Support L3 agent cleanup on shutdown

Add an option to delete all routers on agent shutdown.

Closes-Bug: #1851609
Change-Id: I7a4056680d8453b2ef2dcc853437a0ec4b3e8044

neutron/agent/l3/agent.py

@@ -274,6 +274,7 @@ class L3NATAgent(ha.AgentMixin,
             self.conf = conf
         else:
             self.conf = cfg.CONF
+        self.check_config()
         self.router_info = {}
         self.router_factory = RouterFactory()
         self._register_router_cls(self.router_factory)
@@ -294,6 +295,7 @@ class L3NATAgent(ha.AgentMixin,
 
         self.fullsync = True
         self.sync_routers_chunk_size = SYNC_ROUTERS_MAX_CHUNK_SIZE
+        self._exiting = False
 
         # Get the HA router count from Neutron Server
         # This is the first place where we contact neutron-server on startup
@@ -344,6 +346,12 @@ class L3NATAgent(ha.AgentMixin,
 
         self._check_ha_router_process_status()
 
+    def check_config(self):
+        if self.conf.cleanup_on_shutdown:
+            LOG.warning("cleanup_on_shutdown is set to True, so L3 agent will "
+                        "cleanup all its routers when exiting, "
+                        "data-plane will be affected.")
+
     def _check_ha_router_process_status(self):
         """Check HA router VRRP process status in network node.
 
@@ -668,6 +676,9 @@ class L3NATAgent(ha.AgentMixin,
         self._queue.add(router_update)
 
     def _process_router_update(self):
+        if self._exiting:
+            return
+
         for rp, update in self._queue.each_update_to_next_resource():
             LOG.info("Starting router update for %s, action %s, priority %s, "
                      "update_id %s. Wait time elapsed: %.3f",
@@ -777,7 +788,7 @@ class L3NATAgent(ha.AgentMixin,
 
     def _process_routers_loop(self):
         LOG.debug("Starting _process_routers_loop")
-        while True:
+        while not self._exiting:
             self._pool.spawn_n(self._process_router_update)
 
     # NOTE(kevinbenton): this is set to 1 second because the actual interval
@@ -892,6 +903,13 @@ class L3NATAgent(ha.AgentMixin,
         eventlet.spawn_n(self._process_routers_loop)
         LOG.info("L3 agent started")
 
+    def stop(self):
+        LOG.info("Stopping L3 agent")
+        if self.conf.cleanup_on_shutdown:
+            self._exiting = True
+            for router in self.router_info.values():
+                router.delete()
+
     def create_pd_router_update(self):
         router_id = None
         update = queue.ResourceUpdate(router_id,

neutron/conf/agent/l3/config.py

@@ -100,6 +100,12 @@ OPTS = [
                       '(by default), the user executing the L3 agent will be '
                       'passed. If "root" specified, because radvd is spawned '
                       'as root, no "username" parameter will be passed.')),
+    cfg.BoolOpt('cleanup_on_shutdown', default=False,
+                help=_('Delete all routers on L3 agent shutdown. For L3 HA '
+                       'routers it includes a shutdown of keepalived and '
+                       'the state change monitor. NOTE: Setting to True '
+                       'could affect the data plane when stopping or '
+                       'restarting the L3 agent.')),
 ]
 
 

neutron/manager.py

@@ -72,6 +72,13 @@ class Manager(periodic_task.PeriodicTasks):
         """
         pass
 
+    def stop(self):
+        """Handle stop.
+
+        Child classes can override this method.
+        """
+        pass
+
 
 def validate_pre_plugin_load():
     """Checks if the configuration variables are valid.

neutron/service.py

@@ -427,6 +427,7 @@ class Service(n_rpc.Service):
             except Exception:
                 LOG.exception("Exception occurs when timer stops")
         self.timers = []
+        self.manager.stop()
 
     def wait(self):
         super(Service, self).wait()

neutron/tests/unit/agent/l3/test_agent.py

@@ -3982,3 +3982,20 @@ class TestBasicRouterOperations(BasicRouterOperationsFramework):
                 self.conf, get_networks_callback=mock.ANY)
         funct_partial_mock.assert_called_once_with(
             self.plugin_api.get_networks, agent.context)
+
+    def test_stop_no_cleanup(self):
+        agent = l3_agent.L3NATAgent(HOSTNAME, self.conf)
+        router = mock.Mock()
+        agent.router_info[1] = router
+        agent.stop()
+        self.assertFalse(router.delete.called)
+
+    def test_stop_cleanup(self):
+        self.conf.set_override('cleanup_on_shutdown', True)
+        agent = l3_agent.L3NATAgent(HOSTNAME, self.conf)
+        router = mock.Mock()
+        agent.router_info[1] = router
+        self.assertFalse(agent._exiting)
+        agent.stop()
+        self.assertTrue(router.delete.called)
+        self.assertTrue(agent._exiting)

releasenotes/notes/l3_agent_graceful_shutdown-87bf3304e6fab8a5.yaml

@@ -0,0 +1,10 @@
+---
+features:
+  - |
+    A new configuration option, ``cleanup_on_shutdown``, was added to the
+    L3 agent. If set to True the L3 agent will explicitly delete all routers
+    on shutdown. For L3 HA routers it includes a graceful shutdown of
+    keepalived and the state change monitor, which will allow a faster failover
+    in certain conditions. The default value of ``cleanup_on_shutdown`` is
+    False to maintain backward compatibility. Setting to True could affect
+    the data plane when stopping or restarting the L3 agent.