[DPE-10454] Use better status instead of blocked for transient update failures (#1861)

Use maintenance status instead of blocked one.

The 'failed to update cluster members on member' BlockedStatus was
misleading: these are transient, self-recovering conditions (a RetryError
from update_config(), or a peer IP not yet published) rather than states
needing operator intervention. Replace them with MaintenanceStatus
and a 'updating cluster members' message.

Update test_on_peer_relation_changed accordingly, and set an explicit
BlockedStatus before exercising the early-exit guard so that path stays
covered.

(cherry picked from commit eddf95e)

Assisted-by: Claude:claude-5-fable

Author: taurus-forever

src/charm.py

@@ -749,7 +749,7 @@ def _on_peer_relation_changed(self, event: HookEvent):
             # Update the members of the cluster in the Patroni configuration on this unit.
             self.update_config()
         except RetryError:
-            self.unit.status = BlockedStatus("failed to update cluster members on member")
+            self.unit.status = MaintenanceStatus("cluster member update failed, retrying")
             return
         except ValueError as e:
             self.unit.status = BlockedStatus("Configuration Error. Please check the logs")
@@ -955,7 +955,7 @@ def add_cluster_member(self, member: str) -> None:
         try:
             self.update_config()
         except RetryError:
-            self.unit.status = BlockedStatus("failed to update cluster members on member")
+            self.unit.status = MaintenanceStatus("cluster member update failed, retrying")
 
     def _get_unit_ip(self, unit: Unit) -> str | None:
         """Get the IP address of a specific unit."""

tests/unit/test_charm.py

@@ -1445,11 +1445,12 @@ def test_on_peer_relation_changed(harness):
         _update_config.assert_called_once()
         _start_patroni.assert_not_called()
         _update_new_unit_status.assert_not_called()
-        assert isinstance(harness.model.unit.status, BlockedStatus)
+        assert isinstance(harness.model.unit.status, MaintenanceStatus)
 
         # Test event is early exiting when in blocked status.
         _update_config.side_effect = None
         _member_started.return_value = False
+        harness.model.unit.status = BlockedStatus()
         harness.charm._on_peer_relation_changed(mock_event)
         _start_patroni.assert_not_called()
 
@@ -1791,12 +1792,11 @@ def test_add_cluster_member(harness):
         _update_config.assert_called_once_with()
         _update_config.reset_mock()
 
-        # Charm blocks when update_config fails
+        # Charm goes into maintenance when update_config fails
         _update_config.side_effect = RetryError(last_attempt=None)
         harness.charm.add_cluster_member("postgresql/0")
         _update_config.assert_called_once_with()
-        assert isinstance(harness.charm.unit.status, BlockedStatus)
-        assert harness.charm.unit.status.message == "failed to update cluster members on member"
+        assert isinstance(harness.charm.unit.status, MaintenanceStatus)
         _update_config.reset_mock()
 
         # Not ready error if not all members are ready