New feature: adapting to SLURM based clusters #83
joelfiddes
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
My use case is the the WSL hyperion cluster which uses a slurm based scheduling system. The issue is that nodes are not connected so cannot make use of dask easily (unlike the UIO server for example), There do seem to be ways but I think will need someone cleverer than me to figure that out:
https://jobqueue.dask.org/en/latest/generated/dask_jobqueue.SLURMCluster.html
My approach is to make it possible to distribute a big job (mine is most of the Tian Shan and some of the Pamir) by year to the cluster. This means a couple of things:
run_master.py
like this:
run_worker.py
like this:
Beta Was this translation helpful? Give feedback.
All reactions