bingle-admin wrote:

Prioritization and scheduling of this bug is tracked on Mingle card https://wikimedia.mingle.thoughtworks.com/projects/analytics/cards/1332

We recently reviewed moving Wikimetrics to production and after a discussion with Ops decided that there wasn't justification to do so.

Dan -- I believe we have started backing up data and the chances of losing data are quite remote. (But users should still keep copies of cohorts of course)

Can you comment?

thanks,

-Toby

I never commented on this! Sorry. Ok, so as Toby mentions, we talked this over with Ops and Labs folks. We decided that as long as we had a decent backup strategy in place, Labs would be a fine place to host Wikimetrics. Our backup strategy:

1. we back up report results using features built into Redis
2. we back up wikimetrics metadata (users, cohorts, etc.) daily using a simple cron

Possible improvements:

• increase frequency of wikimetrics database backup
• have the team review the Redis configuration
• perform a recover after a simulated failure

Users should keep copies of anything they absolutely never want to lose, but that's true of any website out there (gmail, facebook, etc.). So while we can't go to the lengths that google or facebook go to, we try as much as possible to keep your data safe. And the Labs team has our back, they think most problems will get addressed quickly.