[DPE-3684] reinit action

actions.yaml

@@ -33,8 +33,13 @@ list-backups:
 pre-upgrade-check:
   description: Run necessary pre-upgrade checks and preparations before executing a charm refresh.
 promote-to-primary:
-  description: Promotes the cluster of choice to a primary cluster. Must be ran against the leader unit.
+  description: Promotes the cluster of choice to a primary cluster. Must be ran against the leader unit when promoting a cluster
+   or against the unit to be promoted within the cluster.
   params:
+    scope:
+      type: string
+      default: cluster
+      description: Whether to promote a unit or a cluster. Must be set to either unit or cluster.
     force:
       type: boolean
       description: Force the promotion of a cluster when there is already a primary cluster.

src/charm.py

@@ -185,6 +185,7 @@ def __init__(self, *args):
         self.framework.observe(self.on.start, self._on_start)
         self.framework.observe(self.on.get_password_action, self._on_get_password)
         self.framework.observe(self.on.set_password_action, self._on_set_password)
+        self.framework.observe(self.on.promote_to_primary_action, self._on_promote_to_primary)
         self.framework.observe(self.on.update_status, self._on_update_status)
         self.cluster_name = self.app.name
         self._member_name = self.unit.name.replace("/", "-")
@@ -626,6 +627,7 @@ def _raft_reinitialisation(self) -> None:
                 and "raft_primary" not in self.unit_peer_data
                 and "raft_followers_stopped" in self.app_peer_data
             ):
+                self.unit.status = MaintenanceStatus("Reinitialising raft")
                 logger.info(f"Reinitialising {self.unit.name} as primary")
                 self._patroni.reinitialise_raft_data()
                 self.unit_peer_data["raft_primary"] = "True"
@@ -1502,6 +1504,27 @@ def _on_set_password(self, event: ActionEvent) -> None:
 
         event.set_results({"password": password})
 
+    def _on_promote_to_primary(self, event: ActionEvent) -> None:
+        if event.params.get("scope") == "cluster":
+            return self.async_replication.promote_to_primary(event)
+        elif event.params.get("scope") == "unit":
+            return self.promote_primary_unit(event)
+        else:
+            event.fail("Scope should be either cluster or unit")
+
+    def promote_primary_unit(self, event: ActionEvent) -> None:
+        """Handles promote to primary for unit scope."""
+        if event.params.get("force"):
+            if self.has_raft_keys():
+                self.unit_peer_data.update({"raft_candidate": "True"})
+                if self.unit.is_leader():
+                    self._raft_reinitialisation()
+                return
+            event.fail("Raft is not stuck.")
+        else:
+            # TODO Regular promotion
+            pass
+
     def _on_update_status(self, _) -> None:
         """Update the unit status message and users list in the database."""
         if not self._can_run_on_update_status():

src/cluster.py

@@ -901,17 +901,9 @@ def remove_raft_member(self, member_ip: str) -> None:
         if not raft_status["has_quorum"] and (
             not raft_status["leader"] or raft_status["leader"].host == member_ip
         ):
-            self.charm.unit.status = MaintenanceStatus("Reinitialising raft")
+            self.charm.unit.status = MaintenanceStatus("Raft majority loss")
             logger.warning("Remove raft member: Stuck raft cluster detected")
             data_flags = {"raft_stuck": "True"}
-            try:
-                candidate = self._get_role() in ["leader", "sync_standby"]
-            except Exception:
-                logger.warning("Remove raft member: Unable to get cluster status")
-                candidate = False
-            if candidate:
-                logger.info(f"{self.charm.unit.name} is raft candidate")
-                data_flags["raft_candidate"] = "True"
             self.charm.unit_peer_data.update(data_flags)
 
             # Leader doesn't always trigger when changing it's own peer data.

src/relations/async_replication.py

@@ -106,9 +106,6 @@ def __init__(self, charm):
         self.framework.observe(
             self.charm.on.create_replication_action, self._on_create_replication
         )
-        self.framework.observe(
-            self.charm.on.promote_to_primary_action, self._on_promote_to_primary
-        )
 
         self.framework.observe(self.charm.on.secret_changed, self._on_secret_changed)
 
@@ -583,7 +580,7 @@ def _on_create_replication(self, event: ActionEvent) -> None:
         # Set the status.
         self.charm.unit.status = MaintenanceStatus("Creating replication...")
 
-    def _on_promote_to_primary(self, event: ActionEvent) -> None:
+    def promote_to_primary(self, event: ActionEvent) -> None:
         """Promote this cluster to the primary cluster."""
         if (
             self.charm.app.status.message != READ_ONLY_MODE_BLOCKING_MESSAGE

tests/unit/test_cluster.py

@@ -783,7 +783,7 @@ def test_remove_raft_member_no_quorum(patroni, harness):
 
         patroni.remove_raft_member("1.2.3.4")
 
-        assert harness.charm.unit_peer_data == {"raft_candidate": "True", "raft_stuck": "True"}
+        assert harness.charm.unit_peer_data == {"raft_stuck": "True"}
 
 
 def test_remove_raft_data(patroni):