Wikimedia\Rdbms\DBTransactionSizeError: Transaction spent 3.6s in writes, exceeding the 3s limit
Closed, ResolvedPublicPRODUCTION ERROR
Actions

Description

Error

mwversion: 1.38.0-wmf.19
reqId: ed3ac035-af1a-4b0d-8442-b055304cc0c8
Find reqId in Logstash

normalized_message

[{reqId}] {exception_url}   Wikimedia\Rdbms\DBTransactionSizeError: Transaction spent 3.6s in writes, exceeding the 3s limit

exception.trace

from /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/loadbalancer/LoadBalancer.php(1724)
#0 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/loadbalancer/LoadBalancer.php(2241): Wikimedia\Rdbms\LoadBalancer->Wikimedia\Rdbms\{closure}(Wikimedia\Rdbms\DatabaseMysqli)
#1 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/loadbalancer/LoadBalancer.php(1747): Wikimedia\Rdbms\LoadBalancer->forEachOpenPrimaryConnection(Closure)
#2 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/lbfactory/LBFactory.php(249): Wikimedia\Rdbms\LoadBalancer->approvePrimaryChanges(array, string, integer)
#3 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/lbfactory/LBFactoryMulti.php(230): Wikimedia\Rdbms\LBFactory::Wikimedia\Rdbms\{closure}(Wikimedia\Rdbms\LoadBalancer, string, array)
#4 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/lbfactory/LBFactory.php(251): Wikimedia\Rdbms\LBFactoryMulti->forEachLB(Closure, array)
#5 /srv/mediawiki/php-1.38.0-wmf.19/includes/libs/rdbms/lbfactory/LBFactory.php(310): Wikimedia\Rdbms\LBFactory->forEachLBCallMethod(string, array)
#6 /srv/mediawiki/php-1.38.0-wmf.19/includes/MediaWiki.php(677): Wikimedia\Rdbms\LBFactory->commitPrimaryChanges(string, array)
#7 /srv/mediawiki/php-1.38.0-wmf.19/includes/MediaWiki.php(647): MediaWiki::preOutputCommit(RequestContext, Closure)
#8 /srv/mediawiki/php-1.38.0-wmf.19/includes/MediaWiki.php(926): MediaWiki->doPreOutputCommit(Closure)
#9 /srv/mediawiki/php-1.38.0-wmf.19/includes/MediaWiki.php(563): MediaWiki->main()
#10 /srv/mediawiki/php-1.38.0-wmf.19/index.php(53): MediaWiki->run()
#11 /srv/mediawiki/php-1.38.0-wmf.19/index.php(46): wfIndexMain()
#12 /srv/mediawiki/w/index.php(3): require(string)
#13 {main}

Impact

Seen multiple occurrences after deploying to group1

Notes

All instances appear to be for categories on Commons.

Details

Request URL: https://commons.wikimedia.org/w/index.php?title=*&action=submit

Subject	Repo	Branch	Lines +/-
rdbms: Normalize DBTransactionSizeError message	mediawiki/core	master	+31 -3
Don't consider lock waits to be write queries	mediawiki/core	master	+5 -3
Don't consider lock waits to be write queries	mediawiki/core	wmf/1.38.0-wmf.19	+5 -3
Revert "rdbms: cleanup the use of QUERY_ flags to query() in Database"	mediawiki/core	wmf/1.38.0-wmf.19	+39 -62

Customize query in gerrit

Related Objects
Search...

		Status	Subtype	Assigned	Task
		Resolved	Release	brennen	T293960 1.38.0-wmf.19 deployment blockers
		Resolved	PRODUCTION ERROR	tstarling	T300194 Wikimedia\Rdbms\DBTransactionSizeError: Transaction spent 3.6s in writes, exceeding the 3s limit

Event Timeline

jeena created this task.Jan 26 2022, 8:26 PM

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJan 26 2022, 8:26 PM

jeena added a parent task: T293960: 1.38.0-wmf.19 deployment blockers.Jan 26 2022, 8:27 PM

Zabe subscribed.Jan 26 2022, 8:30 PM

brennen triaged this task as Unbreak Now! priority.Jan 26 2022, 8:32 PM

brennen updated the task description. (Show Details)

brennen added a project: User-brennen.

brennen subscribed.Jan 26 2022, 8:48 PM

brennen added a project: Commons.Jan 26 2022, 8:54 PM

Similar to T251457?

