Re: [PATCH] mm: per-cgroup memory reclaim stats

[Roman Gushchin]

On Wed, May 17, 2017 at 08:03:03AM +1000, Balbir Singh wrote:
> On Sat, May 13, 2017 at 2:42 AM, Johannes Weiner <hannes@xxxxxxxxxxx> wrote:
> > On Fri, May 12, 2017 at 12:25:22PM +1000, Balbir Singh wrote:
> >>
> >> It sounds like memcg accumlates both global and memcg reclaim driver
> >> counts -- is this what we want?
> >
> > Yes.
> >
> > Consider a fully containerized system that is using only memory.low
> > and thus exclusively global reclaim to enforce the partitioning, NOT
> > artificial limits and limit reclaim. In this case, we still want to
> > know how much reclaim activity each group is experiencing.
> 
> But its also confusing to see memcg.stat's value being greater
> than the global value? At-least for me. For example PGSTEAL_DIRECT
> inside a memcg > global value of PGSTEAL_DIRECT. Do we make
> memcg.stat values sum of all impact on memcg or local to memcg?

Yes, I think that global counters should include both results of
global and per-cgroup reclaim. I will prepare a separate patch for this.

Thank you!

Roman