Re: [PATCH] memcg: handle panic_on_oom=always case

[Nick Piggin]

On Wed, Feb 17, 2010 at 03:04:45PM +0900, KAMEZAWA Hiroyuki wrote:
> tested on mmotm-Feb11.
> 
> Balbir-san, Nishimura-san, I want review from both of you.
> 
> ==
> 
> From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> 
> Now, if panic_on_oom=2, the whole system panics even if the oom happend
> in some special situation (as cpuset, mempolicy....).
> Then, panic_on_oom=2 means painc_on_oom_always.
> 
> Now, memcg doesn't check panic_on_oom flag. This patch adds a check.
> 
> Maybe someone doubts how it's useful. kdump+panic_on_oom=2 is the
> last tool to investigate what happens in oom-ed system. If a task is killed,
> the sysytem recovers and used memory were freed, there will be few hint
> to know what happnes. In mission critical system, oom should never happen.
> Then, investigation after OOM is very important.
> Then, panic_on_oom=2+kdump is useful to avoid next OOM by knowing
> precise information via snapshot.

No I don't doubt it is useful, and I think this probably is the simplest
and most useful semantic. So thanks for doing this.

I hate to pick nits in a trivial patch but I will anyway:


> TODO:
>  - For memcg, it's for isolate system's memory usage, oom-notiifer and
>    freeze_at_oom (or rest_at_oom) should be implemented. Then, management
>    daemon can do similar jobs (as kdump) in safer way or taking snapshot
>    per cgroup.
> 
> CC: Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx>
> CC: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> CC: David Rientjes <rientjes@xxxxxxxxxx>
> Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> ---
>  Documentation/cgroups/memory.txt |    2 ++
>  Documentation/sysctl/vm.txt      |    5 ++++-
>  mm/oom_kill.c                    |    2 ++
>  3 files changed, 8 insertions(+), 1 deletion(-)
> 
> Index: mmotm-2.6.33-Feb11/Documentation/cgroups/memory.txt
> ===================================================================
> --- mmotm-2.6.33-Feb11.orig/Documentation/cgroups/memory.txt
> +++ mmotm-2.6.33-Feb11/Documentation/cgroups/memory.txt
> @@ -182,6 +182,8 @@ list.
>  NOTE: Reclaim does not work for the root cgroup, since we cannot set any
>  limits on the root cgroup.
>  
> +Note2: When panic_on_oom is set to "2", the whole system will panic.
> +

Maybe:

NOTE2: When panic_on_oom is set to "2", the whole system will panic in
case of an oom event in any cgroup.

>  2. Locking
>  
>  The memory controller uses the following hierarchy
> Index: mmotm-2.6.33-Feb11/Documentation/sysctl/vm.txt
> ===================================================================
> --- mmotm-2.6.33-Feb11.orig/Documentation/sysctl/vm.txt
> +++ mmotm-2.6.33-Feb11/Documentation/sysctl/vm.txt
> @@ -573,11 +573,14 @@ Because other nodes' memory may be free.
>  may be not fatal yet.
>  
>  If this is set to 2, the kernel panics compulsorily even on the
> -above-mentioned.
> +above-mentioned. Even oom happens under memoyr cgroup, the whole
> +system panics.
                                           memory

>  
>  The default value is 0.
>  1 and 2 are for failover of clustering. Please select either
>  according to your policy of failover.
> +2 seems too strong but panic_on_oom=2+kdump gives you very strong
> +tool to investigate a system which should never cause OOM.

I don't think you need say 2 seems too strong because as you rightfully
say, it has real uses. The hint about using it to investigate OOM
conditions is good though.

>  
>  =============================================================
>  
> Index: mmotm-2.6.33-Feb11/mm/oom_kill.c
> ===================================================================
> --- mmotm-2.6.33-Feb11.orig/mm/oom_kill.c
> +++ mmotm-2.6.33-Feb11/mm/oom_kill.c
> @@ -471,6 +471,8 @@ void mem_cgroup_out_of_memory(struct mem
>  	unsigned long points = 0;
>  	struct task_struct *p;
>  
> +	if (sysctl_panic_on_oom == 2)
> +		panic("out of memory(memcg). panic_on_oom is selected.\n");
>  	read_lock(&tasklist_lock);
>  retry:
>  	p = select_bad_process(&points, mem);
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/