#3885 [Core feature] Different resources for Ray head pod and worker pods Issue created by aybidi Motivation: Why do you think this is important? We are running ML workflows at my organization that often require spinning up a Ray cluster where the head nodes and worker nodes usually have different resources configuration. Using the Ray plugin, I could use the

limits

arg in the

@task

decorator to set resources for the task.

ray_config = RayJobConfig(
    head_node_config=HeadNodeConfig(...),
    worker_node_config=[WorkerNodeConfig(...),],
    runtime_env={...},
)
@task(task_config=ray_config, limits=Resources(mem="12Gi", cpu="4"))
def ray_task(...):
    ...

However, this way of setting resources sets the same resources for both the head pod and worker pods CR created by the backend Ray plugin. Goal: What should the final outcome look like, ideally? Users should be able to control the cluster specs at a more granular level i.e., specifying different specs for the head node and the worker nodes. Describe alternatives you've considered n/a Propose: Link/Inline OR Additional context No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte