory(), 'replica')
 
+    @patch('os.remove', Mock())
+    @patch('shutil.rmtree', Mock())
+    @patch('os.unlink', Mock(side_effect=OSError))
+    @patch('os.path.isdir', Mock(return_value=True))
+    @patch('os.path.exists', Mock(return_value=True))
     def test_remove_data_directory(self):
-        def _symlink(src, dst):
-            if os.name != 'nt':  # os.symlink under Windows needs admin rights skip it
-                os.symlink(src, dst)
-
-        os.makedirs(os.path.join(self.p.data_dir, 'foo'))
-        _symlink('foo', os.path.join(self.p.data_dir, 'pg_wal'))
-        os.makedirs(os.path.join(self.p.data_dir, 'foo_tsp'))
-        pg_tblspc = os.path.join(self.p.data_dir, 'pg_tblspc')
-        os.makedirs(pg_tblspc)
-        _symlink('../foo_tsp', os.path.join(pg_tblspc, '12345'))
-        self.p.remove_data_directory()
-        open(self.p.data_dir, 'w').close()
-        self.p.remove_data_directory()
-        _symlink('unexisting', self.p.data_dir)
-        with patch('os.unlink', Mock(side_effect=OSError)):
+        with patch('os.path.islink', Mock(return_value=True)):
+            self.p.remove_data_directory()
+        with patch('os.path.isfile', Mock(return_value=True)):
+            self.p.remove_data_directory()
+        with patch('os.path.islink', Mock(side_effect=[False, False, True, True])),\
+                patch('os.listdir', Mock(return_value=['12345'])),\
+                patch('os.path.realpath', Mock(side_effect=['../foo', '../foo_tsp'])):
             self.p.remove_data_directory()
-        self.p.remove_data_directory()
 
     @patch('patroni.postgresql.Postgresql._version_file_exists', Mock(return_value=True))
     def test_controldata(self):

tests/test_postmaster.py

@@ -133,14 +133,20 @@ class TestPostmasterProcess(unittest.TestCase):
         c2.cmdline = Mock(return_value=["postgres: postgres postgres [local] idle"])
         c3 = Mock()
         c3.cmdline = Mock(side_effect=psutil.NoSuchProcess(123))
+        mock_wait.return_value = ([], [c2])
         with patch('psutil.Process.children', Mock(return_value=[c1, c2, c3])):
             proc = PostmasterProcess(123)
-            self.assertIsNone(proc.wait_for_user_backends_to_close())
-            mock_wait.assert_called_with([c2])
+            self.assertIsNone(proc.wait_for_user_backends_to_close(1))
+            mock_wait.assert_called_with([c2], 1)
+
+        mock_wait.return_value = ([c2], [])
+        with patch('psutil.Process.children', Mock(return_value=[c1, c2, c3])):
+            proc = PostmasterProcess(123)
+            proc.wait_for_user_backends_to_close(1)
 
         with patch('psutil.Process.children', Mock(side_effect=psutil.NoSuchProcess(123))):
             proc = PostmasterProcess(123)
-            self.assertIsNone(proc.wait_for_user_backends_to_close())
+            self.assertIsNone(proc.wait_for_user_backends_to_close(None))
 
     @patch('subprocess.Popen')
     @patch('os.setsid', Mock(), create=True)

tests/test_rewind.py

@@ -66,7 +66,7 @@ class TestRewind(BaseTestPostgresql):
 
     def test_pg_rewind(self):
         r = {'user': '', 'host': '', 'port': '', 'database': '', 'password': ''}
-        with patch.object(Postgresql, 'major_version', PropertyMock(return_value=130000)),\
+        with patch.object(Postgresql, 'major_version', PropertyMock(return_value=150000)),\
                 patch.object(CancellableSubprocess, 'call', Mock(return_value=None)):
             with patch('subprocess.check_output', Mock(return_value=b'boo')):
                 self.assertFalse(self.r.pg_rewind(r))
@@ -218,6 +218,62 @@ class TestRewind(BaseTestPostgresql):
         self.r.cleanup_archive_status()
         self.r.cleanup_archive_status()
 
