On 15 August 2017 at 07:41, Robbert Eggermont <[email protected]>
wrote:

>
> On 14-08-17 07:50, Lachlan Musicman wrote:
>
>> We have TaskPlugin=task/cgroup and when testing I noticed that the # of
>> threads/cpus being allocated was rounded up to the nearest even. I presume
>> this was due to cgroups marking a core as a cpu, rather than a thread as a
>> cpu.
>>
>
> Sounds like you're using SelectTypeParameters=CR_Core[_Memory], where
> what you want requires SelectTypeParameters=CR_CPU[_Memory]?
>

No, we are using SelectTypeParameters=CR_CPU_Memory


> I was hoping that the cgroup would kill the job because of too many cpus
>>
>
> task/cgroup just limits the CPUs available to a job; a job can still have
> any number of threads. (The number of *running* threads is of course
> limited to the number of available CPUs.)
>

Ok, this is fine. I think I realised this late - as I read the stress-ng
man page and saw that the threads would just wait for the others to finish.

Is it possible to have granular control at the thread level with cgroups?
>>
>
> Not that I know. Limiting the maximum number of threads can be done via
> resource limits.
>


Interestingly, sinfo is also rounding up

sinfo -Nle -o "%.20n %.15C %.8O %.7t" | uniq
Tue Aug 15 11:05:59 2017
HOSTNAMES   CPUS(A/I/O/T) CPU_LOAD   STATE
compute34         4/4/0/8     6.42     mix
compute35         4/4/0/8     6.45     mix


Cheers
L.


------
"The antidote to apocalypticism is *apocalyptic civics*. Apocalyptic civics
is the insistence that we cannot ignore the truth, nor should we panic
about it. It is a shared consciousness that our institutions have failed
and our ecosystem is collapsing, yet we are still here — and we are
creative agents who can shape our destinies. Apocalyptic civics is the
conviction that the only way out is through, and the only way through is
together. "

*Greg Bloom* @greggish
https://twitter.com/greggish/status/873177525903609857

Reply via email to