Deployment to on-premise Cluster

How severe does this issue affect your experience of using Ray?

  • High: It blocks me to complete my task.

Hi there,
We have an on-premise Ray Cluster running with Windows machines. I’m able to deploy deployments to the cluster using the python API. I ray.init into the Cluster head (using runtime_env “py_modules” : “some custom python scripts”), define a deployment and deploy. That works fine but only as long as the machine from where I deployed stays connected to the Ray Cluster. If ray is stopped on that machine, the actors die. I know that it is possible to deploy an app via serve deploy and a config file but that only works with remote URIs (for py_modules) which is not an option as our system is not connected to the internet.
Do you have any suggestion ?