I love snakemake and have used it locally as well as on HPC with SLURM! However, now we have a particular setup where it is not as easy to use snakemake as we have done before: We need to run some jobs on different HPCs and then sync get the results from one to another. An additional complication is that one of the HPCs has its outgoing internet access blocked. How can we best handle this? Our idea was to run snakemake itself on a different machine and only submit the tasks to the HPC. Is that possible? I did check the plugins, but did only see remote execution for particular platforms like AWS and not 'simply' another HPC. Do you think it would be possible to modify e.g. the snakemake-executor-plugin-aws-batch so that it can be used with a 'normal' HPC? Also what would it take in addition to run jobs via SLURM on that remote platform then? I tried modifying the SLURM plugin, prefixing all the slurm commands with ssh, but that lead to the problem snakemake trying to run python with my local python path... Am I approaching this wrong? Thanks for any ideas! Continue reading...
