Hi Andreas, The Slurm controller has job names in its own data structure.
The Slurm controller job name is set to the name of the job script by default, or to a name you define for example with -J. The environment variable SLURM_JOB_NAME is then set to this Slurm controller job name by job start. So there are two job names you can change for a running job, depending on what you want to do more exactly: scontrol update job=$SLURM_JOB_ID name=myname - Updates the Slurm controller job name, in the Slurm controller's data structure. - Controller calls like squeue will use this new name. For example squeue -n myname will list all jobs with this name. - Doesn't update SLURM_JOB_NAME. So the name that is presented by slurmctld can be changed back and forth without messing up the environment variable. export SLURM_JOB_NAME=myname - Updates the environment variable SLURM_JOB_NAME, just like you did. - The current job's process will get this new environment variable. You can refer to this new name (in code/child processes after the change) using $SLURM_JOB_NAME. - Doesn't update the Slurm controller job name. Best regards, Jessica Nettelblad, UPPMAX On Thu, Mar 22, 2018 at 10:16 PM, Andreas Hilboll < hilboll+sl...@uni-bremen.de> wrote: > Hi, > > I'd like to be able to set the SLURM_JOB_NAME from within the script I'm > submitting to `sbatch`. So, e.g., with the script `myscript.sh`, > > #!/bin/bash > export SLURM_JOB_NAME='myname' > sleep 120 > > and then `sbatch myscript.sh`, I'd like the job's name to be 'myname'. > > Is this somehow possible? When I tried the above approach, it didn't work > (squeue reported the job's name to be 'myscript.sh'). > > Thanks for your help, > Andreas > >
