Re: [patch] memcg: add oom killer delay

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Tue, 8 Mar 2011 22:44:11 -0800 (PST)
David Rientjes <rientjes@google.com> wrote:

> On Wed, 9 Mar 2011, KAMEZAWA Hiroyuki wrote:
 > > Aside from my specific usecase for this tunable, let me pose a question: 
> > > do you believe that the memory controller would benefit from allowing 
> > > users to have a grace period in which to take one of the actions listed 
> > > above instead of killing something itself?  Yes, this would be possible by 
> > > setting and then unsetting memory.oom_control, but that requires userspace 
> > > to always be responsive (which, at our scale, we can unequivocally say 
> > > isn't always possible) and doesn't effectively deal with spikes in memory 
> > > that may only be temporary and doesn't require any intervention of the 
> > > user at all.
> > > 
> > 
> > Please add 'notifier' in kernel space and handle the event by kernel module.
> > It is much better than 'timeout and allow oom-kill again'.
> > 
> 
> A kernel-space notifier would certainly be helpful, but at what point does 
> the kernel choose to oom kill something?  If there's an oom notifier in 
> place, do we always defer killing or for a set period of time? 

For google, as you like.

For me, I want an oom-killall module Or oom-SIGSTOP-all module.
oom-killall will be useful for killing fork-bombs and very quick recovery.

For me, the 1st motivation of oom-disable is to taking core-dump of
memory leaking process and look into it for checking memory leak.
(panic_on_oom -> kdump is used for supporting my customer.)

Maybe my example of notifier user doesn't sounds good to you, 
please find a good one.


> If it's  the latter then we'll still want the timeout, otherwise there's no
> way to  guarantee we haven't killed something by the time userspace has a chance 
> to react to the notification.
> 

You can get a list of tasks in the cgroup and send SIGNALs with filters you like.
List of thread-IDs can be got easily with cgroup_iter_xxx functions.

Anyway, if you add notifier, please give us a user of it. If possible,
it should be a function which can never be implemented in userland even
with sane programmers, admins, and users.

For example, if all process's oom_score_adj was set to -1000 and oom-killer
doesn't work, do you implement a timeout ? I think you'll say it's a
wrong configuration. memcg's oom_disable timeout is the same thing.

Thanks,
-Kame





--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>