On Thu, Jun 6, 2019 at 10:44 PM Michal Hocko <mhocko@suse.com> wrote:

> On Thu 06-06-19 22:18:41, Yafang Shao wrote:
> [...]
> > Well, seems when we introduce new feature for page relciam, we always
> > ignore the node reclaim path.
>
> Yes, node reclaim is quite weird and I am not really sure whether we
> still have many users these days. It used to be mostly driven by
> artificial benchmarks which highly benefit from the local node access.
> We have turned off its automatic enabling when there are nodes with
> higher access latency quite some time ago without anybody noticing
> actually.
>
> > Regarding node reclaim path, we always turn it off on our servers,
> > because we really found some latency spike caused by node reclaim
> > (the reason why node reclaim is turned on is not clear).
>
> Yes, that was the case and the reason it is not enabled by default.
>
> > The reason I expose node reclaim details to userspace is because the user
> > can set node reclaim details now.
>
> Well, just because somebody _can_ enable it doesn't sound like a
> sufficient justification to expose even more implementation details of
> this feature. I am not really sure there is a strong reason to touch the
> code without a real usecase behind.
>
>
Got it.

So should we fix the bugs in node reclaim path then?

Thanks
Yafang