Page MenuHomePhabricator

Zotero OOM
Closed, ResolvedPublic


Zotero just crashed after running out of memory.

Notes from IRC:

[14:08] <akosiaris> I think that's the first one zotero(3)(+0000000): HTTP POST noDoi=yesDoi&doi=10.2307%2F2458932 to
[14:08] <akosiaris> &format=refman&direct=true&singleCitation=true
[14:08] <akosiaris> that got an out of memory error
[14:08] <akosiaris> well, maybe that one
[14:08] <akosiaris> not sure with zotero logging
[14:09] <akosiaris> seems like it was submitted multiple times
[14:09] <akosiaris> right before the out of memory error

Event Timeline

GWicke assigned this task to Mvolz.
GWicke raised the priority of this task from to Medium.
GWicke updated the task description. (Show Details)
GWicke added a project: Citoid.
GWicke added a subscriber: GWicke.
GWicke edited subscribers, added: akosiaris; removed: Aklapper.
Mvolz changed the task status from Open to Stalled.Aug 8 2016, 4:31 PM
Mvolz removed Mvolz as the assignee of this task.
Mvolz added a subscriber: Mvolz.
Deskana added a subscriber: Deskana.

Whatever caused this problem two years ago has likely been fixed now.

Nope. All that has happened is that the infrastructure has now some safeguards to protect itself from it in 2617d4f0a71 and subsequently 57cd2eb7076. The other thing that has happened is that zotero, being unmaintained, can not use jessie but rather requires trusty and as a result lives in it's own infrastructure, minimizing ill effects to other services due to memory leaks. So it's mitigated but definitely not fixed.

Well, also, we haven't updated it in two years. See: T140539

So it might be fixed by now... but if it is we're not running it :).

Change 483703 had a related patch set uploaded (by Mvolz; owner: Mvolz):
[mediawiki/services/zotero@master] Update Zotero; fix OOM issues

Change 483703 merged by jenkins-bot:
[mediawiki/services/zotero@master] Update Zotero; potentially fixes OOM issues

Mvolz removed a project: Patch-For-Review.

Redeploy on the 17th seems to have gotten rid of the big peaks:

These were being caused by the fact that Zotero was loading large pdfs into memory. It now ignores these file types.