Closed stemangiola closed 1 year ago
You could set up a controller for each cluster, then combine them in a controller group: https://wlandau.github.io/crew/articles/groups.html. You will still need to manually decide which controller to send each task, but the interface is a bit more seamless. If you want to make a decision on the fly, you can use the saturated()
method to check if a controller's workers are all busy: https://wlandau.github.io/crew/reference/crew_class_controller_group.html#method-crew_class_controller_group-saturated
Converting this issue to a discussion.
Prework
Proposal
I am not sure if this is already possible, but let's say I have two remote SLURM clusters that have mirrored file systems (same directories and file locations). Would it be possible to spread the pipeline load across those two URLs?
Thanks a lot.