Is this just timing for certain kinds of edits being saved (?) got so much slower that the DB commit self-terminated?

brennen moved this task from Backlog to Logs/Train on the User-brennen board.Jan 26 2022, 10:06 PM

Change 757533 had a related patch set uploaded (by Ladsgroup; author: Amir Sarabadani):

[mediawiki/core@master] rdbms: Normalize DBTransactionSizeError message

https://gerrit.wikimedia.org/r/757533

gerritbot added a project: Patch-For-Review.Jan 26 2022, 10:25 PM

This doesn't fix the issue, just fixes the error messages ^

I don't know exactly what but something changed the edit transaction and added PageEditStash to it. Compare these two:
https://logstash.wikimedia.org/app/discover#/doc/logstash-*/logstash-mediawiki-2022.01.26?id=2doLmH4Bj9Ddf3wKnd0H
https://logstash.wikimedia.org/app/discover#/doc/logstash-*/logstash-mediawiki-2022.01.26?id=LBgNmH4B7YMytsAjMJGw

I thought it could be the changes I did to flaggedrevs to make it use edit stash but lots of these logs are in commons where flaggedrevs is not even installed. I dig a bit but it's quite late here, if anyone wants to take a look. That'd be awesome.

The very likely cause is that https://gerrit.wikimedia.org/r/c/mediawiki/core/+/746985/2/includes/libs/rdbms/database/Database.php#1129

It makes the acquiring and removing locks a write query so it would get into the transaction (I admit I'm not super sure).

I suggest reverting that patch and moving forward and see if it continues to happen.

For a longer term solution, I recommend removing db part of PageEditStash, it should lock using the cache, not the database, at worst it can use redis lock manager we use for files.

My first guess would be https://gerrit.wikimedia.org/r/c/mediawiki/core/+/746985.

Edit: Ladsgroup was apparently faster than me

Change 757473 had a related patch set uploaded (by Ladsgroup; author: Ladsgroup):

[mediawiki/core@wmf/1.38.0-wmf.19] Revert \"rdbms: cleanup the use of QUERY_ flags to query() in Database\"

https://gerrit.wikimedia.org/r/757473

Ladsgroup added a subscriber: tstarling.Jan 26 2022, 10:55 PM

If jenkins passes, we probably can push this https://gerrit.wikimedia.org/r/c/mediawiki/core/+/757473

Thanks @Ladsgroup. We'll make one more attempt at rolling train forward to group1, assuming it passes.

Mentioned in SAL (#wikimedia-operations) [2022-01-26T23:34:35Z] <brennen> train 1.38.0-wmf.19 (T293960): parking the train at group0 until US morning; we have a probable fix for T300194 but CI is having issues

Stashbot mentioned this in T293960: 1.38.0-wmf.19 deployment blockers.Jan 26 2022, 11:34 PM

Change 757556 had a related patch set uploaded (by Tim Starling; author: Tim Starling):

[mediawiki/core@master] Don't consider lock waits to be write queries

https://gerrit.wikimedia.org/r/757556

I reproduced this locally, my fix above seems to be sufficient. I'm pretty sure the intention of Aaron's change was to replace QUERY_IGNORE_DBO_TRX | QUERY_CHANGE_NONE with QUERY_CHANGE_LOCKS purely for brevity, with the same semantics applied.

Change 757476 had a related patch set uploaded (by Ladsgroup; author: Tim Starling):

[mediawiki/core@wmf/1.38.0-wmf.19] Don't consider lock waits to be write queries

https://gerrit.wikimedia.org/r/757476

Marostegui subscribed.Jan 27 2022, 7:53 AM

Change 757473 abandoned by Ladsgroup:

[mediawiki/core@wmf/1.38.0-wmf.19] Revert \"rdbms: cleanup the use of QUERY_ flags to query() in Database\"

Reason:

Proper fix is in place: I29a02c87156a36

https://gerrit.wikimedia.org/r/757473

Change 757556 merged by jenkins-bot:

[mediawiki/core@master] Don't consider lock waits to be write queries

https://gerrit.wikimedia.org/r/757556

Change 757476 merged by jenkins-bot:

[mediawiki/core@wmf/1.38.0-wmf.19] Don't consider lock waits to be write queries

https://gerrit.wikimedia.org/r/757476

Mentioned in SAL (#wikimedia-operations) [2022-01-27T08:13:43Z] <ladsgroup@deploy1002> Synchronized php-1.38.0-wmf.19/includes/libs/rdbms/database/Database.php: Backport: [[gerrit:757476|Don't consider lock waits to be write queries (T300194)]] (duration: 00m 52s)

Ladsgroup closed this task as Resolved.Jan 27 2022, 8:19 AM

Ladsgroup assigned this task to tstarling.

ReleaseTaggerBot added a project: MW-1.38-notes (1.38.0-wmf.19; 2022-01-24).Jan 27 2022, 9:00 AM

Thanks much, @tstarling & @Ladsgroup.