Page MenuHomePhabricator

⬆️ Triage MediaWiki 1.43 staging errors
Closed, ResolvedPublic

Description

Determine if the new errors we are seeing on staging need to be addressed, or if they are simply due to overzealous logging by MediaWiki. The list of new errors can be found at: https://console.cloud.google.com/errors;time=P30D;locations=global?project=wikibase-cloud

n.b.: ignore errors occurring from 1.43 pods accessing 1.39 databases

A/C:

  • categorise top 5 by count new error groups into: fix or ignore

Initial Timebox: 8 hours

Event Timeline

Need more investigation

To ignore

  • All Error 1146 as they are caused by an attempt to read the wrong db pod

@Andrew-WMDE and I checked these 4 examples to make sure these were dues to 143 pods making requests on 139 databases https://console.cloud.google.com/errors/detail/CPmpkcnq-vWsmgE;time=P7D;locations=global?project=wikibase-cloud ~3300 occurences
https://console.cloud.google.com/errors/detail/CPesrdL0srieRQ;time=P7D;locations=global?project=wikibase-cloud ~42
https://console.cloud.google.com/errors/detail/CKW6i_LooYvaMA;time=P7D;locations=global?project=wikibase-cloud ~7
https://console.cloud.google.com/errors/detail/CL6vq9froNPIVg;time=P7D;locations=global?project=wikibase-cloud ~ 3700

Investigate

[rdbms] Expectation (masterConns <= 0) by MediaWiki\MediaWikiEntryPoint::restInPeace not met (actual: 1): [connect to sql-mariadb-primary.default.svc.cluster.local ()]

from https://console.cloud.google.com/errors/detail/COyO-aiI_5q2Zg;time=P7D;locations=global?project=wikibase-cloud
see T408101

[SQLBagOStuff] Duplicate get(): "mwdb_f036984f01-mwt_93f623752b_:echo:seen:message:time:1" fetched 2 times

from https://console.cloud.google.com/errors/detail/CLLJm-eviqOjaw;time=P7D;locations=global?project=wikibase-cloud
see T408101

[RedisBagOStuff] MediaWiki\Extension\OAuth\Backend\MWOAuthToken value for 'OAUTH:mwdb_f036984f01-mwt_93f623752b_:token:TOKEN:request:REQUEST'; serialization is suspect.

from https://console.cloud.google.com/errors/detail/CPbfnf7y8LGyaA;time=P7D;locations=global?project=wikibase-cloud
see T408103

Monitor

Error 1146: Table 'mwdb_cc73c3ce73.mwt_edcc7ed9cb_block' doesn't exist Function

from https://console.cloud.google.com/errors/detail/CPmpkcnq-vWsmgE;time=P7D;locations=global?project=wikibase-cloud

  • Likely caused by the Platform API attempting to use the MediaWiki 1.43 back-end to run jobs for instances that have not yet been upgraded from MediaWiki 1.39
  • Should resolve itself once all instances are upgraded to MediaWiki 1.43 or sooner by T407504
  • Check database version via SQL using: SELECT version FROM apidb.wiki_dbs WHERE name = '';

[rdbms] Error 1146 from MediaWiki\Block\DatabaseBlockStore::newLoad, Table 'mwdb_cc73c3ce73.mwt_edcc7ed9cb_block' doesn't exist

from https://console.cloud.google.com/errors/detail/CL6vq9froNPIVg;time=P7D;locations=global?project=wikibase-cloud

  • Same as Error 1146 above

Error 1054: Unknown column 'pl_namespace' in 'where clause' Function: BacklinkCache::queryLinks Query

from https://console.cloud.google.com/errors/detail/CPvV6dSKuZHJ4gE;time=P30D;locations=global?project=wikibase-cloud

  • Has not occurred since the initial deploy of MediaWiki 1.43 on staging
  • I suspect a straggling job with a mismatched MediaWiki back-end

[runJobs] htmlCacheUpdate Main_Page table=pagelinks recursive=1 rootJobIsSelf=1 rootJobSignature= rootJobTimestamp=causeAction=page-touch namespace=0 title=Main_Page requestId= causeAgent=unknown (id=3,timestamp=20251015090341) t=9 error=Wikimedia\Rdbms\DBQueryError: Error 1054: Unknown column 'pl_namespace' in 'where clause' Function: BacklinkCache::queryLinks

from https://console.cloud.google.com/errors/detail/CKDn5ISY0bjMuAE;time=P30D;locations=global?project=wikibase-cloud

  • Same as Error 1054 above

[ParserCache] Inconsistent revision ID

from https://console.cloud.google.com/errors/detail/CLTtg5LK_d7fFQ;locations=global?project=wikibase-cloud

  • Has not occurred again for over a week
Resolved

Undefined constant MediaWiki\Extension\OAuth\Backend\ConsumerAcceptance::READ_NORMAL

from https://console.cloud.google.com/errors/detail/CP2Tj_KIlOvNEQ;time=P7D;locations=global?project=wikibase-cloud
see wbstack/mediawiki/pull/492 | Call IDBAccessObject::READ_NORMAL directly

Call to undefined method MWHttpRequest::factory()

from https://console.cloud.google.com/errors/detail/CIeK26H_6fWJRQ;time=P7D;locations=global?project=wikibase-cloud
see wbstack/mediawiki/pull/493 | Replace call to removed MWHttpRequest::factory

Ollie.Shotton_WMDE changed the task status from Open to Stalled.Oct 29 2025, 10:20 AM
Ollie.Shotton_WMDE removed Rosalie_WMDE as the assignee of this task.
Ollie.Shotton_WMDE added a subscriber: Rosalie_WMDE.
Ollie.Shotton_WMDE subscribed.

Unassigning and marking as stalled until we can confirm that we don't need to take further action on the "monitor" errors.

Unassigning and marking as stalled until we can confirm that we don't need to take further action on the "monitor" errors.

After talking about it in the daily today I created T408789 as a follow up, so we can move this task here to done

Tarrow claimed this task.