Open githuberj opened 1 month ago
Does the netrc approach work for gitpod?
You can check out https://gitlab.com/gitlab-org/gitlab/-/issues/350582, seems there is a solution there
I tried it but couldn't get it to work with kuberay so I went a step back and tried it on my local:
This code works
import ray
good_env = {"working_dir": "https://github.com/ray-project/serve_config_examples/archive/refs/heads/master.zip"}
# Specify a runtime environment for the entire Ray job
ray.init(runtime_env=good_env,logging_level="DEBUG")
# Create a Ray task, which inherits the above runtime env.
@ray.remote
def f():
# The function will have its working directory changed to its node's
# local copy of /tmp/runtime_env_working_dir.
return open("text_ml.py").read()
print(ray.get(f.remote()))
ray.shutdown()
But if I put the same code into a password protected gitlab repo I only get a RayTaskError(FileNotFoundError): ray::f() (pid=15935, ip=172.26.113.45) exception.
There is also no indication of something going wrong with the runtime environment in the logs. I tried setting NETRC environment variable and using the default location of ~/.netrc
Make sure you do chmod 600 ~/.netrc
. Also can you try wget https://gitlab.com/xxx.zip
first to make sure your netrc file is correct.
Description
We would like to utilize gitlab as our collaboration platform. We already utilize KubeRay and would like to enable users to execute RayJobs and RayServings in a secure fashion similar to the one provided for github in https://docs.ray.io/en/latest/ray-core/runtime_env_auth.html. We use a GitOps based approach and therefore would greatly appreciate this possibility.
I currently got it running through an initcotainer with SSH git-sync but I would greatly appreciate a nicer workflow.
I also found a few related issues:
Use case
A user of Kuberay can provide a valid authentication method (Personal access token, Group access token, Project access token, SSH key) and the operator and ray cluster can execute his code.