Open brcopeland opened 2 years ago
<8 hrs job on glean queue and resubmit for 3 cycles, just in case they got killed ... =>8 hrs job on home queue and increase memory/cores everytime when resubmit ... after maybe 2 resubmit on home report error ... something like that?
I think what you say is sensible but I think those are matters for the snakemake profile. For the pipeline itself I want it to have a good estimate of the walltime and leave the matters of the queue, reruns, etc. up to the profile.
This can be helpful to prioritize using the glean queue (for free) for smaller data files and avoid jobs being killed for larger data files.