Last modified: 2012-05-03 02:42:48 UTC

Wikimedia Bugzilla is closed!

Wikimedia migrated from Bugzilla to Phabricator. Bug reports are handled in Wikimedia Phabricator.
This static website is read-only and for historical purposes. It is not possible to log in and except for displaying bug reports and their history, links might be broken. See T36451, the corresponding Phabricator task for complete and up-to-date bug report information.
Bug 34451 - Heavy DB Usage from API call list=allusers with auactiveusers=
Heavy DB Usage from API call list=allusers with auactiveusers=
Product: MediaWiki
Classification: Unclassified
API (Other open bugs)
All All
: Normal normal (vote)
: 1.19.0 release
Assigned To: Nobody - You can work on this!
Depends on:
Blocks: 31217
  Show dependency treegraph
Reported: 2012-02-16 17:56 UTC by Mark A. Hershberger
Modified: 2012-05-03 02:42 UTC (History)
8 users (show)

See Also:
Web browser: ---
Mobile Platform: ---
Assignee Huggle Beta Tester: ---


Description Mark A. Hershberger 2012-02-16 17:56:44 UTC
It seems to me that the "activeusers" parameter in the allusers API-query [] is quite heavy on the database.

I tried
earlier today which was "Served by mw7 in 16.976 secs". It's probably in an SQL cache now because I can get the result much quicker now.
Seeing that I followed with a request that is not filtered by group, which timeouts:

Looking at the code I see that it's implemented as a grouped JOIN of `recentchanges` and `user` with an un-indexed WHERE; I'm no MYSQL expert but that doesn't seem like a good idea on enWP.
Comment 1 Mark A. Hershberger 2012-02-16 17:57:19 UTC
Oops, meant to say that this is a report I got in email.
Comment 2 Sam Reed (reedy) 2012-02-16 18:43:35 UTC
Hide it behind miser mode?

But we still have
Comment 3 Sam Reed (reedy) 2012-02-16 20:53:28 UTC
Aaron looked at fixing it in 1.18wmf1 in r111667

I merged to trunk in r111673

Merged to 1.19 and 1.19wmf1 in r111675
Comment 4 Amalthea 2012-02-17 00:16:25 UTC
Live hack means it's live on enWP?
A first API call
didn't timeout anymore, but still took 33 seconds.
"only" took 15 seconds.
Comment 5 Sam Reed (reedy) 2012-02-17 00:20:42 UTC
Yeah, it's all live on the cluster
Comment 6 Antoine "hashar" Musso (WMF) 2012-03-01 13:30:42 UTC
With the above fix what prevents us from marking this bug as fixed? Is there anything else to fix?

Note You need to log in before you can comment on or make changes to this bug.