[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Page MenuHomePhabricator

Error: Typed property GrowthExperiments\NewcomerTasks\AddLink\LinkRecommendationUpdater::$linkRecommendationTaskType must not be accessed before initialization
Closed, ResolvedPublicPRODUCTION ERROR

Description

Error
labels.normalized_message
[f9af90088827d1f065d5bb5d] [no req]   Error: Typed property GrowthExperiments\NewcomerTasks\AddLink\LinkRecommendationUpdater::$linkRecommendationTaskType must not be accessed before initialization
error.stack_trace
from /srv/mediawiki/php-1.41.0-wmf.30/extensions/GrowthExperiments/includes/NewcomerTasks/AddLink/LinkRecommendationUpdater.php(266)
#0 /srv/mediawiki/php-1.41.0-wmf.30/extensions/GrowthExperiments/includes/NewcomerTasks/AddLink/LinkRecommendationUpdater.php(158): GrowthExperiments\NewcomerTasks\AddLink\LinkRecommendationUpdater->getLinkRecommendationTaskType()
#1 /srv/mediawiki/php-1.41.0-wmf.30/extensions/GrowthExperiments/includes/NewcomerTasks/AddLink/LinkRecommendationUpdater.php(87): GrowthExperiments\NewcomerTasks\AddLink\LinkRecommendationUpdater->evaluateTitle(MediaWiki\Title\Title, MediaWiki\Revision\RevisionStoreRecord, boolean)
#2 /srv/mediawiki/php-1.41.0-wmf.30/extensions/GrowthExperiments/maintenance/refreshLinkRecommendations.php(285): GrowthExperiments\NewcomerTasks\AddLink\LinkRecommendationUpdater->processCandidate(MediaWiki\Title\Title, boolean)
#3 /srv/mediawiki/php-1.41.0-wmf.30/extensions/GrowthExperiments/maintenance/refreshLinkRecommendations.php(160): GrowthExperiments\Maintenance\RefreshLinkRecommendations->processCandidate(MediaWiki\Title\Title, boolean)
#4 /srv/mediawiki/php-1.41.0-wmf.30/maintenance/includes/MaintenanceRunner.php(703): GrowthExperiments\Maintenance\RefreshLinkRecommendations->execute()
#5 /srv/mediawiki/php-1.41.0-wmf.30/maintenance/run.php(51): MediaWiki\Maintenance\MaintenanceRunner->run()
#6 /srv/mediawiki/multiversion/MWScript.php(159): require_once(string)
#7 {main}
Impact
Notes

Event Timeline

Aklapper triaged this task as Unbreak Now! priority.Oct 12 2023, 8:41 AM

All the errors came from mwmaint2002 from the GrowthExperiments script maintenance/refreshLinkRecommendations.php.

There were some errors at 8:26 and 8:27 followed by a constant stream of errors from 8:29 to 8:34 which caused me to roll back the train. Then I am not sure whether the new code was related since apparently the exceptions stopped just when I started the rollback but it was only effective at 8:53.

GrowthExperiments_maintenance_script.png (383×796 px, 20 KB)

The periodic job from Puppet:

profile::mediawiki::periodic_job { "growthexperiments-refreshLinkRecommendations-${name}":
    command  => "/usr/local/bin/foreachwikiindblist 'growthexperiments & ${name}' extensions/GrowthExperiments/maintenance/refreshLinkRecommendations.php --verbose",
    interval => '*-*-* *:27:00',
}

There were no exceptions previously. So I guess it is indeed related to wmf.30. It does not look like anything meaningful got introduced in the extension.

We have promoted the wikis to wmf.30 at 8:28 which matches the start of the exception. I have triggered the rollback at 8:34 which for some reason matches the end of the exception stream but maybe it is unrelated. The sync of code happend between 8:49-8:50. So I don't know :(

Maybe it is because some wmf.29 cached values got retrieved by the wmf.30 code and are not compatible leading to the error?

Change 965476 had a related patch set uploaded (by Phuedx; author: Phuedx):

[mediawiki/extensions/GrowthExperiments@master] LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration

https://gerrit.wikimedia.org/r/965476

Change 965476 merged by jenkins-bot:

[mediawiki/extensions/GrowthExperiments@master] LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration

https://gerrit.wikimedia.org/r/965476

FWIW, this patch is good in the sense that I agree that linkRecommendationTaskType should be nullable, but in local dev, it looks like the task type still isn't getting set correctly. I now get the exception thrown in the getLinkRecommendationTaskType helper:

'link-recommendation' is not a link recommendation task type

This may be local config; I'm double checking now.

Okay, yes, I verified that I was missing some configuration for link recommendations specifically, in my GrowthExperiments config. Locally,
docker compose exec mediawiki php extensions/GrowthExperiments/maintenance/refreshLinkRecommendations.php --verbose
is now processing topics and exiting with status 0.

Change 965217 had a related patch set uploaded (by Phuedx; author: Phuedx):

[mediawiki/extensions/GrowthExperiments@wmf/1.41.0-wmf.30] LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration

https://gerrit.wikimedia.org/r/965217

Nice thank you! I will approve and deploy the backport then I guess promote the group2 wikis again.

Change 965217 merged by jenkins-bot:

[mediawiki/extensions/GrowthExperiments@wmf/1.41.0-wmf.30] LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration

https://gerrit.wikimedia.org/r/965217

Mentioned in SAL (#wikimedia-operations) [2023-10-12T13:11:55Z] <hashar@deploy2002> Started scap: Backport for [[gerrit:965217|LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration (T348719)]]

Mentioned in SAL (#wikimedia-operations) [2023-10-12T13:13:15Z] <hashar@deploy2002> phuedx and hashar: Backport for [[gerrit:965217|LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration (T348719)]] synced to the testservers (https://wikitech.wikimedia.org/wiki/Mwdebug)

Mentioned in SAL (#wikimedia-operations) [2023-10-12T13:18:47Z] <hashar@deploy2002> Finished scap: Backport for [[gerrit:965217|LinkRecommendationUpdater: Update $linkRecommendationTaskType declaration (T348719)]] (duration: 06m 51s)

I have promoted the wikis. The cron job will starts at 14:27 UTC (23 minutes from now).

Mentioned in SAL (#wikimedia-operations) [2023-10-12T14:11:44Z] <urbanecm> mwmaint2002: stop previous instance of refreshLinkRecommendations maintenance job (T348719)

hashar assigned this task to phuedx.

The scheduled job should have triggered at 14:27 and I don't see the error in the logs which thus looks like the issue got fixed.

Maybe the maintenance job had nothing to do for this iteration. If the issue has not been fixed, the next time it does process things that will log an error which will point us back to this task for more investigation.

Lets assume that got fixed for real (and that got tested via MediaWiki docker).