diff options
Diffstat (limited to 'ironic/conductor')
-rw-r--r-- | ironic/conductor/manager.py | 6 | ||||
-rw-r--r-- | ironic/conductor/utils.py | 21 |
2 files changed, 25 insertions, 2 deletions
diff --git a/ironic/conductor/manager.py b/ironic/conductor/manager.py index d11224852..7f6470e38 100644 --- a/ironic/conductor/manager.py +++ b/ironic/conductor/manager.py @@ -1604,7 +1604,8 @@ class ConductorManager(base_manager.BaseConductorManager): :param context: request context. """ - offline_conductors = self.dbapi.get_offline_conductors() + offline_conductors = utils.exclude_current_conductor( + self.host, self.dbapi.get_offline_conductors()) if not offline_conductors: return @@ -3436,7 +3437,8 @@ class ConductorManager(base_manager.BaseConductorManager): :param context: request context. """ - offline_conductors = self.dbapi.get_offline_conductors(field='id') + offline_conductors = utils.exclude_current_conductor( + self.conductor.id, self.dbapi.get_offline_conductors(field='id')) for conductor_id in offline_conductors: filters = {'state': states.ALLOCATING, 'conductor_affinity': conductor_id} diff --git a/ironic/conductor/utils.py b/ironic/conductor/utils.py index 4a0d68a5d..b418d9d0a 100644 --- a/ironic/conductor/utils.py +++ b/ironic/conductor/utils.py @@ -1671,3 +1671,24 @@ def update_image_type(context, node): 'image_type', images.IMAGE_TYPE_WHOLE_DISK if iwdi else images.IMAGE_TYPE_PARTITION) return True + + +def exclude_current_conductor(current_conductor, offline_conductors): + """Wrapper to exclude current conductor from offline_conductors + + In some cases the current conductor may have failed to update + the heartbeat timestamp due to failure or resource starvation. + When this occurs the dbapi get_offline_conductors method will + include the current conductor in its return value. + + :param current_conductor: id or hostname of the current conductor + :param offline_conductors: List of offline conductors. + :return: List of offline conductors, excluding current conductor + """ + if current_conductor in offline_conductors: + LOG.warning('Current conductor %s will be excluded from offline ' + 'conductors. Conductor heartbeat has failed to update the ' + 'database timestamp. This is sign of resource starvation.', + current_conductor) + + return [x for x in offline_conductors if x != current_conductor] |