Re: [PATCH] memcg: introduce non-blocking limit setting interfaces

[Roman Gushchin <roman.gushchin@linux.dev>]

On Fri, Apr 18, 2025 at 01:30:03PM -0700, Shakeel Butt wrote:
> On Fri, Apr 18, 2025 at 01:18:53PM -0700, Greg Thelen wrote:
> > On Fri, Apr 18, 2025 at 1:00 PM Shakeel Butt <shakeel.butt@linux.dev> wrote:
> > >
> > > Setting the max and high limits can trigger synchronous reclaim and/or
> > > oom-kill if the usage is higher than the given limit. This behavior is
> > > fine for newly created cgroups but it can cause issues for the node
> > > controller while setting limits for existing cgroups.
> > >
> > > In our production multi-tenant and overcommitted environment, we are
> > > seeing priority inversion when the node controller dynamically adjusts
> > > the limits of running jobs of different priorities. Based on the system
> > > situation, the node controller may reduce the limits of lower priority
> > > jobs and increase the limits of higher priority jobs. However we are
> > > seeing node controller getting stuck for long period of time while
> > > reclaiming from lower priority jobs while setting their limits and also
> > > spends a lot of its own CPU.
> > >
> > > One of the workaround we are trying is to fork a new process which sets
> > > the limit of the lower priority job along with setting an alarm to get
> > > itself killed if it get stuck in the reclaim for lower priority job.
> > > However we are finding it very unreliable and costly. Either we need a
> > > good enough time buffer for the alarm to be delivered after setting
> > > limit and potentialy spend a lot of CPU in the reclaim or be unreliable
> > > in setting the limit for much shorter but cheaper (less reclaim) alarms.
> > >
> > > Let's introduce new limit setting interfaces which does not trigger
> > > reclaim and/or oom-kill and let the processes in the target cgroup to
> > > trigger reclaim and/or throttling and/or oom-kill in their next charge
> > > request. This will make the node controller on multi-tenant
> > > overcommitted environment much more reliable.
> > 
> > Would opening the typical synchronous files (e.g. memory.max) with
> > O_NONBLOCK be a more general way to tell the kernel that the user
> > space controller doesn't want to wait? It's not quite consistent with
> > traditional use of O_NONBLOCK, which would make operations to
> > fully succeed or fail, rather than altering the operation being requested.
> > But O_NONBLOCK would allow for a semantics of non-blocking
> > reclaim, if that's fast enough for your controller.

+1

> > 
> 
> We actually thought about O_NONBLOCK but the challenge with that is how
> would the node controller knows if the underlying kernel has O_NONBLOCK
> implying no-reclaim/no-oom-kill feature. I don't think opening
> memory.max with O_NONBLOCK will fail today, so the node controller would
> still need to implement the complicated fork+set-limit+alarm logic
> until the whole fleet has moved away from older kernel. Also I have
> checked with systemd folks and they are not happy to implement that
> complicated fork+set-limit+alarm logic.

/sys/kernel/cgroup/features ?