+    @patch('os.path.isfile', Mock(return_value=True))
+    @patch('shutil.move', Mock(side_effect=OSError))
+    @patch('patroni.postgresql.rewind.logger.info')
+    def test_archive_ready_wals(self, mock_logger_info):
+        with patch('os.listdir', Mock(side_effect=OSError)), \
+              patch.object(Postgresql, 'get_guc_value', Mock(side_effect=['on', 'command %f'])):
+            self.r._archive_ready_wals()
+            mock_logger_info.assert_not_called()
+
+        # each assert_not_called() calls get_guc_value('archive_mode') + get_guc_value('archive_command')
+        get_guc_value_res = [
+            '', 'command %f',
+            'on', '',
+        ]
+        with patch.object(Postgresql, 'get_guc_value', Mock(side_effect=get_guc_value_res)):
+            for _ in range(len(get_guc_value_res)//2):
+                self.r._archive_ready_wals()
+                mock_logger_info.assert_not_called()
+
+        with patch('os.listdir', Mock(return_value=['000000000000000000000000.ready'])):
+            # successful archive_command call
+            with patch.object(CancellableSubprocess, 'call',  Mock(return_value=0)):
+                get_guc_value_res = [
+                    'on', 'command %f',
+                    'always', 'command %f',
+                ]
+                with patch.object(Postgresql, 'get_guc_value', Mock(side_effect=get_guc_value_res)):
+                    for _ in range(len(get_guc_value_res)//2):
+                        self.r._archive_ready_wals()
+                        mock_logger_info.assert_called_once()
+                        self.assertEqual(('Trying to archive %s: %s',
+                                          '000000000000000000000000', 'command 000000000000000000000000'),
+                                         mock_logger_info.call_args[0])
+                        mock_logger_info.reset_mock()
+
+            # failed archive_command call
+            with patch.object(CancellableSubprocess, 'call', Mock(return_value=1)):
+                with patch.object(Postgresql, 'get_guc_value', Mock(side_effect=['on', 'command %f'])):
+                    self.r._archive_ready_wals()
+                    self.assertEqual(('Trying to archive %s: %s',
+                                      '000000000000000000000000', 'command 000000000000000000000000'),
+                                     mock_logger_info.call_args_list[0][0])
+                    self.assertEqual(('Failed to archive WAL segment %s', '000000000000000000000000'),
+                                     mock_logger_info.call_args_list[1][0])
+                    mock_logger_info.reset_mock()
+
+        wal_files_to_skip = [
+            '000000000000000000000000.done',
+            '000000000000000000000001.partial.done',
+            '002.ready',
+            'U00000000000000000000001.ready',
+        ]
+        with patch('os.listdir', Mock(return_value=wal_files_to_skip)):
+            self.r._archive_ready_wals()
+            mock_logger_info.assert_not_called()
+
     @patch('os.unlink', Mock())
     @patch('os.listdir', Mock(return_value=[]))
     @patch('os.path.isfile', Mock(return_value=True))

tests/test_slots.py

@@ -4,17 +4,19 @@ import unittest
 
 
 from mock import Mock, PropertyMock, patch
+from threading import Thread
 
 from patroni import psycopg
 from patroni.dcs import Cluster, ClusterConfig, Member
 from patroni.postgresql import Postgresql
-from patroni.postgresql.slots import SlotsHandler, fsync_dir
+from patroni.postgresql.slots import SlotsAdvanceThread, SlotsHandler, fsync_dir
 
 from . import BaseTestPostgresql, psycopg_connect, MockCursor
 
 
 @patch('subprocess.call', Mock(return_value=0))
 @patch('patroni.psycopg.connect', psycopg_connect)
+@patch.object(Thread, 'start', Mock())
 @patch.object(Postgresql, 'is_running', Mock(return_value=True))
 class TestSlotsHandler(BaseTestPostgresql):
 
@@ -27,6 +29,9 @@ class TestSlotsHandler(BaseTestPostgresql):
         super(TestSlotsHandler, self).setUp()
         self.s = self.p.slots_handler
         self.p.start()
+        config = ClusterConfig(1, {'slots': {'ls': {'database': 'a', 'plugin': 'b'}}}, 1)
+        self.cluster = Cluster(True, config, self.leader, 0,
+                               [self.me, self.other, self.leadermem], None, None, None, {'ls': 12345})
 
