[OVN] Cleanup old Hash Ring node entries

This patch introduces a maintenance task that runs once a day and is
responsible for cleaning up Hash Ring nodes that haven't been updated in
5 days or more.

Change-Id: Ibed9e0d77500570c3d0f9f39bfe40cb9239d0d7a
Closes-Bug: #2033281
Signed-off-by: Lucas Alvares Gomes <lucasagomes@gmail.com>
(cherry picked from commit 7f777c223e)

neutron/db/ovn_hash_ring_db.py

@@ -60,6 +60,15 @@ def remove_node_by_uuid(context, node_uuid):
     LOG.info('Node "%s" removed from the Hash Ring', node_uuid)
 
 
+@db_api.retry_if_session_inactive()
+def cleanup_old_nodes(context, days):
+    age = timeutils.utcnow() - datetime.timedelta(days=days)
+    with db_api.CONTEXT_WRITER.using(context):
+        context.session.query(ovn_models.OVNHashRing).filter(
+            ovn_models.OVNHashRing.updated_at < age).delete()
+    LOG.info('Cleaned up Hash Ring nodes older than %d days', days)
+
+
 @db_api.retry_if_session_inactive()
 def _touch(context, updated_at=None, **filter_args):
     if updated_at is None:

neutron/plugins/ml2/drivers/ovn/mech_driver/ovsdb/maintenance.py

@@ -1070,6 +1070,20 @@ class DBInconsistenciesPeriodics(SchemaAwarePeriodicsBase):
                     for table in ('Chassis_Private', 'Chassis'):
                         txn.add(self._sb_idl.db_destroy(table, ch.name))
 
+    @periodics.periodic(spacing=86400, run_immediately=True)
+    def cleanup_old_hash_ring_nodes(self):
+        """Daily task to cleanup old stable Hash Ring node entries.
+
+        Runs once a day and clean up Hash Ring entries that haven't
+        been updated in more than 5 days. See LP #2033281 for more
+        information.
+
+        """
+        if not self.has_lock:
+            return
+        context = n_context.get_admin_context()
+        hash_ring_db.cleanup_old_nodes(context, days=5)
+
 
 class HashRingHealthCheckPeriodics(object):
 

neutron/tests/unit/db/test_ovn_hash_ring_db.py

@@ -285,3 +285,29 @@ class TestHashRing(testlib_api.SqlTestCaseLight):
             self.admin_ctx, interval=60, group_name=HASH_RING_TEST_GROUP)
         self.assertEqual(2, len(active_nodes))
         self.assertNotIn(node_to_remove, [n.node_uuid for n in active_nodes])
+
+    def test_cleanup_old_nodes(self):
+        # Add 2 new nodes
+        self._add_nodes_and_assert_exists(count=2)
+
+        # Subtract 5 days from utcnow() and touch the nodes to make
+        # them to appear stale
+        fake_utcnow = timeutils.utcnow() - datetime.timedelta(days=5)
+        with mock.patch.object(timeutils, 'utcnow') as mock_utcnow:
+            mock_utcnow.return_value = fake_utcnow
+            ovn_hash_ring_db.touch_nodes_from_host(self.admin_ctx,
+                                                   HASH_RING_TEST_GROUP)
+
+        # Add 3 new nodes
+        self._add_nodes_and_assert_exists(count=3)
+
+        # Assert we have 5 nodes in the hash ring
+        self.assertEqual(5, ovn_hash_ring_db.count_nodes_from_host(
+            self.admin_ctx, HASH_RING_TEST_GROUP))
+
+        # Clean up the 2 stale nodes
+        ovn_hash_ring_db.cleanup_old_nodes(self.admin_ctx, days=5)
+
+        # Assert we only have 3 node entries after the clean up
+        self.assertEqual(3, ovn_hash_ring_db.count_nodes_from_host(
+            self.admin_ctx, HASH_RING_TEST_GROUP))

releasenotes/notes/hash-ring-cleanup-1079d2375082cebe.yaml

@@ -0,0 +1,6 @@
+---
+other:
+  - |
+    Adds a maintenance task that runs once a day and is responsible for
+    cleaning up Hash Ring nodes that haven't been updated in 5 days or
+    more. See LP #2033281 for more information.