[TPU Webhook] Fix KubeRay headless worker svc truncation bug

[ryanaoleary]

This PR fixes a bug where the KubeRay RayCluster controller's truncation of generated service names (here) resulted in incorrectly generated TPU_WORKER_HOSTNAMES. This bug was discovered because RayJobs and RayServices with names > 13 chars result in long generated RayCluster names and a truncated TPU headless worker service name. This PR applies the same truncation when generating the name within the webhook so that the names remain consistent.

Related Issues:
ray-project/kuberay#2923

Testing Process:

• Unit tests - added unit test coverage
• Manual tests

[ryanaoleary]

Manual testing process:

• created a RayCluster with the following spec:

# This template contains a Kuberay cluster using a 2x2x1 TPU v4 PodSlice.
# To get access to TPU resources, please follow instructions in this link:
# https://cloud.google.com/kubernetes-engine/docs/how-to/tpus
apiVersion: ray.io/v1
kind: RayCluster
metadata:
  # Label required for TPU webhook to initialize environments.
  labels:
    app.kubernetes.io/name: kuberay
  name: extremely-long-test-raycluster-name
spec:
  headGroupSpec:
    rayStartParams:
      {}
    template:
      spec:
        imagePullSecrets:
          []
        containers:
          - volumeMounts:
            - mountPath: /tmp/ray
              name: ray-logs
            name: ray-head
            image: rayproject/ray:2.9.0-py310
            imagePullPolicy: IfNotPresent
            resources:
              limits:
                cpu: "8"
                ephemeral-storage: 20Gi
                memory: 40G
              requests:
                cpu: "8"
                ephemeral-storage: 10Gi
                memory: 40G
            securityContext:
              {}
            env:
              - name: RAY_memory_monitor_refresh_ms
                value: "0"
              - name: RAY_GRAFANA_IFRAME_HOST
                value: http://${grafana_host}
              - name: RAY_GRAFANA_HOST
                value: http://grafana:80
              - name: RAY_PROMETHEUS_HOST
                value: http://frontend:9090
            ports:
              - containerPort: 6379
                name: gcs
              - containerPort: 8265
                name: dashboard
              - containerPort: 10001
                name: client
              - containerPort: 8000
                name: serve
        volumes:
          - emptyDir: {}
            name: ray-logs
      metadata:
        labels:
          cloud.google.com/gke-ray-node-type: head
          app.kubernetes.io/name: kuberay
          app.kubernetes.io/instance: example-cluster

  workerGroupSpecs:
  - rayStartParams:
      {}
    replicas: 1
    minReplicas: 1
    maxReplicas: 1
    numOfHosts: 2
    groupName: workergroup
    template:
      spec:
        imagePullSecrets:
          []
        containers:
          - volumeMounts:
            - mountPath: /tmp/ray
              name: ray-logs
            name: ray-worker
            image: rayproject/ray:2.9.0-py310
            imagePullPolicy: IfNotPresent
            resources:
              limits:
                cpu: "1"
                ephemeral-storage: 20Gi
                google.com/tpu: "4"
                memory: 40G
              requests:
                cpu: "1"
                ephemeral-storage: 10Gi
                google.com/tpu: "4"
                memory: 40G
            securityContext:
              {}
            env:
            ports:
              null
        volumes:
          - emptyDir: {}
            name: ray-logs
        nodeSelector:
          cloud.google.com/gke-tpu-accelerator: tpu-v4-podslice
          cloud.google.com/gke-tpu-topology: 2x2x2
      metadata:
        labels:
          cloud.google.com/gke-ray-node-type: worker
          app.kubernetes.io/name: kuberay
          app.kubernetes.io/instance: example-cluster

Checked headless service name created for multi-host TPU group:

(myenv) ryanaoleary@ryanaoleary1:~/Desktop/forks/kuberay/ray-operator/config/samples$ kubectl get svc
NAME                                                 TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)                                         AGE
extremely-long-test-raycluster-name-head-svc         ClusterIP   None           <none>        10001/TCP,8265/TCP,6379/TCP,8080/TCP,8000/TCP   61s
kuberay-operator                                     ClusterIP   34.118.226.3   <none>        8080/TCP                                        7d2h
kubernetes                                           ClusterIP   34.118.224.1   <none>        443/TCP                                         26d
mely-long-test-raycluster-name-headless-worker-svc   ClusterIP   None           <none>        <none>                                          61s

Checked created Pods:

(myenv) ryanaoleary@ryanaoleary1:~/Desktop/forks/kuberay/ray-operator/config/samples$ kubectl get pods
NAME                                                           READY   STATUS    RESTARTS   AGE
extremely-long-test-raycluster-name-head-7jcqq                 0/1     Pending   0          47s
extremely-long-test-raycluster-name-workergroup-worker-cr5vh   0/1     Pending   0          46s
extremely-long-test-raycluster-name-workergroup-worker-kh7fs   0/1     Pending   0          47s
kuberay-operator-77d8944bfc-dfssn                              1/1     Running   0          5d8h

Verified TPU_WORKER_HOSTNAMES matches service name:

kubectl describe pods extremely-long-test-raycluster-name-workergroup-worker-kh7fs
...
      TPU_WORKER_HOSTNAMES:                 workergroup-0-0.mely-long-test-raycluster-name-headless-worker-svc,workergroup-0-1.mely-long-test-raycluster-name-headless-worker-svc
      TPU_WORKER_ID:                        0
      TPU_NAME:                             workergroup-0
...

[ryanaoleary]

cc: @andrewsykim