Description of problem: start recording error notifications projects (specifically nova) send notifications on error topic when an error occurs. we should capture this in events. Discussion at https://review.openstack.org/#/c/152525/ The following ones were tried during OSP 7 kilo cycle (looked and even applied the diffs for) https://review.openstack.org/#/c/153362/, \https://review.openstack.org/#/c/152525/ but it turns out we do not have the required code on OSP 6 Juno in e.g. /usr/lib/python2.7/site-packages/ceilometer/ to deal with it. I've since reverted those diffs. rabbitmq know of no consumers to notifications.error, e.g: [root@lab ~]# rabbitmqctl list_queues | grep notifications notifications.error 100 notifications.info 0 [root@mlab ~]# rabbitmqctl list_consumers | grep notifications.info | wc -l 225 [root@mlab ~]# rabbitmqctl list_consumers | grep notifications.error | wc -l 0
The problem has nothing to do with Ceilometer unfortunately. As soon as you enable notifications in any OpenStack project, and if that projects emits an "error" notification, it ends up into that queue. No project in OpenStack consume that queue. Ceilometer neither. So it might en up being big for nothing. There should probably be a way to disable this queue, or to make Ceilometer consume it eventually. I'm gonna reassign to oslo.messaging as this is likely the place where such (dis|)enablements should be done.
(In reply to Julien Danjou from comment #8) > The problem has nothing to do with Ceilometer unfortunately. As soon as you > enable notifications in any OpenStack project, and if that projects emits an > "error" notification, it ends up into that queue. > > No project in OpenStack consume that queue. Ceilometer neither. So it might > en up being big for nothing. > > There should probably be a way to disable this queue, or to make Ceilometer > consume it eventually. > > I'm gonna reassign to oslo.messaging as this is likely the place where such > (dis|)enablements should be done. To disable emitting notifications to the unconsumed error queue, you can set on OSP6: notification_driver = noop In the configuration for nova (and others). Does that resolve this issue?
Changing notification_driver to noop will also disable the notification.info queue. The issue is about other notification queues (error/audit/...). Ceilometer upstream decides to consume all notifications queues with Ceilometer since they are no way to just consume one queue and disable other queues without breaking oslo.messaging API/contract.
Thanks
https://git.openstack.org/cgit/openstack/ceilometer/commit/?id=09afee9a8fd665097f912defc844bb80e54fc23d is present in 5.0.5 rebase
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHBA-2016-2714.html