     def test_sync_replication_slots(self):
         config = ClusterConfig(1, {'slots': {'test_3': {'database': 'a', 'plugin': 'b'},
@@ -39,8 +44,10 @@ class TestSlotsHandler(BaseTestPostgresql):
         self.p.set_role('standby_leader')
         self.s.sync_replication_slots(cluster, False)
         self.p.set_role('replica')
-        with patch.object(Postgresql, 'is_leader', Mock(return_value=False)):
-            self.s.sync_replication_slots(cluster, False)
+        with patch.object(Postgresql, 'is_leader', Mock(return_value=False)),\
+                patch.object(SlotsHandler, 'drop_replication_slot') as mock_drop:
+            self.s.sync_replication_slots(cluster, False, paused=True)
+            mock_drop.assert_not_called()
         self.p.set_role('master')
         with mock.patch('patroni.postgresql.Postgresql.role', new_callable=PropertyMock(return_value='replica')):
             self.s.sync_replication_slots(cluster, False)
@@ -81,41 +88,46 @@ class TestSlotsHandler(BaseTestPostgresql):
     @patch.object(Postgresql, 'is_leader', Mock(return_value=False))
     def test__ensure_logical_slots_replica(self):
         self.p.set_role('replica')
-        config = ClusterConfig(1, {'slots': {'ls': {'database': 'a', 'plugin': 'b'}}}, 1)
-        cluster = Cluster(True, config, self.leader, 0,
-                          [self.me, self.other, self.leadermem], None, None, None, {'ls': 12346})
-        self.assertEqual(self.s.sync_replication_slots(cluster, False), [])
+        self.cluster.slots['ls'] = 12346
+        with patch.object(SlotsHandler, 'check_logical_slots_readiness', Mock()):
+            self.assertEqual(self.s.sync_replication_slots(self.cluster, False), [])
         self.s._schedule_load_slots = False
         with patch.object(MockCursor, 'execute', Mock(side_effect=psycopg.OperationalError)),\
+                patch.object(SlotsAdvanceThread, 'schedule', Mock(return_value=(True, ['ls']))),\
                 patch.object(psycopg.OperationalError, 'diag') as mock_diag:
             type(mock_diag).sqlstate = PropertyMock(return_value='58P01')
-            self.assertEqual(self.s.sync_replication_slots(cluster, False), ['ls'])
-        cluster.slots['ls'] = 'a'
-        self.assertEqual(self.s.sync_replication_slots(cluster, False), [])
+            self.assertEqual(self.s.sync_replication_slots(self.cluster, False), ['ls'])
+        self.cluster.slots['ls'] = 'a'
+        self.assertEqual(self.s.sync_replication_slots(self.cluster, False), [])
         with patch.object(MockCursor, 'rowcount', PropertyMock(return_value=1), create=True):
-            self.assertEqual(self.s.sync_replication_slots(cluster, False), ['ls'])
+            self.assertEqual(self.s.sync_replication_slots(self.cluster, False), ['ls'])
 
-    @patch.object(MockCursor, 'execute', Mock(side_effect=psycopg.OperationalError))
     def test_copy_logical_slots(self):
-        self.s.copy_logical_slots(self.leader, ['foo'])
+        self.cluster.config.data['slots']['ls']['database'] = 'b'
+        self.s.copy_logical_slots(self.cluster, ['ls'])
+        with patch.object(MockCursor, 'execute', Mock(side_effect=psycopg.OperationalError)):
+            self.s.copy_logical_slots(self.cluster, ['foo'])
 
     @patch.object(Postgresql, 'stop', Mock(return_value=True))
     @patch.object(Postgresql, 'start', Mock(return_value=True))
     @patch.object(Postgresql, 'is_leader', Mock(return_value=False))
     def test_check_logical_slots_readiness(self):
-        self.s.copy_logical_slots(self.leader, ['ls'])
-        config = ClusterConfig(1, {'slots': {'ls': {'database': 'a', 'plugin': 'b'}}}, 1)
-        cluster = Cluster(True, config, self.leader, 0,
-                          [self.me, self.other, self.leadermem], None, None, None, {'ls': 12345})
-        self.assertEqual(self.s.sync_replication_slots(cluster, False), [])
-        with patch.object(MockCursor, 'rowcount', PropertyMock(return_value=1), create=True):
-            self.s.check_logical_slots_readiness(cluster, False, None)
+        self.s.copy_logical_slots(self.cluster, ['ls'])
+        with patch.object(MockCursor, '__iter__', Mock(return_value=iter([('postgresql0', None)]))),\
+                patch.object(MockCursor, 'fetchone', Mock(side_effect=Exception)):
+            self.assertIsNone(self.s.check_logical_slots_readiness(self.cluster, False, None))
+        with patch.object(MockCursor, '__iter__', Mock(return_value=iter([('postgresql0', None)]))),\
+                patch.object(MockCursor, 'fetchone', Mock(return_value=(False,))):
+            self.assertIsNone(self.s.check_logical_slots_readiness(self.cluster, False, None))
+        with patch.object(MockCursor, '__iter__', Mock(return_value=iter([('ls', 100)]))):
+            self.s.check_logical_slots_readiness(self.cluster, False, None)
 
