scarlehoff/pyHepGrid

Untangle SLURM backend

Opened this issue · 0 comments

Currently the slurm backend contains quite a lot of NNLOJET specific content, as it does not use a runfile like nnlorun.py etc. Instead it directly writes a .sh file for submission, which is pre-tailored to NNLOJET runs (hardcoding socket details, executable command and so on).

This because one cannot pass additional arguments to sbatch like you can with a .jdl or .xrsl file, they have to be hardcoded in the .sh file at runtime. Ideally it should be generalised in order to allow all programs straightforward access to SLURM runs.

Low priority, as HEJ are not using SLURM systems currently.