Page MenuHomePhabricator

Re-enable daily updates of formerly slow enwiki QueryPages
Open, Needs TriagePublic

Description

Background: in T142936: Generate Special:LonelyPages on a more regular basis, we switched very slow enwiki QueryPages from biyearly to monthly. Currently the config looks like:

# Disable all the query pages that take more than about 15 minutes to update
# Specify a "run mode" when a cron job is set up for the special page and
# also create a message for the "run mode" in WikimediaMessages extension with
# the message key 'querypage-updates-<run mode>'
# wgDisableQueryPageUpdate @{
'wgDisableQueryPageUpdate' => [
	'default' => [
		'Ancientpages' => 'half-monthly',
		'Deadendpages' => 'half-monthly',
		'Fewestrevisions' => 'half-monthly',
		'Mostlinked' => 'half-monthly',
		'Mostrevisions' => 'half-monthly',
		'Wantedpages' => 'half-monthly',
	],
	'small' => [], // T45668
	'medium' => [], // T48094

	// Individual wikis
	' enwiki' => [
		'Lonelypages' => 'monthly',
		'Mostcategories' => 'monthly',
		'Mostlinkedcategories' => 'monthly',
		'Mostlinkedtemplates' => 'monthly',
		'Uncategorizedcategories' => 'monthly',
		'Wantedtemplates' => 'monthly',
	],
        ...
],

For whatever reason, whether it's MediaWiki schema changes, better hardware, etc., these pages are much faster now.

  • Lonelypages: 8 minutes
  • Mostcategories: 6-7 minutes
  • Mostlinkedcategories: no longer expensive, the cronjob literally does nothing.
  • Mostlinkedtemplates: 20-24 minutes
  • Uncategorizedcategories: 1-3 minutes
  • Wantedtemplates: 10-13 minutes

Question for DBA: can we move these pages to the standard update timeline (every 3 days) now? Note that in T307314#7997375 I've already asked to increase that frequency over time to daily. If we need to leave Mostlinkedtemplates on the slow queue that would be understandable.

Event Timeline

Change 804803 had a related patch set uploaded (by Legoktm; author: Legoktm):

[operations/puppet@production] mediawiki: Disable useless mostlinkedcategories update job

https://gerrit.wikimedia.org/r/804803

Change 804805 had a related patch set uploaded (by Legoktm; author: Legoktm):

[operations/mediawiki-config@master] Remove misleading "disable" of Special:Mostlinkedcategories

https://gerrit.wikimedia.org/r/804805

I think that's fine as long as they go to the vslow replica. Further, if we can make all this to reload the DB config (T298485) often that'd be great so it doesn't mess up with schema changes (even if the runtime times you posted are quite fast, it would be good to have it implemented).

Change 804803 merged by Legoktm:

[operations/puppet@production] mediawiki: Disable useless mostlinkedcategories update job

https://gerrit.wikimedia.org/r/804803

Change 804805 merged by jenkins-bot:

[operations/mediawiki-config@master] Remove misleading "disable" of Special:Mostlinkedcategories

https://gerrit.wikimedia.org/r/804805

Mentioned in SAL (#wikimedia-operations) [2023-04-05T23:50:46Z] <legoktm@deploy2002> Started scap: Backport for [[gerrit:804805|Remove misleading "disable" of Special:Mostlinkedcategories (T310456)]]

Mentioned in SAL (#wikimedia-operations) [2023-04-05T23:52:08Z] <legoktm@deploy2002> legoktm: Backport for [[gerrit:804805|Remove misleading "disable" of Special:Mostlinkedcategories (T310456)]] synced to the testservers: mwdebug1002.eqiad.wmnet, mwdebug2001.codfw.wmnet, mwdebug2002.codfw.wmnet, mwdebug1001.eqiad.wmnet

Mentioned in SAL (#wikimedia-operations) [2023-04-05T23:58:41Z] <legoktm@deploy2002> Finished scap: Backport for [[gerrit:804805|Remove misleading "disable" of Special:Mostlinkedcategories (T310456)]] (duration: 07m 55s)