Re: [systemd-devel] cgroup use case for preventing a system lock up if say a browser exhausts memory
On Mon, 12.11.12 12:54, Kai Hendry (hen...@iki.fi) wrote: Hi there, I hope this is the right forum for a systemd question, to address a particular problem I'm trying to solve. Problem is I've found Firefox (for example, this probably can be extrapolated to any browser) to lock up machines when memory is exhausted. Ideally when the browser hits a certain limit (only 5% of free RAM left), Firefox is killed and restarted, to solve my particular kiosk use case. https://github.com/Webconverger/webc/issues/83 It might be a good idea to simply bump the OOM score for firefox, so that it is always killed first in case of OOM. You can achieve that by wrapping ffox in shell script that first echos somthing to /proc/self/oom_score_adj, and then execs the real ffox binary. For now, systemd can do the same only for system services really. Lennart -- Lennart Poettering - Red Hat, Inc. ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] cgroup use case for preventing a system lock up if say a browser exhausts memory
On 12 November 2012 19:35, Colin Guthrie gm...@colin.guthr.ie wrote: When launching firefox becomes a user unit, then it will be contained within it's own cgroup and can have resource limits imposed with relative ease. Thanks Colin for your reply. I'm interested in a general systemd framework example of a system process resource limited. Do you have one you can point me to? What's important for me here, since this can run on machines with varying amount of memory that it's basically restarted when it's nearing memory exhaustion. Tbh I think the OOM killer should just do this, but I gather it's more complicated than that. Kind regards, ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] cgroup use case for preventing a system lock up if say a browser exhausts memory
Am 13.11.2012 07:53, schrieb Kai Hendry: I'm interested in a general systemd framework example of a system process resource limited. Do you have one you can point me to? http://0pointer.de/blog/projects/resources.html signature.asc Description: OpenPGP digital signature ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
Re: [systemd-devel] cgroup use case for preventing a system lock up if say a browser exhausts memory
'Twas brillig, and Kai Hendry at 12/11/12 04:54 did gyre and gimble: Hi there, I hope this is the right forum for a systemd question, to address a particular problem I'm trying to solve. Problem is I've found Firefox (for example, this probably can be extrapolated to any browser) to lock up machines when memory is exhausted. Ideally when the browser hits a certain limit (only 5% of free RAM left), Firefox is killed and restarted, to solve my particular kiosk use case. https://github.com/Webconverger/webc/issues/83 I've looked and tested varying unsatisfactory solutions and since I'm an Archlinux user who has been put onto systemd (tbh I'm impressed), I was eager to know if systemd can help. I can't help but notice how busy my `mount` looks now: http://ix.io/3nt Yeah, topic discussed/flamed to death here to be honest! The answer this is to make the tools more intelligent - e.g. tools like df and similar to avoid showing the same device multiple times if it's mounted multiple times in different paths etc. I've done a quick couple of searches trying to find how to use cgroups for controlling limits with systemd and I'm still in the dark. I was half expecting a service file example to have a resource limit stanza. Am I looking in the right place? Many thanks, Well, I suspect systemd can help eventually here, but likely not yet. The reason is that Firefox is just one process that runs in the user session, not the system. systemd is designed in such a way to eventually replace the home-grown user session implementations (e.g. gnome-session etc.) as the problems faced and the solutions for them are all rather similar. When launching firefox becomes a user unit, then it will be contained within it's own cgroup and can have resource limits imposed with relative ease. When this happens I think you'll have a good framework to implement the needed safeguards. See Auke's work on systemd for the user session for some more background (including his talk at this years Plumbers conference). HTHs Col -- Colin Guthrie gmane(at)colin.guthr.ie http://colin.guthr.ie/ Day Job: Tribalogic Limited http://www.tribalogic.net/ Open Source: Mageia Contributor http://www.mageia.org/ PulseAudio Hacker http://www.pulseaudio.org/ Trac Hacker http://trac.edgewall.org/ ___ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel