External-Mirrors/awx
2
0
Fork 0
mirror of https://github.com/ansible/awx.git synced 2026-04-11 13:09:21 -02:30
Code Issues Packages Projects Releases Wiki Activity

Make sure that the health checks handle the state transitions properly

Browse Source 
- nodes with states Provisioning, Provisioning Fail, Deprovisioning,
  and Deprovisioning Fail should bypass health checks and should never
  transition due to the existing machinery
- nodes with states Unavailable and Installed can transition to Ready
  if they check out as healthy
- nodes in the Ready state should transition to Unavailable if they
  fail a check
This commit is contained in:
Jeff Bradberry
2022-07-27 17:17:13 -04:00
parent 81e68cb9bf
commit 3bcd539b3d
3 changed files with 18 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
awx/main/models/ha.py
View File

@@ -242,15 +242,18 @@ class Instance(HasPolicyEditsMixin, BaseModel):
return self.last_seen < ref_time - timedelta(seconds=grace_period)
def mark_offline(self, update_last_seen=False, perform_save=True, errors=''):
if self.cpu_capacity == 0 and self.mem_capacity == 0 and self.capacity == 0 and self.errors == errors and (not update_last_seen):
if self.node_state not in (Instance.States.READY, Instance.States.UNAVAILABLE, Instance.States.INSTALLED):
return
if self.node_state == Instance.States.UNAVAILABLE and self.errors == errors and (not update_last_seen):
return
self.node_state = Instance.States.UNAVAILABLE
self.cpu_capacity = self.mem_capacity = self.capacity = 0
self.errors = errors
if update_last_seen:
self.last_seen = now()
if perform_save:
update_fields = ['capacity', 'cpu_capacity', 'mem_capacity', 'errors']
update_fields = ['node_state', 'capacity', 'cpu_capacity', 'mem_capacity', 'errors']
if update_last_seen:
update_fields += ['last_seen']
self.save(update_fields=update_fields)
@@ -307,6 +310,9 @@ class Instance(HasPolicyEditsMixin, BaseModel):
if not errors:
self.refresh_capacity_fields()
self.errors = ''
if self.node_state in (Instance.States.UNAVAILABLE, Instance.States.INSTALLED):
self.node_state = Instance.States.READY
update_fields.append('node_state')
else:
self.mark_offline(perform_save=False, errors=errors)
update_fields.extend(['cpu_capacity', 'mem_capacity', 'capacity'])
@@ -325,7 +331,7 @@ class Instance(HasPolicyEditsMixin, BaseModel):
# playbook event data; we should consider this a zero capacity event
redis.Redis.from_url(settings.BROKER_URL).ping()
except redis.ConnectionError:
errors = _('Failed to connect ot Redis')
errors = _('Failed to connect to Redis')
self.save_health_data(awx_application_version, get_cpu_count(), get_mem_in_bytes(), update_last_seen=True, errors=errors)