Last modified: 2014-06-24 14:09:53 UTC
Redis configuration on wikimetrics in puppet is incorrect for both production and staging. Problems are two: - redis does not persist well to disk - keeps getting restarted
Prioritization and scheduling of this bug is tracked on Mingle card https://wikimedia.mingle.thoughtworks.com/projects/analytics/cards/cards/1532
Additional things changed in Redis configuration on production: save 900 1 save 300 10 save 60 20 stop-writes-on-bgsave-error yes Additional things changed in Redis configuration on staging save 900 1 save 300 1 save 60 2 stop-writes-on-bgsave-error yes
This is having the effect of deleting people's data too soon. In 1 day instead of 30 days.
Change 140173 had a related patch set uploaded by Milimetric: Stop relying on limited redis module https://gerrit.wikimedia.org/r/140173
Change 140173 merged by Ottomata: Stop relying on limited redis module https://gerrit.wikimedia.org/r/140173
Change 140380 had a related patch set uploaded by Ottomata: Use custom redis.conf.erb template for wikimetrics in labs https://gerrit.wikimedia.org/r/140380
Change 140380 merged by Ottomata: Use custom redis.conf.erb template for wikimetrics in labs https://gerrit.wikimedia.org/r/140380
This turned out to be a problem in both the redis configuration AND the celery configuration. Yay team-bugslaying :)
Change 140764 had a related patch set uploaded by Nuria: Reports in prod should be stored on redis 30 days https://gerrit.wikimedia.org/r/140764
Change 140764 merged by Ottomata: Reports in prod should be stored on redis 30 days https://gerrit.wikimedia.org/r/140764