Long term support

[mtwest2718]

While I would like to see this sort of API as part of the standard Slurm install, that doesn't seem to fit their unixy design philosophy. And currently this seems to be just the personal project of a few individuals.

What vision is there to make this a viable long term production product that would track changes in Slurm that system administrators would be comfortable including as a necessary part of their system configuration?

• Funding?
• Contributors?
• An institutional home?

[giovtorres]

Thank you for raising this issue. A few year back, I reached out to one of the original authors of PySlurm because I was working at an HPC site that used the Slurm scheduler. Over time, I added features to it as well as a build and test system. I haven't been working on HPC clusters for a while and have been depending more on contributors to keep the project up to date.

I feel the project needs to reach a certain point before it can be easier to maintain. There's also the challenge of having to backport features, as some HPC sites won't upgrade Slurm relatively often.

There is some on going work by other contributors to make the library more manageable and more Pythonic. The goal is break up the large extension file into smaller, easier to maintain, modules while defining an interface that is easy for the user to consume.

Contributors would be most valuable at this point, and folks can contribute in a variety of ways.

[carlosgmartin]

@giovtorres @tazend Is this project still actively maintained? Is there anything you currently need help with?

[tazend]

Hi @carlosgmartin

kind of, to be honest. I have not really continued work on other parts of the API for quite some time, like slurm database things, that are mostly only relevant to admins I think. The most important parts are implemented for users I believe: querying jobs (live and from the database), submitting jobs, checking node/partition status etc.

I will however continue to support users when they have questions, fix bugs, upgrade to a new major slurm version (24.05 has released a few days ago), or perhaps implement new features if there are enough people who'd like to see something specific in the library.

One thing that would be cool to have back working is the CI/CD, especially for building and pushing the documentation to pyslurm.github.io, which I have done manually in the past but is not really the way to go when trying to maintain multiple slurm versions.