External-Mirrors/awx
2
0
Fork 0
mirror of https://github.com/ansible/awx.git synced 2026-05-19 14:57:39 -02:30
Code Issues Packages Projects Releases Wiki Activity

only reap non-netsplit nodes

Browse Source
This commit is contained in:
Chris Meyers
2017-08-14 10:37:19 -04:00
parent 5ad29c9081
commit 9314db646b
2 changed files with 148 additions and 95 deletions
Download Patch File Download Diff File Expand all files Collapse all files

94
awx/main/scheduler/__init__.py
View File

@@ -69,14 +69,43 @@ class TaskManager():
'''
Tasks that are running and SHOULD have a celery task.
{
'execution_node': [j1, j2,...],
'execution_node': [j3],
...
}
'''
def get_running_tasks(self):
execution_nodes = {}
now = tz_now()
return list(UnifiedJob.objects.filter(Q(status='running') |
jobs = list(UnifiedJob.objects.filter(Q(status='running') |
(Q(status='waiting', modified__lte=now - timedelta(seconds=60)))))
for j in jobs:
if j.execution_node in execution_nodes:
execution_nodes[j.execution_node].append(j)
elif j.execution_node not in execution_nodes:
execution_nodes[j.execution_node] = [j]
return execution_nodes
'''
Tasks that are currently running in celery
Transform:
{
"celery@ec2-54-204-222-62.compute-1.amazonaws.com": [],
"celery@ec2-54-163-144-168.compute-1.amazonaws.com": [{
...
"id": "5238466a-f8c7-43b3-9180-5b78e9da8304",
...
}]
}
to:
{
"celery@ec2-54-204-222-62.compute-1.amazonaws.com": [
"5238466a-f8c7-43b3-9180-5b78e9da8304",
]
}
'''
def get_active_tasks(self):
inspector = inspect()
@@ -86,15 +115,26 @@ class TaskManager():
logger.warn("Ignoring celery task inspector")
active_task_queues = None
active_tasks = set()
queues = None
if active_task_queues is not None:
queues = {}
for queue in active_task_queues:
active_tasks = set()
map(lambda at: active_tasks.add(at['id']), active_task_queues[queue])
# queue is of the form celery@myhost.com
queue_name = queue.split('@')
if len(queue_name) > 1:
queue_name = queue_name[1]
else:
queue_name = queue_name[0]
queues[queue_name] = active_tasks
else:
if not hasattr(settings, 'CELERY_UNIT_TEST'):
return (None, None)
return (active_task_queues, active_tasks)
return (active_task_queues, queues)
def get_latest_project_update_tasks(self, all_sorted_tasks):
project_ids = Set()
@@ -380,32 +420,38 @@ class TaskManager():
logger.debug("Failing inconsistent running jobs.")
celery_task_start_time = tz_now()
active_task_queues, active_tasks = self.get_active_tasks()
active_task_queues, active_queues = self.get_active_tasks()
cache.set('last_celery_task_cleanup', tz_now())
if active_tasks is None:
if active_queues is None:
logger.error('Failed to retrieve active tasks from celery')
return None
all_running_sorted_tasks = self.get_running_tasks()
for task in all_running_sorted_tasks:
if (task.celery_task_id not in active_tasks and not hasattr(settings, 'IGNORE_CELERY_INSPECTOR')):
# TODO: try catch the getting of the job. The job COULD have been deleted
if isinstance(task, WorkflowJob):
continue
if task.modified > celery_task_start_time:
continue
task.status = 'failed'
task.job_explanation += ' '.join((
'Task was marked as running in Tower but was not present in',
'Celery, so it has been marked as failed.',
))
task.save()
awx_tasks._send_notification_templates(task, 'failed')
task.websocket_emit_status('failed')
logger.error("%s appears orphaned... marking as failed", task.log_format)
'''
Only consider failing tasks on instances for which we obtained a task
list from celery for.
'''
execution_nodes_jobs = self.get_running_tasks()
for node, node_jobs in execution_nodes_jobs.iteritems():
if node not in active_queues:
continue
active_tasks = active_queues[node]
for task in node_jobs:
if (task.celery_task_id not in active_tasks and not hasattr(settings, 'IGNORE_CELERY_INSPECTOR')):
# TODO: try catch the getting of the job. The job COULD have been deleted
if isinstance(task, WorkflowJob):
continue
if task.modified > celery_task_start_time:
continue
task.status = 'failed'
task.job_explanation += ' '.join((
'Task was marked as running in Tower but was not present in',
'Celery, so it has been marked as failed.',
))
task.save()
awx_tasks._send_notification_templates(task, 'failed')
task.websocket_emit_status('failed')
logger.error("Task %s appears orphaned... marking as failed" % task)
def calculate_capacity_used(self, tasks):
for rampart_group in self.graph: