Skip to content

Reject conflicting healthcheck ports in the DRA kubelet-plugin daemonset #2979

Open

Description

@cdesiniotis

When ComputeDomains are enabled in the GPUCluster CR (draDriver.computeDomains.enabled=true), the kubelet-plugin daemonset will have two main containers (one for GPUs and one for ComputeDomains). Both containers run a gRPC healthcheck service that is checked by a liveness probe (see here and here). The GPUCluster controller should validate that different ports are configured for the two healthcheck services, and return an error if there is a collision.

Activity

  1. jramkissoon commented on Oct 8, 2026

    @jramkissoon

    Hello all. This is my first contribution to this project. If my submitted PR for this issue fix is incorrectly formatted please advise. I have been able to replicate this issue with a local test bed and in the same test environment, validate both good and bad configs in the fixed version proposed. Thanks

  2. added a commit that references this issue on Oct 29, 2026
    152cc10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions