Re: [RFC] [PATCH] memory controller statistics

[Balbir Singh <balbir-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>]

箕浦真 wrote:
> Takashi is AFK for a while; i'm replying for him as possible.
> 
>> Thanks for doing this. We are building containerstats for
>> per container statistics. It would be really nice to provide
>> the statistics using that interface. I am not opposed to
>> memory.stat, but Paul Menage recommends that one file has
>> just one meaningful value.
> 
> Thanks, we'll check it.  The interface is not important for
> us.
> 
>> The other thing is that could you please report all the
>> statistics in bytes, we are moving to that interface,
>> I've posted patches to do that. If we are going to push
>> a bunch of statistics in one file, please use a format
>> separator like
> 
>> name: value
> 
>>> YAMOMOTO Takshi
>>>
>>> todo: something like nr_active/inactive in /proc/vmstat.
>>>
> 
>> This would be really nice to add.
> 
> `something' here could be # of pages (in bytes according to
> your advice) on the global active (or inactive) list that
> are charged to a container, and/or # of pages on the
> per-container active/inactive list.  The latter is easy to
> implement but I'm afraid it's somewhat confusing for users.
> 

I think it would be useful for administrators to get a rough
idea of the working set (active bytes), to configure the size
of the container.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL