Last modified: 2013-05-21 16:41:24 UTC
Production has been replicating memcached writes to redis. We might want to install redis in labs and start doing the same.
Not really needed for now -> lowest priority.
It's actually needed for the jobqueue -- for example in centralnotice when I insert new translation cache regen jobs meta.beta dies because it cannot talk to the backend job queue.
Well if someone volunteer, I guess it needs a dedicated instance for redis then apply whatever puppet manifest and adapt it. Once done mediawiki-config can be tweaked to point beta to that new instance.
Ariel wrote the jobqueue configuration for labs. The conf was previously broken because labs was using jobqueue-pmtpa.php which is for production. https://gerrit.wikimedia.org/r/#/c/64044/
There were some issues with file layout on the job runner too which Antoine fixed up. Anyways there is now a local redis instance in beta but it needs to be puppetized, so I am leaving this bug open for a bit yet.
Assigning to Ariel who is working on this.
This is waiting on a review from Asher of the changeset at https://gerrit.wikimedia.org/r/#/c/64267/ which is in use by labs on deployment-redisdb-test right now.
Thanks for the review.. pushed and live, jobs appear to be working, closing.