Re: single aio thread is migrated crazily by scheduler

From: Phil Auld
Date: Wed Nov 20 2019 - 17:03:31 EST


Hi Peter,

On Wed, Nov 20, 2019 at 08:16:36PM +0100 Peter Zijlstra wrote:
> On Tue, Nov 19, 2019 at 07:40:54AM +1100, Dave Chinner wrote:
> > On Mon, Nov 18, 2019 at 10:21:21AM +0100, Peter Zijlstra wrote:
>
> > > We typically only fall back to the active balancer when there is
> > > (persistent) imbalance and we fail to migrate anything else (of
> > > substance).
> > >
> > > The tuning mentioned has the effect of less frequent scheduling, IOW,
> > > leaving (short) tasks on the runqueue longer. This obviously means the
> > > load-balancer will have a bigger chance of seeing them.
> > >
> > > Now; it's been a while since I looked at the workqueue code but one
> > > possible explanation would be if the kworker that picks up the work item
> > > is pinned. That would make it runnable but not migratable, the exact
> > > situation in which we'll end up shooting the current task with active
> > > balance.
> >
> > Yes, that's precisely the problem - work is queued, by default, on a
> > specific CPU and it will wait for a kworker that is pinned to that
>
> I'm thinking the problem is that it doesn't wait. If it went and waited
> for it, active balance wouldn't be needed, that only works on active
> tasks.

Since this is AIO I wonder if it should queue_work on a nearby cpu by
default instead of unbound.

>
> > specific CPU to dispatch it. We've already tested that queuing on a
> > different CPU (via queue_work_on()) makes the problem largely go
> > away as the work is not longer queued behind the long running fio
> > task.
> >
> > This, however, is not at viable solution to the problem. The pattern
> > of a long running process queuing small pieces of individual work
> > for processing in a separate context is pretty common...
>
> Right, but you're putting the scheduler in a bind. By overloading the
> CPU and only allowing the one task to migrate, it pretty much has no
> choice left.
>
> Anyway, I'm still going to have try and reproduce -- I got side-tracked
> into a crashing bug, I'll hopefully get back to this tomorrow. Lastly,
> one other thing to try is -next. Vincent reworked the load-balancer
> quite a bit.
>

I've tried it with the lb patch series. I get basically the same results.
With the high granularity settings I get 3700 migrations for the 30
second run at 4k. Of those about 3200 are active balance on stock 5.4-rc7.
With the lb patches it's 3500 and 3000, a slight drop.

Using the default granularity settings 50 and 22 for stock and 250 and 25.
So a few more total migrations with the lb patches but about the same active.


On this system I'm getting 100k migrations using 512 byte blocksize. Almost
all not active. I haven't looked into that closely yet but it's like 3000
per second looking like this:

...
64.19641 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19694 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19746 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19665 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19718 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19772 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19800 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19828 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19856 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19882 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19909 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19937 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.19967 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.19995 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.20023 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.20053 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.20079 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.20107 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.20135 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.20163 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
64.20192 386 386 kworker/15:1 sched_migrate_task fio/2784 cpu 15->19
64.20221 389 389 kworker/19:1 sched_migrate_task fio/2784 cpu 19->15
...

Which is roughly equal to the number if iops it's doing.

Cheers,
Phil

--