Last weekend, there was an heap issue with the Camus mapper JVMs that caused the mediawiki_events to fail for a while. Luca fixed this, but it seems the cirrussearch-request topics never caught back up.
Andrew noticed that the mediawiki_events camus job was poorly configured since the addition of more and higher volume mediawiki event topics. He has created a new camus job 'mediawiki_analytics_events' for these topics: mediawiki.cirrussearch-request and mediawiki.api-request.
After running over night, this job is running, but isn't quite going fast enough.
Andrew will try to get the current job as is to catch up by running it manually and giving it more time to run.
We should also add some partitions to these topics to allow more camus mappers to run in parallel.