It seems odd that such a modest batch size with wfWaitForSlaves() could cause so much lag anywhere. It seems like something is off somewhere.
On Tue, Jul 22, 2014 at 12:54 PM, James Forrester <jforres...@wikimedia.org> wrote: > On 22 July 2014 09:24, Antoine Musso <hashar+...@free.fr> wrote: > >> Hello, >> >> Yesterday a patch landed in MediaWiki core that causes the database >> update script to update a bunch of tables and do a lot of queries on the >> database. >> >> The slave database is thus lagged out by 4+ hours as I write this and >> that causes a bunch of errors when browsing the beta sites. >> >> There is a few more details on: >> >> populateBacklinkNamespace script causing massive slave lag on beta >> https://bugzilla.wikimedia.org/show_bug.cgi?id=68349 >> >> Bryan Davis pointed me to a quite useful link to check the database lag: >> >> http://en.wikipedia.beta.wmflabs.org/w/index.php?maxlag=-1 >> >> >> Jenkins is still processing the upgrade of simplewiki which is the >> largest wiki on beta cluster. The job output is: >> >> >> https://integration.wikimedia.org/ci/job/beta-update-databases-eqiad/label=deployment-bastion-eqiad,wikidb=simplewiki/2742/console >> >> Whenever that is completed, the database slave will have to catch up and >> we will eventually resume normal operations. >> > > As an update, the database upgrade finally finished about 20 minutes ago, > and slave lag is slowly adjusting back to normal. Unfortunately the lag is > currently nearly four hours long, so this will take a while. > > J. > -- > James D. Forrester > Product Manager, Editing > Wikimedia Foundation, Inc. > > jforres...@wikimedia.org | @jdforrester > > _______________________________________________ > Engineering mailing list > engineer...@lists.wikimedia.org > https://lists.wikimedia.org/mailman/listinfo/engineering > > -- -Aaron S
_______________________________________________ QA mailing list QA@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/qa