https://bugzilla.wikimedia.org/show_bug.cgi?id=47534

--- Comment #19 from Bawolff (Brian Wolff) <[email protected]> ---
>btw, I believe ghostscript error code 9 is for invalid file access.

Actually nevermind, in guile, if you kill a subprocess executed with the
(system) command, the return status is the signal that killed it, so in this
case SIGKILL (signal 9), which is what would happen if ghostscript exceeded
resource limits when using cgroups. (Internally ghostscript also uses 9 to mean
invalid file access I believe, but I don't think it actually uses that as a
return code)

Which brings up the question of why is it only gs that is exceeding resource
limits. I would expect it to be other parts of the rendering process as well.
And specifically which limit is being exceeded. Is ghostscript really that
expensive? [I don't have cgroups set up locally, but using ulimits, the process
completes with much less memory then the limit is set to on wmf wikis]

To further add mystery, on test.wikipedia.org, this issue does not happen. I
occasionally got a return status from lilypond of 137 (128 + signal number 9
for bash signalling that it was killed, again presumably due to cgroups). In
general though rendering succeeds. Which is odd, since even though
test.wikipedia.org is not the same as a normal cluster wiki, afaik it has
resource limits configured in the same way [Unless it is perhaps configured as
an image scalar?]

[As an aside this provides a very poor workaround, since once you get something
rendered on test, it gets cached, and can be used on the other wikis]


-----
In case its later useful, it can be determined what the precise gs command
being executed by lilypond by addding -dverbose to the lilypond command line.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are on the CC list for the bug.
_______________________________________________
Wikibugs-l mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikibugs-l

Reply via email to