Last modified: 2014-11-14 10:16:46 UTC
Vanadium (the main machine for collecting EventLogging data) used to have logrotation of EventLogging data set to 90 days of EventLogging data. Due to disk space getting low, it is now at 45 days [1]. That's not an issue per se, as we're not using vanadium to keeping history. We have full history at stat1002+stat1003 for that. Should we go back to holding 90 days of EventLogging data on vanadium? (Ops do not like the current setup of logging onto the root mount. And a transparent setup to logging to a different mount with more capacity would be easily possible. We just need to make sure to that the switch to the new mountpoint does not cause data loss.) --------------------------------- [1] The free disk space was getting low, Ops got alerted. andrewbogott handled the issue, and decreased the window after checking back with me. The relavant change in operations/puppet is Ia0bd336ab9df5d2f6fa41590351cd0d6c5980b98 The relevant logs are at http://bots.wmflabs.org/~wm-bot/logs/%23wikimedia-operations/20140801.txt starting on [20:00:43]
The RT ticket for moving the mount is RT #8063.
Change 172707 had a related patch set uploaded by QChris: Retain 90 days of EventLogging logs https://gerrit.wikimedia.org/r/172707
Change 172707 abandoned by Ottomata: Retain 90 days of EventLogging logs Reason: This has been done in templates/logrotate.erb https://gerrit.wikimedia.org/r/172707
Change 172707 restored by QChris: Retain 90 days of EventLogging logs Reason: Seems the template still says 45 days. https://gerrit.wikimedia.org/r/172707
Change 172707 merged by Ottomata: Retain 90 days of EventLogging logs https://gerrit.wikimedia.org/r/172707