On Mon 05-08-19 20:28:40, Yang Shi wrote:
> On Mon, Aug 5, 2019 at 7:32 AM Michal Hocko <mho...@kernel.org> wrote:
> >
> > On Fri 02-08-19 11:56:28, Yang Shi wrote:
> > > On Fri, Aug 2, 2019 at 2:35 AM Michal Hocko <mho...@kernel.org> wrote:
> > > >
> > > > On Thu 01-08-19 14:00:51, Yang Shi wrote:
> > > > > On Mon, Jul 29, 2019 at 11:48 AM Michal Hocko <mho...@kernel.org> 
> > > > > wrote:
> > > > > >
> > > > > > On Mon 29-07-19 10:28:43, Yang Shi wrote:
> > > > > > [...]
> > > > > > > I don't worry too much about scale since the scale issue is not 
> > > > > > > unique
> > > > > > > to background reclaim, direct reclaim may run into the same 
> > > > > > > problem.
> > > > > >
> > > > > > Just to clarify. By scaling problem I mean 1:1 kswapd thread to 
> > > > > > memcg.
> > > > > > You can have thousands of memcgs and I do not think we really do 
> > > > > > want
> > > > > > to create one kswapd for each. Once we have a kswapd thread pool 
> > > > > > then we
> > > > > > get into a tricky land where a determinism/fairness would be non 
> > > > > > trivial
> > > > > > to achieve. Direct reclaim, on the other hand is bound by the 
> > > > > > workload
> > > > > > itself.
> > > > >
> > > > > Yes, I agree thread pool would introduce more latency than dedicated
> > > > > kswapd thread. But, it looks not that bad in our test. When memory
> > > > > allocation is fast, even though dedicated kswapd thread can't catch
> > > > > up. So, such background reclaim is best effort, not guaranteed.
> > > > >
> > > > > I don't quite get what you mean about fairness. Do you mean they may
> > > > > spend excessive cpu time then cause other processes starvation? I
> > > > > think this could be mitigated by properly organizing and setting
> > > > > groups. But, I agree this is tricky.
> > > >
> > > > No, I meant that the cost of reclaiming a unit of charges (e.g.
> > > > SWAP_CLUSTER_MAX) is not constant and depends on the state of the memory
> > > > on LRUs. Therefore any thread pool mechanism would lead to unfair
> > > > reclaim and non-deterministic behavior.
> > >
> > > Yes, the cost depends on the state of pages, but I still don't quite
> > > understand what does "unfair" refer to in this context. Do you mean
> > > some cgroups may reclaim much more than others?
> >
> > > Or the work may take too long so it can't not serve other cgroups in time?
> >
> > exactly.
> 
> Actually, I'm not very concerned by this. In our design each memcg has
> its dedicated work (memcg->wmark_work), so the reclaim work for
> different memcgs could be run in parallel since they are *different*
> work in fact although they run the same function. And, We could queue
> them to a dedicated unbound workqueue which may have maximum 512 or
> scale with nr cpus active works. Although the system may have
> thousands of online memcgs, I'm supposed it should be rare to have all
> of them trigger reclaim at the same time.

I do believe that it might work for your particular usecase but I do not
think this is robust enough for the upstream kernel, I am afraid.

As I've said I am open to discuss an opt-in per memcg pro-active reclaim
(a kernel thread that belongs to the memcg) but it has to be a dedicated
worker bound by all the cgroup resource restrictions.

-- 
Michal Hocko
SUSE Labs

Reply via email to