Messages in this thread | | | Date | Fri, 26 Feb 2010 14:47:52 +0900 | From | Daisuke Nishimura <> | Subject | Re: [RFC][PATCH 1/2] memcg: oom kill handling improvement |
| |
On Fri, 26 Feb 2010 14:23:39 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote: > On Fri, 26 Feb 2010 13:15:52 +0900 > Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp> wrote: > > > On Wed, 24 Feb 2010 16:59:21 +0900, KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> wrote: > > > These are dump of patches just for showing concept, what I want to do. > > > But not tested. please see if you have free time. (you can ignore ;) > > > > > > Anyway, this will HUNK to the latest mmotm, Kirill's work is merged. > > > > > > This is not related to David's work. I don't hesitate to rebase mine > > > to the mmotm if his one is merged, it's easy. > > > But I'm not sure his one goes to mm soon. > > > > > > 1st patch is for better handling oom-kill under memcg. > > It's bigger than I expected, but it basically looks good to me. > > > > BTW, do you think we need quick fix ? I can't think of a very easy/small fix > which is very correct... To be honest, yes. IMHO, casing global oom because of memcg's oom is a very bad behavior in the sence of resource isolation. But I agree it's hard to fix in simple way, so leave it as it is for now.. hmm.. I must admit that I've not done enough oom test under very high pressure.
Thanks, Daisuke Nishimura.
| |