I submit a job in a slurm cluster and all output and error messages go into the same file. Is there a way I can separate the error messages from the output?
ORIGINAL AUTHOR of this ANSWER: toreliza
With SLURM, the default stdout and stderr are directed to the same file. The file has a .out extension and the filename is dependent on the job ID and array index, if applicable. The default can be changed in the preamble of the SLURM submit script (the βtopβ part where β#SBATCHβ is used). Below is a code fragment showing a redirect of stderr (-e
) separately from a redirect of stdout (-o
). The %j
will be replaced by the SLURM_JOB_ID
.
#SBATCH -e slurm-%j.err
#SBATCH -o slurm-%j.out