Page MenuHomePhabricator

Zuul: statsd metrics for Gerrit events are broken
Closed, DeclinedPublic

Description

Whenever a Gerrit event is received by Zuul, it increments a counter in statsd. The format is:

gerrit.event.<event type>

For some reason the metrics are broken and stuck to 1. Apparently since March 2014.

I give it a try with:

echo "gerrit.test.hashar:5|c"| nc -w1 -q0 -u statsd.eqiad.wmnet 8125

Nothing is created.

I guess we want to drop the whole gerrit.* metric namespace and try again. CCing Ori since he knows about statsd.


Version: wmf-deployment
Severity: normal
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=62667

Details

Reference
bz66174

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 22 2014, 3:15 AM
bzimport set Reference to bz66174.
bzimport added a subscriber: Unknown Object (MLST).

Matt Walker has been talking to ops about problems that sound related in using txstatsd. I think there are plans in the works to either fix txstatsd or find a new statsd implementation. I don't know if there is an RT ticket or bug to track that yet.

Maybe we can drop the whole gerrit. namespace and hope that it clears things up?

I have no need for the history, the main purpose is to display a graph of patchset/comment added on the Zuul status page though I removed it. A live example at openstack http://status.openstack.org/zuul/ (browse to the bottom).

I have dropped the graph from https://integration.wikimedia.org/zuul/ which was the only use case.