     @patch.object(Postgresql, 'stop', Mock(return_value=True))
     @patch.object(Postgresql, 'start', Mock(return_value=True))
     @patch.object(Postgresql, 'is_leader', Mock(return_value=False))
     def test_on_promote(self):
-        self.s.copy_logical_slots(self.leader, ['ls'])
+        self.s.schedule_advance_slots({'foo': {'bar': 100}})
+        self.s.copy_logical_slots(self.cluster, ['ls'])
         self.s.on_promote()
 
     @unittest.skipIf(os.name == 'nt', "Windows not supported")
@@ -124,3 +136,18 @@ class TestSlotsHandler(BaseTestPostgresql):
     @patch('os.fsync', Mock(side_effect=OSError))
     def test_fsync_dir(self):
         self.assertRaises(OSError, fsync_dir, 'foo')
+
+    def test_slots_advance_thread(self):
+        with patch.object(MockCursor, 'execute', Mock(side_effect=psycopg.OperationalError)),\
+                patch.object(psycopg.OperationalError, 'diag') as mock_diag:
+            type(mock_diag).sqlstate = PropertyMock(return_value='58P01')
+            self.s.schedule_advance_slots({'foo': {'bar': 100}})
+            self.s._advance.sync_slots()
+
+        with patch.object(SlotsAdvanceThread, 'sync_slots', Mock(side_effect=Exception)):
+            self.s._advance._condition.wait = Mock()
+            self.assertRaises(Exception, self.s._advance.run)
+
+        with patch.object(SlotsHandler, 'get_local_connection_cursor', Mock(side_effect=Exception)):
+            self.s.schedule_advance_slots({'foo': {'bar': 100}})
+            self.s._advance.sync_slots()

tests/test_validator.py

@@ -63,6 +63,7 @@ config = {
     "postgresql": {
         "listen": "127.0.0.2,::1:543",
         "connect_address": "127.0.0.2:543",
+        "proxy_address": "127.0.0.2:5433",
         "authentication": {
             "replication": {"username": "user"},
             "superuser": {"username": "user"},
@@ -141,14 +142,14 @@ class TestValidator(unittest.TestCase):
         del directories[:]
 
     def test_empty_config(self, mock_out, mock_err):
-        schema({})
-        output = mock_out.getvalue()
+        errors = schema({})
+        output = "\n".join(errors)
         expected = list(sorted(['name', 'postgresql', 'restapi', 'scope'] + available_dcs))
         self.assertEqual(expected, parse_output(output))
 
     def test_complete_config(self, mock_out, mock_err):
-        schema(config)
-        output = mock_out.getvalue()
+        errors = schema(config)
+        output = "\n".join(errors)
         self.assertEqual(['postgresql.bin_dir', 'raft.bind_addr', 'raft.self_addr'], parse_output(output))
 
     def test_bin_dir_is_file(self, mock_out, mock_err):
@@ -156,10 +157,11 @@ class TestValidator(unittest.TestCase):
         files.append(config["postgresql"]["bin_dir"])
         c = copy.deepcopy(config)
         c["restapi"]["connect_address"] = 'False:blabla'
+        c["postgresql"]["listen"] = '*:543'
         c["etcd"]["hosts"] = ["127.0.0.1:2379", "1244.0.0.1:2379", "127.0.0.1:invalidport"]
         c["kubernetes"]["pod_ip"] = "127.0.0.1111"
-        schema(c)
-        output = mock_out.getvalue()
+        errors = schema(c)
+        output = "\n".join(errors)
         self.assertEqual(['etcd.hosts.1', 'etcd.hosts.2', 'kubernetes.pod_ip', 'postgresql.bin_dir',
                           'postgresql.data_dir', 'raft.bind_addr', 'raft.self_addr',
                           'restapi.connect_address'], parse_output(output))
@@ -176,8 +178,8 @@ class TestValidator(unittest.TestCase):
         c["etcd"]["host"] = "127.0.0.1:237"
         c["postgresql"]["listen"] = "127.0.0.1:5432"
         with patch('patroni.validator.open', mock_open(read_data='9')):
-            schema(c)
-        output = mock_out.getvalue()
+            errors = schema(c)
+        output = "\n".join(errors)
         self.assertEqual(['consul.host', 'etcd.host', 'postgresql.bin_dir', 'postgresql.data_dir', 'postgresql.listen',
                           'raft.bind_addr', 'raft.self_addr', 'restapi.connect_address'], parse_output(output))
 
