Fix branch used to trigger release to 5/edge (#221)

* Split cluster initialization into two patrs:

1. Cluster initialization
2. User creation

* Fix lint issue

* Fix lint issue

* Fix unit tests

* Fix branch used to trigger release to 5/edge

* Update CI build scripts to use 'build withoug cache' workflow

* Free disk space

* Update src/charm.py

Updated doc string for `_initialise_users` function

Co-authored-by: Mia Altieri <[email protected]>

* Fix lint issues

---------

Co-authored-by: Mia Altieri <[email protected]>

.github/workflows/ci.yaml

@@ -41,10 +41,16 @@ jobs:
         run: tox run -e unit
 
   build:
-    name: Build charms
-    uses: canonical/data-platform-workflows/.github/workflows/build_charms_with_cache.yaml@v5
+    strategy:
+      fail-fast: true
+      matrix:
+        charms: [".", "tests/integration/ha_tests/application_charm", "tests/integration/relation_tests/application-charm"]
+    name: Build ${{matrix.charms}} charm
+    uses: canonical/data-platform-workflows/.github/workflows/build_charm_without_cache.yaml@v5
     with:
       charmcraft-snap-channel: "latest/edge"
+      path-to-charm-directory: ${{matrix.charms}}
+
 
   integration-test:
     strategy:
@@ -78,6 +84,17 @@ jobs:
         uses: actions/download-artifact@v3
         with:
           name: ${{ needs.build.outputs.artifact-name }}
+      - name: Free disk space
+        run: |
+          echo "Free disk space before cleanup"
+          df -T
+          # free space in the runner
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /usr/local/share/boost
+          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
+          echo "Free disk space after cleanup"
+          df -T
       - name: Select tests
         id: select-tests
         run: |

.github/workflows/release_5_edge.yaml

@@ -3,7 +3,7 @@ name: Release to 5/edge
 on:
   push:
     branches:
-      - main
+      - 5/edge
 
 jobs:
   lib-check:

src/charm.py

@@ -271,6 +271,21 @@ def db_initialised(self, value):
                 f"'db_initialised' must be a boolean value. Proivded: {value} is of type {type(value)}"
             )
 
+    @property
+    def users_initialized(self) -> bool:
+        """Check if MongoDB users are created."""
+        return "users_initialized" in self.app_peer_data
+
+    @users_initialized.setter
+    def users_initialized(self, value):
+        """Set the users_initialized flag."""
+        if isinstance(value, bool):
+            self.app_peer_data["users_initialized"] = str(value)
+        else:
+            raise ValueError(
+                f"'users_initialized' must be a boolean value. Proivded: {value} is of type {type(value)}"
+            )
+
     # END: properties
 
     # BEGIN: generic helper methods
@@ -387,6 +402,7 @@ def _on_start(self, event) -> None:
             return
 
         self._initialise_replica_set(event)
+        self._initialise_users(event)
 
         # mongod is now active
         self.unit.status = ActiveStatus()
@@ -590,6 +606,48 @@ def _on_secret_changed(self, event):
     # END: actions
 
     # BEGIN: user management
+    @retry(
+        stop=stop_after_attempt(3),
+        wait=wait_fixed(5),
+        reraise=True,
+        before=before_log(logger, logging.DEBUG),
+    )
+    def _initialise_users(self, event: StartEvent) -> None:
+        """Create users.
+
+        User creation can only be completed after the replica set has
+        been initialised which requires some time.
+        In race conditions this can lead to failure to initialise users.
+        To prevent these race conditions from breaking the code, retry on failure.
+        """
+        if not self.db_initialised:
+            return
+
+        if self.users_initialized:
+            return
+
+        # only leader should create users
+        if not self.unit.is_leader():
+            return
+
+        logger.info("User initialization")
+
+        try:
+            self._init_operator_user()
+            self._init_backup_user()
+            self._init_monitor_user()
+            logger.info("Reconcile relations")
+            self.client_relations.oversee_users(None, event)
+            self.users_initialized = True
+        except ExecError as e:
+            logger.error("Deferring on_start: exit code: %i, stderr: %s", e.exit_code, e.stderr)
+            event.defer()
+            return
+        except PyMongoError as e:
+            logger.error("Deferring on_start since: error=%r", e)
+            event.defer()
+            return
+
     @retry(
         stop=stop_after_attempt(3),
         wait=wait_fixed(5),
@@ -757,12 +815,6 @@ def _initialise_replica_set(self, event: StartEvent) -> None:
             try:
                 logger.info("Replica Set initialization")
                 direct_mongo.init_replset()
-                logger.info("User initialization")
-                self._init_operator_user()
-                self._init_backup_user()
-                self._init_monitor_user()
-                logger.info("Reconcile relations")
-                self.client_relations.oversee_users(None, event)
             except ExecError as e:
                 logger.error(
                     "Deferring on_start: exit code: %i, stderr: %s", e.exit_code, e.stderr

tests/unit/test_charm.py

@@ -295,6 +295,7 @@ def test_start_already_initialised(self, connection, init_user, provider, defer)
         self.harness.charm.unit.get_container = mock_container
 
         self.harness.charm.app_peer_data["db_initialised"] = "True"
+        self.harness.charm.app_peer_data["users_initialized"] = "True"
 
         self.harness.charm.on.start.emit()
 
@@ -394,7 +395,7 @@ def test_start_mongod_error_initalising_user(self, connection, init_user, provid
         defer.assert_called()
 
         # verify app data
-        self.assertEqual("db_initialised" in self.harness.charm.app_peer_data, False)
+        self.assertEqual("users_initialized" in self.harness.charm.app_peer_data, False)
 
     @patch("ops.framework.EventBase.defer")
     @patch("charm.MongoDBProvider")
@@ -423,7 +424,7 @@ def test_start_mongod_error_overseeing_users(self, connection, init_user, provid
             defer.assert_called()
 
             # verify app data
-            self.assertEqual("db_initialised" in self.harness.charm.app_peer_data, False)
+            self.assertEqual("users_initialized" in self.harness.charm.app_peer_data, False)
 
     @patch("ops.framework.EventBase.defer")
     @patch("charm.MongoDBConnection")