Last modified: 2014-11-17 09:21:20 UTC
Brought up originally at https://www.mediawiki.org/wiki/Talk:Release_Management_RFP/2014/Consortium#In-place_updates - I thought it deserved a bug. *We should do schema updates first, then populate-fields updates second. Most population updates are semi-optional, where the schema changes cause actual errors. If we do all the schema updates first, this minimizes the amount of time that the wiki will serve errors in the common scenario where a user replaces the live mediawiki directory with the new tarball and runs update.php *Some of our population maintenance scripts work in batches, followed by calling wfWaitForSlaves(); . If the server doesn't use slave dbs, then that should be a no-op. There won't be any slave lag, but if the server is also serving real requests at the same time, perhaps we should replace it with a small sleep() call to reduces load on the server when mass updating fields. *Probably other things.
