Abort startup if nodename conflict is detected

We do run update_available_resource() synchronously during service
startup, but we only allow certain exceptions to abort startup. This
makes us abort for InvalidConfiguration, and makes the resource
tracker raise that for the case where the compute node create failed
due to a duplicate entry.

This also modifies the object to raise a nova-specific error for that
condition to avoid the compute node needing to import oslo_db stuff
just to be able to catch it.

Change-Id: I5de98e6fe52e45996bc2e1014fa8a09a2de53682

Author: kk7ds

nova/compute/manager.py

@@ -10480,6 +10480,14 @@ def _update_available_resource_for_node(self, context, nodename,
             LOG.exception(
                 "Error updating PCI resources for node %(node)s.",
                 {'node': nodename})
+        except exception.InvalidConfiguration as e:
+            if startup:
+                # If this happens during startup, we need to let it raise to
+                # abort our service startup.
+                raise
+            else:
+                LOG.error("Error updating resources for node %s: %s",
+                          nodename, e)
         except Exception:
             LOG.exception("Error updating resources for node %(node)s.",
                           {'node': nodename})

nova/compute/resource_tracker.py

@@ -728,7 +728,13 @@ def _init_compute_node(self, context, resources):
         cn = objects.ComputeNode(context)
         cn.host = self.host
         self._copy_resources(cn, resources, initial=True)
-        cn.create()
+        try:
+            cn.create()
+        except exception.DuplicateRecord:
+            raise exception.InvalidConfiguration(
+                'Duplicate compute node record found for host %s node %s' % (
+                cn.host, cn.hypervisor_hostname))
+
         # Only map the ComputeNode into compute_nodes if create() was OK
         # because if create() fails, on the next run through here nodename
         # would be in compute_nodes and we won't try to create again (because

nova/exception.py

@@ -2512,6 +2512,10 @@ class InvalidNodeConfiguration(NovaException):
     msg_fmt = _('Invalid node identity configuration: %(reason)s')
 
 
+class DuplicateRecord(NovaException):
+    msg_fmt = _('Unable to create duplicate record for %(target)s')
+
+
 class NotSupportedComputeForEvacuateV295(NotSupported):
     msg_fmt = _("Starting to microversion 2.95, evacuate API will stop "
                 "instance on destination. To evacuate before upgrades are "

nova/objects/compute_node.py

@@ -12,6 +12,7 @@
 #    License for the specific language governing permissions and limitations
 #    under the License.
 
+from oslo_db import exception as db_exc
 from oslo_serialization import jsonutils
 from oslo_utils import uuidutils
 from oslo_utils import versionutils
@@ -339,7 +340,12 @@ def create(self):
         self._convert_supported_instances_to_db_format(updates)
         self._convert_pci_stats_to_db_format(updates)
 
-        db_compute = db.compute_node_create(self._context, updates)
+        try:
+            db_compute = db.compute_node_create(self._context, updates)
+        except db_exc.DBDuplicateEntry:
+            target = 'compute node %s:%s' % (updates['hypervisor_hostname'],
+                                             updates['uuid'])
+            raise exception.DuplicateRecord(target=target)
         self._from_db_object(self._context, self, db_compute)
 
     @base.remotable

nova/tests/unit/compute/test_resource_tracker.py

@@ -1552,6 +1552,20 @@ def test_undelete_node_move_host(self, mock_pci, mock_update):
         self.assertEqual('fake-host', node.host)
         mock_update.assert_called()
 
+    @mock.patch.object(resource_tracker.ResourceTracker,
+                       '_get_compute_node',
+                       return_value=None)
+    @mock.patch('nova.objects.compute_node.ComputeNode.create')
+    def test_create_failed_conflict(self, mock_create, mock_getcn):
+        self._setup_rt()
+        resources = {'hypervisor_hostname': 'node1',
+                     'uuid': uuids.node1}
+        mock_create.side_effect = exc.DuplicateRecord(target='foo')
+        self.assertRaises(exc.InvalidConfiguration,
+                          self.rt._init_compute_node,
+                          mock.MagicMock,
+                          resources)
+
 
 @ddt.ddt
 class TestUpdateComputeNode(BaseTestCase):

nova/tests/unit/objects/test_compute_node.py

@@ -16,6 +16,7 @@
 from unittest import mock
 
 import netaddr
+from oslo_db import exception as db_exc
 from oslo_serialization import jsonutils
 from oslo_utils.fixture import uuidsentinel
 from oslo_utils import timeutils
@@ -341,6 +342,14 @@ def test_recreate_fails(self, mock_get, mock_create):
                       'uuid': uuidsentinel.fake_compute_node}
         mock_create.assert_called_once_with(self.context, param_dict)
 
+    @mock.patch('nova.db.main.api.compute_node_create')
+    def test_create_duplicate(self, mock_create):
+        mock_create.side_effect = db_exc.DBDuplicateEntry
+        compute = compute_node.ComputeNode(context=self.context)
+        compute.service_id = 456
+        compute.hypervisor_hostname = 'node1'
+        self.assertRaises(exception.DuplicateRecord, compute.create)
+
     @mock.patch.object(db, 'compute_node_update')
     @mock.patch(
         'nova.db.main.api.compute_node_get', return_value=fake_compute_node)