Submitting SLURM jobs with STDOUT & STDERR written to files wrt JOB_ID

Updated on September 17, 2021

There are different ways to submit SLURM jobs onto a cluster and the main way is by submitting a script with the sbatch command as shown below:

sbatch runcscript.sh

In my recent attempt to develop a web application with backend as SLURM scheduler for job management, referred SLURM documentation to redirect the STDOUT and STDERR to a file w.r.t to the job id as shown below:

#!/bin/bash
#SBATCH --job-name=qsim
#SBATCH --partition=standard-low
#SBATCH -o $SLURM_JOB_ID.output
#SBATCH -e $SLURM_JOB_ID.error
python UWVr6QCFKLGgx6sRtsnRZyRrajJdbPF4CsKGUqd7S4r.py

Unfortunately, the output didn’t get substituted with the variable $SLURM_JOB_ID.

$ls -lrt
-rw-rw-r-- 1 vivekn vivekn 737 Sep 17 12:43 $SLURM_JOB_ID.output
-rw-rw-r-- 1 vivekn vivekn 580 Sep 17 12:43 $SLURM_JOB_ID.error

After quite extensive research and the right guide from Harvard showed me that the way am referring was wrong, and had to change the SLURM Jobs script as below:

#!/bin/bash
#SBATCH --job-name=qsim
#SBATCH --partition=standard-low
#SBATCH -o %j.output
#SBATCH -e %j.error
python UWVr6QCFKLGgx6sRtsnRZyRrajJdbPF4CsKGUqd7S4r.py

The %j in the filename will be substituted by the JobID at runtime.

However, I am yet to find out why the SLURM Environment variable $SLURM_JOB_ID didn’t work. I Will update once I have an explanation. If you have any suggestions or answers please write in the comment section below.

Was this article helpful?

Related Articles

Leave a Comment