ReproNim / reproman

ReproMan (AKA NICEMAN, AKA ReproNim TRD3)
https://reproman.readthedocs.io
Other
24 stars 14 forks source link

Provide explicit job array "serialization" #558

Open yarikoptic opened 3 years ago

yarikoptic commented 3 years ago

As discovered in #550 slurm on tacc seems to not allow (with obscure error message) batch array specification and uses Launcher (see #557 ) which was developed years ago, and possibly even not for SLURM initially (LSF?). Since apparently it could be quite common on well established HPCs, we should probably provide similar to launcher functionality to "serialize" job array execution where we would

either

and we would need to provide "aggregation" helper to monitor "overall job", not individual ones