Clean up nodes in DELETING on conductor restart

Change-Id: Iad6cb02ee4f17b7c5987719b3a0603df210c840d
(cherry picked from commit 4f6374a209)

ironic/common/states.py

@@ -235,7 +235,7 @@ UNSTABLE_STATES = (DEPLOYING, DEPLOYWAIT, CLEANING, CLEANWAIT, VERIFYING,
 """States that can be changed without external request."""
 
 STUCK_STATES_TREATED_AS_FAIL = (DEPLOYING, CLEANING, VERIFYING, INSPECTING,
-                                ADOPTING, RESCUING, UNRESCUING)
+                                ADOPTING, RESCUING, UNRESCUING, DELETING)
 """States that cannot be resumed once a conductor dies.
 
 If a node gets stuck with one of these states for some reason
@@ -383,7 +383,7 @@ machine.add_transition(DEPLOYWAIT, DELETING, 'delete')
 machine.add_transition(DEPLOYFAIL, DELETING, 'delete')
 
 # This state can also transition to error
-machine.add_transition(DELETING, ERROR, 'error')
+machine.add_transition(DELETING, ERROR, 'fail')
 
 # When finished deleting, a node will begin cleaning
 machine.add_transition(DELETING, CLEANING, 'clean')

ironic/conductor/manager.py

@@ -1033,7 +1033,7 @@ class ConductorManager(base_manager.BaseConductorManager):
                 LOG.exception('Error in tear_down of node %(node)s: %(err)s',
                               {'node': node.uuid, 'err': e})
                 node.last_error = _("Failed to tear down. Error: %s") % e
-                task.process_event('error')
+                task.process_event('fail')
         else:
             # NOTE(deva): When tear_down finishes, the deletion is done,
             # cleaning will start next

ironic/tests/unit/conductor/test_base_manager.py

@@ -217,6 +217,7 @@ class StartStopTestCase(mgr_utils.ServiceSetUpMixin, db_base.DbTestCase):
             (states.ADOPTING, states.ADOPTFAIL),
             (states.RESCUING, states.RESCUEFAIL),
             (states.UNRESCUING, states.UNRESCUEFAIL),
+            (states.DELETING, states.ERROR),
         ]
         nodes = [obj_utils.create_test_node(self.context, uuid=uuid.uuid4(),
                                             driver='fake-hardware',

ironic/tests/unit/conductor/test_manager.py

@@ -3434,7 +3434,7 @@ class DoNodeCleanTestCase(mgr_utils.ServiceSetUpMixin, db_base.DbTestCase):
     def test_continue_node_clean_wrong_state(self, mock_spawn):
         # Test the appropriate exception is raised if node isn't already
         # in CLEANWAIT state
-        prv_state = states.DELETING
+        prv_state = states.ACTIVE
         tgt_prv_state = states.AVAILABLE
         node = obj_utils.create_test_node(self.context, driver='fake-hardware',
                                           provision_state=prv_state,

releasenotes/notes/deleting-dcdb9cf0d2a6a1a6.yaml

@@ -0,0 +1,4 @@
+---
+fixes:
+  - |
+    Cleans up nodes stuck in the ``deleting`` state on conductor restart.