Tweaked SLURM jobscript
This tackles the problem, that SLURM incorrectly assigned the job state COMPLETED
for jobs with conda and/or Jupyter commands that did not succeed. This triggered a misleading "JupyterLab has already been shutdown..." message. I have now set the set -e
option, so that the FAILED
job state is triggered correctly.
I have switched the order of the usage instructions and the specification of the default machine resources, which are also a bit larger now. Other things are mainly cosmetic.