[DPE-10663] fix(patroni): handle unreachable cluster in get_member_status (#179)

* fix(patroni): handle unreachable cluster in get_member_status

get_member_status() called cluster_status() with no try/except, unlike
its siblings get_member_ip / get_primary / cluster_members which all
catch the RetryError cluster_status() raises when no Patroni endpoint
responds. The exception then propagated uncaught into callers.

During a PITR restore, Patroni is intentionally stopped while the
restore runs; the ops framework re-emits the deferred
database-peers-relation-changed event inside the restore action hook,
which calls _was_restore_successful() -> get_member_status() ->
cluster_status(), raising RetryError. That crashed the restore action
hook with a non-zero exit, discarding the staged restore-status result
and exhausting the integration test's 10x retry loop
(test_backups_pitr_{aws,gcp}).

Catch the RetryError and return "" — the value the method's docstring
already promises when the status cannot be retrieved — so callers see
a clean early-exit instead of a hook crash. This fixes every caller of
get_member_status, not just the restore path.

Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>

* chore: bump version to 16.3.4

The get_member_status RetryError fix is a behavioral change to the
library, so it needs its own release. 16.3.3 is already released, so
bump to 16.3.4 for the fix to ship under a new version.

Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>

---------

Signed-off-by: Marcelo Henrique Neppel <marcelo.neppel@canonical.com>

Author: marceloneppel

pyproject.toml

@@ -4,7 +4,7 @@
 [project]
 name = "postgresql-charms-single-kernel"
 description = "Shared and reusable code for PostgreSQL-related charms"
-version = "16.3.3"
+version = "16.3.4"
 readme = "README.md"
 license = {file = "LICENSE"}
 authors = [

single_kernel_postgresql/managers/patroni.py

@@ -251,7 +251,11 @@ def get_member_status(self, member_name: str) -> str:
                 couldn't be retrieved yet.
         """
         # Request info from cluster endpoint (which returns all members of the cluster).
-        cluster_status = self.cluster_status()
+        try:
+            cluster_status = self.cluster_status()
+        except RetryError:
+            logger.debug("Unable to get member status. Cluster status unreachable")
+            return ""
         if cluster_status:
             for member in cluster_status:
                 if member["name"] == member_name:

tests/unit/test_patroni_manager.py

@@ -105,6 +105,46 @@ def test_get_member_ip(patroni):
         assert patroni.get_member_ip("postgresql-0") == "1.1.1.1"
 
 
+def test_get_member_status(patroni):
+    with (
+        patch(
+            "single_kernel_postgresql.managers.patroni.parallel_patroni_get_request",
+            return_value=None,
+        ) as _parallel_patroni_get_request,
+        patch(
+            "single_kernel_postgresql.core.peer_relation.PostgreSQLApplication.patroni_password",
+            new_callable=PropertyMock,
+            return_value="test-pass",
+        ),
+        patch(
+            "single_kernel_postgresql.core.peer_relation.PostgreSQLApplication.endpoints",
+            new_callable=PropertyMock,
+            return_value=["endpoint1", "endpoint2", "endpoint3"],
+        ),
+        patch(
+            "single_kernel_postgresql.core.peer_relation.PostgreSQLApplication.members_ips",
+            new_callable=PropertyMock,
+            return_value=["ip1", "ip2", "ip3"],
+        ),
+        patch(
+            "single_kernel_postgresql.core.state.CharmState.endpoint",
+            new_callable=PropertyMock,
+            return_value="endpoint",
+        ),
+    ):
+        # An unreachable cluster (cluster_status raises RetryError) yields an empty
+        # status instead of propagating the exception.
+        assert patroni.get_member_status("postgresql-0") == ""
+
+        _parallel_patroni_get_request.return_value = {
+            "members": [
+                {"name": "postgresql-1", "state": "running"},
+                {"name": "postgresql-0", "state": "streaming"},
+            ]
+        }
+        assert patroni.get_member_status("postgresql-0") == "streaming"
+
+
 def test_get_patroni_health(patroni):
     with (
         patch(