Page MenuHomePhabricator

Recentchanges purges cause lock wait timeout
Closed, ResolvedPublic

Description

Author: bugzilla_wikipedia_org.to.jamesd

Description:
I've observed the periodic deletes from recentchanges taking in excess of 200
seconds at only moderately busy times and lock wait timeouts have been reported
for en Wikipedia with recentchanges specified as the affected table. Perhaps
increasing their frequency would help, by reducing the work per update.

As an interim step I've increased the Wikimedia lock wait timeout value.


Version: 1.3.x
Severity: normal

Details

Reference
bz237

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 21 2014, 6:45 PM
bzimport set Reference to bz237.
bzimport added a subscriber: Unknown Object (MLST).

zigger wrote:

(Set MW Version to 1.3.1 based on report date and release history.)
AFAICT, 1.4 & 1.5 still do this on average every 1000 edits, so with constant
editing, an average of 1000 rows would be deleted each time along with the index
changes. v1.4 added another index.

Is this still an issue? Increasing the frequency seems like an easy change to
make, and 1000x seems like a very high factor for removing the overhead.

Closing bug as the frequency got increased + we now run
the biggest mediawiki site (en.wikipedia) under 1.6 ;)