search

ray-project / kuberay

A toolkit to run Ray applications on Kubernetes
Apache License 2.0
1.15k stars 373 forks source link

[Feature] Rayjob GitHub Enterprise integration #2204

Closed HCharlie closed 3 months ago

HCharlie commented 3 months ago

Search before asking

Description

Rayjob take care of the ray cluster creation, code submission, and cluster decommision. My goal is to have a disposable way to create jobs without worrying about the cluster resource provision and decommision. In an ideal world, everytime I only need apply the rayjob manifest, I could get the result and it's done. However, in the official documentation, I see it's supporting mounting code as a configmap, which doesn't look like a scalable solution when the project contains complex file structures. It also supports using git clone from public Github inside the rayjob entrypoint. But I didn't find any information on GitHub Enterprise support/integration, so I would like to have this feature to support more enterprise use cases.

Use case

similar to https://github.com/ray-project/kuberay/issues/2200, I also come from an ML platform team, and we would like our users to clone their code from GHE in the rayjob entrypoint, and run the code from there.

Related issues

https://github.com/ray-project/kuberay/issues/2200

Are you willing to submit a PR?

andrewsykim commented 3 months ago

@HCharlie try the tips here? https://docs.ray.io/en/latest/ray-core/runtime_env_auth.html

HCharlie commented 3 months ago

nice! this looks like what I am looking for! thanks for the reply! :)

kevin85421 commented 3 months ago

I think this issue is resolved. Close this issue. @HCharlie feel free to reopen the issue if it is not resolved.