[10:01:14] <Amir1>	 chrisalbon: I actually do (sorry I missed your ping somehow)
[10:01:36] <Amir1>	 We should reduce number of requests a uwsgi worker takes before restarting
[10:01:44] <Amir1>	 it's currently 200
[10:03:13] <Amir1>	 I prepare a patch
[10:29:51] <Amir1>	 https://gerrit.wikimedia.org/r/c/operations/puppet/+/638467
[10:51:12] <wikibugs>	 10MediaWiki-extensions-ORES, 10Machine Learning Platform: Ores gives 500 error - https://phabricator.wikimedia.org/T267118 (10Hawkeye7)
[11:05:14] <wikibugs>	 10MediaWiki-extensions-ORES, 10Machine Learning Platform, 10User-Ladsgroup: Ores gives 500 error - https://phabricator.wikimedia.org/T267118 (10Ladsgroup) 05Open→03Resolved a:03Ladsgroup I restarted ores and it should be back to normal. The actual long term issue is {T263910}
[14:56:01] <chrisalbon>	 Awesome thank you Amir!
[19:40:45] <chrisalbon>	 Alright, I’ve jerry-rigged a solution for today so I can sleep in peace tonight. If ORES is critical (~1k busy workers) it’ll restart itself. I’ll be manually monitoring throughout the day.
[19:45:05] <chrisalbon>	 I'm going to set the restart threshold at 500 workers, which based on past data is when clearly a tipping point has been reached and ORES needs restart.