Merge lp:~jameinel/maas/ignore_results into lp:~maas-committers/maas/trunk

[John A Meinel]

Pass ignore_result=True to all current celery tasks.

With celery, you are meant to fire off a job, and then grab the
result object and wait for it. The design of all current jobs is
that they just trigger and you never wait for the result.

If you don't ignore the results, then the queues build up forever
and stay alive because they wait for someone to read the result.

See commit message

[Jeroen T. Vermeulen (jtv)]

I thought we had a global setting for this somewhere. If there's any more of this, maybe we need our own wrapper for @task.

review: Approve

[John A Meinel (jameinel)]

I saw something about setting an expiry time for queues. I'm not sure that ignore_result can be set in config.

[Raphaël Badin (rvb)]

Not sure why you need to do that, CELERY_IGNORE_RESULT is set to True in etc/celeryconfig_common.py.

?

[John A Meinel (jameinel)]

Re-approving to revert the original change.

=== modified file 'src/provisioningserver/tasks.py'
--- src/provisioningserver/tasks.py	2012-10-04 05:49:09 +0000
+++ src/provisioningserver/tasks.py	2012-10-05 10:05:26 +0000
@@ -64,7 +64,7 @@
 celery_config = app_or_default().conf
 
 
-@task
+@task(ignore_result=True)
 def refresh_secrets(**kwargs):
     """Update the worker's knowledge of various secrets it needs.
 
@@ -135,13 +135,13 @@
     # TODO: signal to webapp that it worked.
 
 
-@task
+@task(ignore_result=True)
 def power_on(power_type, **kwargs):
     """Turn a node on."""
     issue_power_action(power_type, 'on', **kwargs)
 
 
-@task
+@task(ignore_result=True)
 def power_off(power_type, **kwargs):
     """Turn a node off."""
     issue_power_action(power_type, 'off', **kwargs)
@@ -340,7 +340,7 @@
     boot_images.report_to_server()
 
 
-@task
+@task(ignore_result=True)
 def update_node_tags(tag_name, tag_definition):
     """Update the nodes for a new/changed tag definition.