update

ray-operator/controllers/ray/raycluster_controller.go

@@ -754,8 +754,8 @@ func (r *RayClusterReconciler) createWorkerPod(ctx context.Context, instance ray
 // Build head instance pod(s).
 func (r *RayClusterReconciler) buildHeadPod(instance rayv1alpha1.RayCluster) corev1.Pod {
 	podName := strings.ToLower(instance.Name + common.DashSymbol + string(rayv1alpha1.HeadNode) + common.DashSymbol)
-	podName = utils.CheckName(podName)                                            // making sure the name is valid
-	fqdnRayIP := utils.GenerateFQDNServiceName(instance.Name, instance.Namespace) // Fully Qualified Domain Name
+	podName = utils.CheckName(podName)                                       // making sure the name is valid
+	fqdnRayIP := utils.GenerateFQDNServiceName(instance, instance.Namespace) // Fully Qualified Domain Name
 	// The Ray head port used by workers to connect to the cluster (GCS server port for Ray >= 1.11.0, Redis port for older Ray.)
 	headPort := common.GetHeadPort(instance.Spec.HeadGroupSpec.RayStartParams)
 	autoscalingEnabled := instance.Spec.EnableInTreeAutoscaling
@@ -787,8 +787,8 @@ func getCreator(instance rayv1alpha1.RayCluster) string {
 // Build worker instance pods.
 func (r *RayClusterReconciler) buildWorkerPod(instance rayv1alpha1.RayCluster, worker rayv1alpha1.WorkerGroupSpec) corev1.Pod {
 	podName := strings.ToLower(instance.Name + common.DashSymbol + string(rayv1alpha1.WorkerNode) + common.DashSymbol + worker.GroupName + common.DashSymbol)
-	podName = utils.CheckName(podName)                                            // making sure the name is valid
-	fqdnRayIP := utils.GenerateFQDNServiceName(instance.Name, instance.Namespace) // Fully Qualified Domain Name
+	podName = utils.CheckName(podName)                                       // making sure the name is valid
+	fqdnRayIP := utils.GenerateFQDNServiceName(instance, instance.Namespace) // Fully Qualified Domain Name
 	// The Ray head port used by workers to connect to the cluster (GCS server port for Ray >= 1.11.0, Redis port for older Ray.)
 	headPort := common.GetHeadPort(instance.Spec.HeadGroupSpec.RayStartParams)
 	autoscalingEnabled := instance.Spec.EnableInTreeAutoscaling

ray-operator/controllers/ray/utils/dashboard_httpclient.go

@@ -69,6 +69,9 @@ type RayDashboardClient struct {
 func FetchHeadServiceURL(ctx context.Context, log *logr.Logger, cli client.Client, rayCluster *rayv1alpha1.RayCluster, defaultPortName string) (string, error) {
 	headSvc := &corev1.Service{}
 	headSvcName := GenerateServiceName(rayCluster.Name)
+	if rayCluster.Spec.HeadGroupSpec.HeadService != nil && rayCluster.Spec.HeadGroupSpec.HeadService.Name != "" {
+		headSvcName = rayCluster.Spec.HeadGroupSpec.HeadService.Name
+	}
 	if err := cli.Get(ctx, client.ObjectKey{Name: headSvcName, Namespace: rayCluster.Namespace}, headSvc); err != nil {
 		if errors.IsNotFound(err) {
 			log.Error(err, "Head service is not found", "head service name", headSvcName, "namespace", rayCluster.Namespace)

ray-operator/controllers/ray/utils/util.go

@@ -135,8 +135,12 @@ func GenerateServiceName(clusterName string) string {
 }
 
 // GenerateFQDNServiceName generates a Fully Qualified Domain Name.
-func GenerateFQDNServiceName(clusterName string, namespace string) string {
-	return fmt.Sprintf("%s.%s.svc.%s", GenerateServiceName(clusterName), namespace, GetClusterDomainName())
+func GenerateFQDNServiceName(cluster rayv1alpha1.RayCluster, namespace string) string {
+	headSvcName := GenerateServiceName(cluster.Name)
+	if cluster.Spec.HeadGroupSpec.HeadService != nil && cluster.Spec.HeadGroupSpec.HeadService.Name != "" {
+		headSvcName = cluster.Spec.HeadGroupSpec.HeadService.Name
+	}
+	return fmt.Sprintf("%s.%s.svc.%s", headSvcName, namespace, GetClusterDomainName())
 }
 
 // ExtractRayIPFromFQDN extracts the head service name (i.e., RAY_IP, deprecated) from a fully qualified