[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Tim Starling tstarl...@wikimedia.org changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #8 from Tim Starling tstarl...@wikimedia.org --- All done now. -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 --- Comment #7 from Faidon Liambotis fai...@wikimedia.org --- Tim, I know you were in the process of doing some work related to this. Can we get your opinion on Jan's patches either here or in Gerrit? I'd like to push this forward, let me know if I can help coordinating this in a better way. -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Andre Klapper aklap...@wikimedia.org changed: What|Removed |Added Priority|Normal |High CC||aklap...@wikimedia.org -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Antoine hashar Musso has...@free.fr changed: What|Removed |Added CC|has...@free.fr | -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 --- Comment #6 from Jan Gerber jger...@wikimedia.org --- pushed a patch to use cgroup memory limits via ulimit4.sh if setup: core: https://gerrit.wikimedia.org/r/#/c/40785/ puppet: https://gerrit.wikimedia.org/r/#/c/40784/ -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 --- Comment #5 from Faidon Liambotis fai...@wikimedia.org --- cgroups are nice and I agree that it'd be nice to use them. We can even use LXC, which sits on top of them and provides an easy way to run processes better isolated. However, I also think that a timeout mechanism should be employed to kill processes that hang, for whatever reason that can't be caught otherwise (e.g. maybe they're just sitting idle, not consuming memory or CPU). If we really feel like overengineering, we can even use cgroups to properly kill the whole group, including potentially runaway processes :) Jan, from a quick glance timelimit is basically signal handlers and alarm(), like what I suggested earlier. My idea was writing that in PHP, but using that helper might be easier indeed. -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 --- Comment #3 from Jan Gerber jger...@wikimedia.org --- one simple addition might be to use something like timelimit[1] to enforce a real time limit not a cpu time limit, that way processes that hang and don't use any cpu anymore would still get killed. [1] http://devel.ringlet.net/sysutils/timelimit/ -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Tim Starling tstarl...@wikimedia.org changed: What|Removed |Added CC||tstarl...@wikimedia.org --- Comment #4 from Tim Starling tstarl...@wikimedia.org --- Cgroups are a fairly nice option for memory limiting. With a little bit of per-server setup, an unprivileged user can create a cgroup dynamically for each convert/avconv process. It measures memory in an intelligent way, better than RSS and much better than vsize. And when the limit is reached, the kernel OOM killer kills it, so there's no chance of a deadlock. -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Andre Klapper aklap...@wikimedia.org changed: What|Removed |Added Keywords||performance Priority|Unprioritized |Normal Version|unspecified |1.21-git Severity|normal |enhancement -- You are receiving this mail because: You are the assignee for the bug. You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 --- Comment #1 from Faidon Liambotis fai...@wikimedia.org --- #40099 (AppArmor profile for avconv) is also related to this. The AppArmor idea is nice and could be expanded to image scalers too. However, note that it probably won't be enough, as it can't impose resource limits. -- You are receiving this mail because: You are the assignee for the bug. You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Rob Lanphier ro...@wikimedia.org changed: What|Removed |Added CC||ro...@wikimedia.org Assignee|wikibugs-l@lists.wikimedia. |jger...@wikimedia.org |org | -- You are receiving this mail because: You are the assignee for the bug. You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l
[Bug 43188] Enhance imagescaler processes containment
https://bugzilla.wikimedia.org/show_bug.cgi?id=43188 Antoine hashar Musso has...@free.fr changed: What|Removed |Added CC||has...@free.fr --- Comment #2 from Antoine hashar Musso has...@free.fr --- Would it make sense to run such huge scaling process on a dedicated box that will run them one at a time and thus with a lot more memory? We could probably get MediaWiki to catch such failing processes and create a dedicated job queue that will be process by such server. Of course if scaling fail on that one, we would abandon. -- You are receiving this mail because: You are watching all bug changes. ___ Wikibugs-l mailing list Wikibugs-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikibugs-l