translatewiki.net suffered degraded service today after changes to Template:Identical caused a huge amount of jobs generated.
The issue was that I had 10+ parallel runJobs.php running for many minutes. First I thought that --exclusive should have prevented this, but on further inspection that never stayed in: https://www.mediawiki.org/wiki/Special:Code/MediaWiki/73198
Then my second thought was that --maxtime=50 should prevent this, but it didn't. It looks like $sTime is correctly set at the beginning, but then it is overwritten again for each job, which means it actually functions like "stop if the last job took longer than maxtime":
// Run the job...
wfProfileIn( METHOD . '-' . get_class( $job ) );
$sTime = microtime( true );
My CRON entry is:
- * * * * betawiki nice php /www/translatewiki.net/w/maintenance/runJobs.php --exclusive --maxtime=50 --procs=1 --memory-limit=250M >> /www/translatewiki.net/w/logs/jobqueue 2> /dev/null
For now I have disabled the CRON entry and running jobqueue manually until it drains.
Version: 1.24rc
Severity: major