facebookincubator / submitit

Python 3.8+ toolbox for submitting jobs to Slurm
MIT License
1.27k stars 120 forks source link

Submit Over SSH? #1711

Open JRJacoby opened 2 years ago

JRJacoby commented 2 years ago

Hi All,

Apologies if this is not the right forum for this sort of question. I'm wondering if there's a way to submit a function as a job to a cluster where the cluster must be accessed via ssh. So some sort of Python equivalent of ssh {cluster_address} sbatch {job_script}. I like my scripts to stay running and monitor the jobs and potentially do something if they fail, and I don't think our sysadmins appreciate long-running scripts on the cluster login node. So ideally the script doing the monitoring would be able to run on my local machine. Thanks for anything you might be able to point me to.

Best, John

bjudkewitz commented 1 year ago

+1 ! Being able to submit over SSH would be fabulous!

gwenzek commented 1 year ago

I agree that would be really useful, but nobody has started working on it yet.

things that would need to be done:

  1. prefix sbatch calls with ssh cluster
  2. prefix sacct calls with ssh cluster
  3. rewrite file reading/writing/copying to work across machines
  4. find a good strategy to synchronize local files with cluster files
YannDubs commented 1 year ago

+1 that would really be amazing! 💯

firstmover commented 1 year ago

+1 I like to write code on the local machine. submitit with ssh would be helpful.