[slurm-users] How to check the percent cpu of a job?

[Peter Kjellström]

On Thu, 22 Nov 2018 01:51:59 +0800 (GMT+08:00)
宋亚磊 <songyl11 at lzu.edu.cn> wrote:

> Hello everyone,
> 
> How to check the percent cpu of a job in slurm? I tried sacct, sstat,
> squeue, but I can't find that how to check. Can someone help me?

I've written a small tool, jobload, that takes a jobid and outputs
current per node cpu% and memory use.

Example for a two node job using 2x 32 core nodes but with only 12
active processes on each. Using 10G RAM:

 $ jobload 1234
 n161: cpu% 1205 of 3200   memG 10 of 90
 n160: cpu% 1206 of 3200   memG 10 of 90

The tool uses srun to collect cgroup usage stats from within the job.
Pro: it works even with shared nodes. Con: it requires the use of
cgroups for job containment.

/Peter K