Re: [RFC PATCH v2 1/2] tick/nohz: Add a policy dependency for full dynticks

[Peng Wang]

On Thu, Oct 29, 2026 at 05:49:10PM +0100, Frederic Weisbecker wrote:
> Le Fri, Oct 30, 2026 at 12:02:53AM +0800, Peng Wang a 殂rit :
> > > Ok in that case, do you know about the current work in progress
> > > to make nohz_full opt-able through cpusets? It implies taking the
> > > target CPUs offline shortly but the policy is to be setup before VM
> > > is launched, right?
> > 
> > Not always. Waiman mentioned the cpuset work earlier; it could
> > cover selecting the policy before a new VM starts.
> > 
> > Some customers express a preference for host tick suppression
> > through the management interface when changing workloads within
> > a running VM. The host operator decides whether to allow
> > suppression on the CPUs backing that VM.
> 
> I think people should know ahead if they will run latency critical workloads
> or common workloads in their VMs. Because those are completely different.
> 
> > 
> > We want to support revisiting that decision while keeping those
> > CPUs online and the VM running, with nohz_full membership
> > unchanged. Could such an operation fit into the cpuset work?
> 
> If that is really necessary then I guess we can have a file to control that
> in cpuset once the nohz_full interface is ready. Though I'm a bit skeptical due
> to the overhead of user <-> kernel transitions and housekeeping.

Hi Frederic,

Thanks for clarifying.

For some of our database workloads, we have observed a net
benefit from nohz_full despite the transition and housekeeping
overheads. We understand that those overheads remain when the
tick is kept running.

We'll follow the cpuset work and see whether this control
would fit.

Thanks,
Peng

> 
> Thanks.
> 
> > 
> > Thanks,
> > Peng
> > 
> > > 
> > > Thanks.
> > > 
> > > -- 
> > > Frederic Weisbecker
> > > SUSE Labs
> 
> -- 
> Frederic Weisbecker
> SUSE Labs