Hi,

I just upgraded my cluster form 19.05 to 20.02.

Now, in the prolog/epilog scripts, the variables SLURM_JOB_GPUS, CUDA_VISIBLE_DEVICES and GPU_DEVICE_ORDINAL are missing.

I am setting access to the GPUs via cgroups.

The only variables in prolog available are
SLURMD_NODENAME
SLURM_CLUSTER_NAME
SLURM_CONF
SLURM_JOBID
SLURM_JOB_CONSTRAINTS
SLURM_JOB_GID
SLURM_JOB_ID
SLURM_JOB_PARTITION
SLURM_JOB_UID
SLURM_JOB_USER
SLURM_NODELIST
SLURM_SCRIPT_CONTEXT
SLURM_UID

I switched to a configless setup and moved the slurm controller host during the upgrade, but did not change any configuration.

Any chance to get the old variables back? I use them in my prolog scripts...

Thanks
Quirin

--
Quirin Lohr
Systemadministration
Technische Universität München
Fakultät für Informatik
Lehrstuhl für Bildverarbeitung und Künstliche Intelligenz

Boltzmannstrasse 3
85748 Garching

Tel. +49 89 289 17769
Fax +49 89 289 17757

quirin.l...@in.tum.de
www.vision.in.tum.de

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to