So we notice if it falls over.
Description
Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | Jdforrester-WMF | T90281 Improve citoid production service | |||
Resolved | ori | T87496 Provide service alerting/statistics for the citoid and zotero services |
Event Timeline
Something like
https://wikitech.wikimedia.org/wiki/Parsoid#Monitoring ?
I don't have access to the current living versions of citoid except for the one in the services project on novawiki (a.k.a. not the one Roan set up for deploy) so I'm not sure how much I can do here- although I am happy to be given access and mess around :).
Also we probably need to monitor Zotero too. Moreso, I think. I've had some random Zotero dying happen on me and I was never able to replicate the circumstances that caused it.
There are different ways to monitor it. High-level by putting metrics in ganglia and then checking for changes there or a bit lower-level with Icinga to check stuff like "is the process even running". I can do the latter, while others are more familiar with the former. If you want something in Icinga let me know a typical line from the output of "ps" please that indicates zotero process is running normally.
The icinga stuff for zotero is done in https://gerrit.wikimedia.org/r/#/c/194495/10/modules/nagios_common/files/checkcommands.cfg,cm
It is an HTTP check in that it tries to convert (yes it is actually converting, not exporting as the /export url suggests) an empty citation to the "wikipedia" format. Getting metrics out of zotero in order to have meaningful graphs in graphite or ganglia, I don't think is doable. I haven't found yet a way to query zotero about it's internal state so I am guessing if we want something we will have to rely on log parsing.
Change 197126 had a related patch set uploaded (by Mobrovac):
Report metrics using StatsD
Change 197310 had a related patch set uploaded (by Mobrovac):
Citoid: set the StatsD host to statsd.eqiad.wmnet
Change 197310 merged by Alexandros Kosiaris:
Citoid: set the StatsD host to statsd.eqiad.wmnet