Job killed: Always request 1 node for verification jobs
Hi @vagudets,
I am checking to use launch_SUNSET.sh in MN5.
I got the error:
/scratch/slurm/job13599118/slurm_script: line 16: 3025295 Killed Rscript ${script} ${atomic_recipe}
slurmstepd: error: Detected 1 oom_kill event in StepId=13599118.batch. Some of the step tasks have been OOM Killed.
I think the reason is that the jobs are running in two different nodes:
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
13599118 gpp SUNSET_v bsc03233 R 12:28 2 gs16r1b[55,58]
13599119 gpp SUNSET_v bsc03233 R 12:28 2 gs23r3b[67,71]
To avoid this behaviour, I added the line #SBATCH -N 1
to my sunset.sh that I was using before.
Could you please take a look at this? Is this line already included in the launcher?
Thanks a lot,
Núria