Page MenuHomePhabricator

another round of activemq queue decrufting needed
Closed, ResolvedPublic

Description

We're over 30% store percent utilization again which is perturbing our early warning alerts, could we do another round of queue cleanup?

Event Timeline

Change 289566 had a related patch set uploaded (by Ejegg):
Validate banner history messages before insert

https://gerrit.wikimedia.org/r/289566

Jgreen triaged this task as Unbreak Now! priority.May 19 2016, 12:59 AM

Looks like the offending queue is banner-history which has been growing steadily since the weekend and is now up to 150K messages.

ActiveMQ crashed earlier today and from the ramp-up in RAM utilization I think we're headed for another soon. We are going to take banners down until the situation can be fixed.

Looks like the offending queue is banner-history which has been growing steadily since the weekend and is now up to 150K messages.

ActiveMQ crashed earlier today and from the ramp-up in RAM utilization I think we're headed for another soon. We are going to take banners down until the situation can be fixed.

https://ganglia.wikimedia.org/latest/graph.php?r=month&z=xlarge&c=Fundraising+eqiad&h=silicon.frack.eqiad.wmnet&jr=&js=&event=hide&ts=0&v=151484&m=ActiveMQ+QueueSize+banner-history&vl=Messages

Change 289566 merged by Ejegg:
Validate banner history messages before insert

https://gerrit.wikimedia.org/r/289566