Update README & CHANGELOG for quorum commit

CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## unreleased
 
+### Added
+
+* Support for quorum synchronous réplication (#80)
+
 ### Fixed
 
 * Update the documentation to clarify that if patroni cannot be reached, we consider

README.md

@@ -212,14 +212,15 @@ Options:
 ```
 Usage: check_patroni cluster_has_replica [OPTIONS]
 
-  Check if the cluster has healthy replicas and/or if some are sync standbies
+  Check if the cluster has healthy replicas and/or if some are sync or quorum
+  standbies
 
   For patroni (and this check):
   * a replica is `streaming` if the `pg_stat_wal_receiver` say's so.
   * a replica is `in archive recovery`, if it's not `streaming` and has a `restore_command`.
 
   A healthy replica:
-  * has a `replica` or `sync_standby` role
+  * has a `replica`, `quorum_standby` or `sync_standby` role
   * has the same timeline as the leader and
     * is in `running` state (patroni < V3.0.4)
     * is in `streaming` or `in archive recovery` state (patroni >= V3.0.4)
@@ -236,27 +237,41 @@ Usage: check_patroni cluster_has_replica [OPTIONS]
   switchover or failover and the standbies are in the process of catching up
   with the new leader. The alert shouldn't last long.
 
+  In PostgreSQL, synchronous replication has two modes: on and quorum and is
+  configured with the gucs `synchronous_standby_names` and
+  `synchronous_commit`. Patroni uses the parameter `synchronous_mode`, which
+  can be set to `on`, `quorum` and `off`, and has `synchronous_node_count` to
+  configure the synchronous replication factor. Please note that, in
+  synchronous replication, the number of servers tagged as
+  "{sync|quorum}_standby" (what we measure) is not always equal tot
+  `synchronous_node_count`.
+
   Check:
   * `OK`: if the healthy_replica count and their lag are compatible with the replica count threshold.
-          and if the sync_replica count is compatible with the sync replica count threshold.
+          and if the synchronous replica count is compatible with the sync replica count threshold.
   * `WARNING` / `CRITICAL`: otherwise
 
   Perfdata:
   * healthy_replica & unhealthy_replica count
-  * the number of sync_replica, they are included in the previous count
+  * the number of sync_replica (sync or quorum depending on `--sync-type`), they are included
+    in the previous count
   * the lag of each replica labelled with "member name"_lag
   * the timeline of each replica labelled with "member name"_timeline
   * a boolean to tell if the node is a sync stanbdy labelled with "member name"_sync
 
 Options:
-  -w, --warning TEXT    Warning threshold for the number of healthy replica
-                        nodes.
-  -c, --critical TEXT   Critical threshold for the number of healthy replica
-                        nodes.
-  --sync-warning TEXT   Warning threshold for the number of sync replica.
-  --sync-critical TEXT  Critical threshold for the number of sync replica.
-  --max-lag TEXT        maximum allowed lag
-  --help                Show this message and exit.
+  -w, --warning TEXT             Warning threshold for the number of healthy
+                                 replica nodes.
+  -c, --critical TEXT            Critical threshold for the number of healthy
+                                 replica nodes.
+  --sync-warning TEXT            Warning threshold for the number of sync
+                                 replica.
+  --sync-critical TEXT           Critical threshold for the number of sync
+                                 replica.
+  --sync-type [any|sync|quorum]  Synchronous replication mode used to filter
+                                 and count sync standbies.  [default: any]
+  --max-lag TEXT                 maximum allowed lag
+  --help                         Show this message and exit.
 ```
 
 ### cluster_has_scheduled_action
@@ -310,6 +325,7 @@ Usage: check_patroni cluster_node_count [OPTIONS]
   * replica
   * standby_leader
   * sync_standby
+  * quorum_standby
   * demoted
   * promoted
   * uninitialized
@@ -327,7 +343,7 @@ Usage: check_patroni cluster_node_count [OPTIONS]
   The "healthy" checks only ensures that:
   * a leader has the running state
   * a standby_leader has the running or streaming (V3.0.4) state
-  * a replica or sync-standby has the running or streaming (V3.0.4) state
+  * a replica, quorum_standby or sync_standby has the running or streaming (V3.0.4) state
 
   Since we dont check the lag or timeline, "in archive recovery" is not
   considered a valid state for this service. See cluster_has_leader and
@@ -468,10 +484,11 @@ Usage: check_patroni node_is_replica [OPTIONS]
   noloadbalance tag and the lag is under the maximum threshold, 0 otherwise.
 
 Options:
-  --max-lag TEXT  maximum allowed lag
-  --is-sync       check if the replica is synchronous
-  --is-async      check if the replica is asynchronous
-  --help          Show this message and exit.
+  --max-lag TEXT                 maximum allowed lag
+  --is-sync                      check if the replica is synchronous
+  --sync-type [any|sync|quorum]  Synchronous replication mode.  [default: any]
+  --is-async                     check if the replica is asynchronous
+  --help                         Show this message and exit.
 ```
 
 ### node_patroni_version