Update node_is_replica to account for the quorum_standby state

check_patroni/cli.py

@@ -626,6 +626,13 @@ def node_is_leader(ctx: click.Context, check_standby_leader: bool) -> None:
     default=False,
     help="check if the replica is synchronous",
 )
+@click.option(
+    "--sync-type",
+    type=click.Choice(["any", "sync", "quorum"], case_sensitive=True),
+    default="any",
+    show_default=True,
+    help="Synchronous replication mode.",
+)
 @click.option(
     "--is-async",
     "check_is_async",
@@ -636,7 +643,11 @@ def node_is_leader(ctx: click.Context, check_standby_leader: bool) -> None:
 @click.pass_context
 @nagiosplugin.guarded
 def node_is_replica(
-    ctx: click.Context, max_lag: str, check_is_sync: bool, check_is_async: bool
+    ctx: click.Context,
+    max_lag: str,
+    check_is_sync: bool,
+    check_is_async: bool,
+    sync_type: str,
 ) -> None:
     """Check if the node is a replica with no noloadbalance tag.
 
@@ -675,9 +686,11 @@ def node_is_replica(
 
     check = nagiosplugin.Check()
     check.add(
-        NodeIsReplica(ctx.obj.connection_info, max_lag, check_is_sync, check_is_async),
+        NodeIsReplica(
+            ctx.obj.connection_info, max_lag, check_is_sync, check_is_async, sync_type
+        ),
         nagiosplugin.ScalarContext("is_replica", None, "@0:0"),
-        NodeIsReplicaSummary(max_lag, check_is_sync, check_is_async),
+        NodeIsReplicaSummary(max_lag, check_is_sync, check_is_async, sync_type),
     )
     check.main(verbose=ctx.obj.verbose, timeout=ctx.obj.timeout)
 

check_patroni/node.py

@@ -65,35 +65,53 @@ def __init__(
         max_lag: str,
         check_is_sync: bool,
         check_is_async: bool,
+        sync_type: str,
     ) -> None:
         super().__init__(connection_info)
         self.max_lag = max_lag
         self.check_is_sync = check_is_sync
         self.check_is_async = check_is_async
+        self.sync_type = sync_type
 
     def probe(self) -> Iterable[nagiosplugin.Metric]:
+        item_dict = {}
         try:
-            if self.check_is_sync:
-                api_name = "synchronous"
-            elif self.check_is_async:
-                api_name = "asynchronous"
-            else:
-                api_name = "replica"
-
             if self.max_lag is None:
-                self.rest_api(api_name)
+                item_dict = self.rest_api("replica")
             else:
-                self.rest_api(f"{api_name}?lag={self.max_lag}")
+                item_dict = self.rest_api(f"replica?lag={self.max_lag}")
         except APIError:
             return [nagiosplugin.Metric("is_replica", 0)]
-        return [nagiosplugin.Metric("is_replica", 1)]
+
+        if self.check_is_sync:
+            if (
+                self.sync_type in ["sync", "any"] and "sync_standby" in item_dict.keys()
+            ) or (
+                self.sync_type in ["quorum", "any"]
+                and "quorum_standby" in item_dict.keys()
+            ):
+                return [nagiosplugin.Metric("is_replica", 1)]
+            else:
+                return [nagiosplugin.Metric("is_replica", 0)]
+        elif self.check_is_async:
+            if (
+                "sync_standby" in item_dict.keys()
+                or "quorum_standby" in item_dict.keys()
+            ):
+                return [nagiosplugin.Metric("is_replica", 0)]
+            else:
+                return [nagiosplugin.Metric("is_replica", 1)]
+        else:
+            return [nagiosplugin.Metric("is_replica", 1)]
 
 
 class NodeIsReplicaSummary(nagiosplugin.Summary):
-    def __init__(self, lag: str, check_is_sync: bool, check_is_async: bool) -> None:
+    def __init__(
+        self, lag: str, check_is_sync: bool, check_is_async: bool, sync_type: str
+    ) -> None:
         self.lag = lag
         if check_is_sync:
-            self.replica_kind = "synchronous replica"
+            self.replica_kind = f"synchronous replica of kind '{sync_type}'"
         elif check_is_async:
             self.replica_kind = "asynchronous replica"
         else:

tests/json/node_is_replica_ok_quorum.json

@@ -0,0 +1,22 @@
+{
+  "state": "running",
+  "postmaster_start_time": "2024-12-23 10:07:07.660665+00:00",
+  "role": "replica",
+  "server_version": 140013,
+  "xlog": {
+    "received_location": 251660416,
+    "replayed_location": 251660416,
+    "replayed_timestamp": "2024-12-23 15:43:43.152572+00:00",
+    "paused": false
+  },
+  "quorum_standby": true,
+  "timeline": 9,
+  "replication_state": "streaming",
+  "dcs_last_seen": 1734972473,
+  "database_system_identifier": "7421168130564934130",
+  "patroni": {
+    "version": "4.0.2",
+    "scope": "patroni-demo",
+    "name": "p2"
+  }
+}

tests/json/node_is_replica_ok_sync.json

@@ -0,0 +1,22 @@
+{
+  "state": "running",
+  "postmaster_start_time": "2024-12-23 10:07:07.660665+00:00",
+  "role": "replica",
+  "server_version": 140013,
+  "xlog": {
+    "received_location": 251660416,
+    "replayed_location": 251660416,
+    "replayed_timestamp": "2024-12-23 15:43:43.152572+00:00",
+    "paused": false
+  },
+  "sync_standby": true,
+  "timeline": 9,
+  "replication_state": "streaming",
+  "dcs_last_seen": 1734972473,
+  "database_system_identifier": "7421168130564934130",
+  "patroni": {
+    "version": "4.0.2",
+    "scope": "patroni-demo",
+    "name": "p2"
+  }
+}