[slurm-users] How can I tell SLURM PowerUP to boot GPU Image

[J.R. W]

Hello,

I have an AWS burst cluster and I was wondering if there was any new information on if it’s possible to conditionally run different ResumeScripts or pass them an argument. I see from PoweSave documentation <https://slurm.schedmd.com/power_save.html>, the end calls for conditional AWS images to be booted with a “sophisticated” script that:


Determine(s) which jobs are associated with the nodes to be booted...

My thinking is that in your ResumeScript you would go through all the jobs in the configuring state, “squeue -t CF" maybe? Then asks which of those have asked for a GPU. You can then boot up GPU images based on how many GPU jobs are configuring.

Would this be considered best practice on how to conditionally boot an AWS image? Is there something I’m missing? Has anyone written a Resume script anywhere like this I can take a look at?

Thanks so much SLURM,
Jordan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20190118/fae697a1/attachment.html>