SIGHUP n-cpu to clear provider tree cache

An earlier change [1] allowed
[compute]resource_provider_association_refresh to be set to zero to
disable the resource tracker's periodic refresh of its local copy of
provider traits and aggregates. To allow for out-of-band changes to
placement (e.g. via the CLI) to be picked up by the resource tracker in
this configuration (or a configuration where the timer is set to a high
value) this change clears the provider tree cache when SIGHUP is sent to
the compute service. The next periodic will repopulate it afresh from
placement.

[1] Iec33e656491848b26686fbf6fb5db4a4c94b9ea8

Change-Id: I65a7ee565ca5b3ec6c33a2fd9e39d461f7d90ed2

nova/compute/manager.py

@@ -548,6 +548,7 @@ class ComputeManager(manager.Manager):
         LOG.info('Reloading compute RPC API')
         compute_rpcapi.LAST_VERSION = None
         self.compute_rpcapi = compute_rpcapi.ComputeAPI()
+        self._get_resource_tracker().reportclient.clear_provider_cache()
 
     def _get_resource_tracker(self):
         if not self._resource_tracker:

nova/conf/compute.py

@@ -652,6 +652,8 @@ Related options:
         default=300,
         min=0,
         mutable=True,
+        # TODO(efried): Provide more/better explanation of what this option is
+        # all about. Reference bug(s). Unless we're just going to remove it.
         help="""
 Interval for updating nova-compute-side cache of the compute node resource
 provider's aggregates and traits info.
@@ -662,6 +664,9 @@ node.
 
 A value of zero disables cache refresh completely.
 
+The cache can be cleared manually at any time by sending SIGHUP to the compute
+process, causing it to be repopulated the next time the data is accessed.
+
 Possible values:
 
 * Any positive integer in seconds, or zero to disable refresh.

nova/scheduler/client/report.py

@@ -274,6 +274,11 @@ class SchedulerReportClient(object):
         # NOTE(danms): Keep track of how naggy we've been
         self._warn_count = 0
 
+    def clear_provider_cache(self):
+        LOG.info("Clearing the report client's provider cache.")
+        self._provider_tree = provider_tree.ProviderTree()
+        self._association_refresh_time = {}
+
     @utils.synchronized(PLACEMENT_CLIENT_SEMAPHORE)
     def _create_client(self):
         """Create the HTTP session accessing the placement service."""

nova/tests/unit/compute/test_compute_mgr.py

@@ -4744,6 +4744,20 @@ class ComputeManagerUnitTestCase(test.NoDBTestCase):
             mock_rpc.assert_called_once_with()
             self.assertIsNot(orig_rpc, self.compute.compute_rpcapi)
 
+    def test_reset_clears_provider_cache(self):
+        # Seed the cache so we can tell we've cleared it
+        reportclient = self.compute._get_resource_tracker().reportclient
+        ptree = reportclient._provider_tree
+        ptree.new_root('foo', uuids.foo)
+        self.assertEqual([uuids.foo], ptree.get_provider_uuids())
+        times = reportclient._association_refresh_time
+        times[uuids.foo] = time.time()
+        self.compute.reset()
+        ptree = reportclient._provider_tree
+        self.assertEqual([], ptree.get_provider_uuids())
+        times = reportclient._association_refresh_time
+        self.assertEqual({}, times)
+
     @mock.patch('nova.objects.BlockDeviceMappingList.get_by_instance_uuid')
     @mock.patch('nova.compute.manager.ComputeManager._delete_instance')
     def test_terminate_instance_no_bdm_volume_id(self, mock_delete_instance,

releasenotes/notes/disable-rt-cache-refresh-9f6633e585516760.yaml

@@ -7,4 +7,8 @@ features:
     1767309`_ allowing more aggressive reduction in the amount of traffic to
     the placement service.
 
+    The cache can be cleared manually at any time by sending SIGHUP to the
+    compute process. This will cause the cache to be repopulated the next time
+    the data is accessed.
+
     .. _`bug 1767309`: https://bugs.launchpad.net/nova/+bug/1767309