@@ -195,8 +197,8 @@ class TestValidator(unittest.TestCase):
         files.append(os.path.join(config["postgresql"]["bin_dir"], "postgres"))
         files.append(os.path.join(config["postgresql"]["bin_dir"], "pg_isready"))
         with patch('patroni.validator.open', mock_open(read_data='12')):
-            schema(config)
-        output = mock_out.getvalue()
+            errors = schema(config)
+        output = "\n".join(errors)
         self.assertEqual(['raft.bind_addr', 'raft.self_addr'], parse_output(output))
 
     @patch('subprocess.check_output', Mock(return_value=b"postgres (PostgreSQL) 12.1"))
@@ -208,11 +210,12 @@ class TestValidator(unittest.TestCase):
         files.append(os.path.join(config["postgresql"]["data_dir"], "PG_VERSION"))
         c = copy.deepcopy(config)
         c["etcd"]["hosts"] = []
+        c["postgresql"]["listen"] = '127.0.0.2,*:543'
         del c["postgresql"]["bin_dir"]
         with patch('patroni.validator.open', mock_open(read_data='11')):
-            schema(c)
-        output = mock_out.getvalue()
-        self.assertEqual(['etcd.hosts', 'postgresql.data_dir',
+            errors = schema(c)
+        output = "\n".join(errors)
+        self.assertEqual(['etcd.hosts', 'postgresql.data_dir', 'postgresql.listen',
                           'raft.bind_addr', 'raft.self_addr'], parse_output(output))
 
     @patch('subprocess.check_output', Mock(return_value=b"postgres (PostgreSQL) 12.1"))
@@ -224,8 +227,8 @@ class TestValidator(unittest.TestCase):
         c = copy.deepcopy(config)
         del c["postgresql"]["bin_dir"]
         with patch('patroni.validator.open', mock_open(read_data='11')):
-            schema(c)
-        output = mock_out.getvalue()
+            errors = schema(c)
+        output = "\n".join(errors)
         self.assertEqual(['postgresql.data_dir', 'raft.bind_addr', 'raft.self_addr'], parse_output(output))
 
     def test_data_dir_is_empty_string(self, mock_out, mock_err):
@@ -236,7 +239,7 @@ class TestValidator(unittest.TestCase):
         c["postgresql"]["pg_hba"] = ""
         c["postgresql"]["data_dir"] = ""
         c["postgresql"]["bin_dir"] = ""
-        schema(c)
-        output = mock_out.getvalue()
+        errors = schema(c)
+        output = "\n".join(errors)
         self.assertEqual(['kubernetes', 'postgresql.bin_dir', 'postgresql.data_dir',
                           'postgresql.pg_hba', 'raft.bind_addr', 'raft.self_addr'], parse_output(output))

tests/test_watchdog.py

@@ -35,7 +35,7 @@ def mock_ioctl(fd, op, arg=None, mutate_flag=False):
     sys.stderr.write("Ioctl %d %d %r\n" % (fd, op, arg))
     if op == linuxwd.WDIOC_GETSUPPORT:
         sys.stderr.write("Get support\n")
-        assert(mutate_flag is True)
+        assert (mutate_flag is True)
         arg.options = sum(map(linuxwd.WDIOF.get, ['SETTIMEOUT', 'KEEPALIVEPING']))
         arg.identity = (ctypes.c_ubyte*32)(*map(ord, 'Mock Watchdog'))
     elif op == linuxwd.WDIOC_GETTIMEOUT:

tests/test_zookeeper.py

@@ -124,9 +124,11 @@ class TestPatroniSequentialThreadingHandler(unittest.TestCase):
         self.assertIsNotNone(self.handler.create_connection((), 40))
         self.assertIsNotNone(self.handler.create_connection(timeout=40))
 
-    @patch.object(SequentialThreadingHandler, 'select', Mock(side_effect=ValueError))
     def test_select(self):
-        self.assertRaises(select.error, self.handler.select)
+        with patch.object(SequentialThreadingHandler, 'select', Mock(side_effect=ValueError)):
+            self.assertRaises(select.error, self.handler.select)
+        with patch.object(SequentialThreadingHandler, 'select', Mock(side_effect=IOError)):
+            self.assertRaises(Exception, self.handler.select)
 
 
 class TestPatroniKazooClient(unittest.TestCase):