address comments

Author: cchen777

ray-operator/apis/ray/v1/raycluster_types.go

@@ -173,7 +173,7 @@ const (
 
 const (
 	// HeadPodReady is added in a RayCluster when its Head Pod is ready for requests.
-	HeadPodReady RayClusterConditionType = "Ready"
+	HeadPodReady RayClusterConditionType = "HeadPodReady"
 	// RayClusterReplicaFailure is added in a RayCluster when one of its pods fails to be created or deleted.
 	RayClusterReplicaFailure RayClusterConditionType = "ReplicaFailure"
 )

ray-operator/controllers/ray/raycluster_controller.go

@@ -1209,8 +1209,18 @@ func (r *RayClusterReconciler) calculateStatus(ctx context.Context, instance *ra
 		if err != nil {
 			return nil, err
 		}
-		replicaHeadPodReadyCondition := utils.FindPodReadyCondition(headPod, rayv1.HeadPodReady)
-		meta.SetStatusCondition(&newInstance.Status.Conditions, replicaHeadPodReadyCondition)
+		// GetRayClusterHeadPod can return nil, nil when pod is not found, we handle it separately.
+		if headPod == nil {
+			meta.SetStatusCondition(&newInstance.Status.Conditions, metav1.Condition{
+				Type:    string(rayv1.HeadPodReady),
+				Status:  metav1.ConditionFalse,
+				Reason:  "HeadPodNotFound",
+				Message: "Head Pod not found",
+			})
+		} else {
+			replicaHeadPodReadyCondition := utils.FindPodReadyCondition(headPod, rayv1.HeadPodReady)
+			meta.SetStatusCondition(&newInstance.Status.Conditions, replicaHeadPodReadyCondition)
+		}
 	}
 
 	if newInstance.Spec.Suspend != nil && *newInstance.Spec.Suspend && len(runtimePods.Items) == 0 {

ray-operator/controllers/ray/rayservice_controller.go

@@ -1061,7 +1061,7 @@ func (r *RayServiceReconciler) reconcileServe(ctx context.Context, rayServiceIns
 
 	// check the latest condition of the head Pod to see if it is ready.
 	if features.Enabled(features.RayClusterStatusConditions) {
-		if meta.IsStatusConditionPresentAndEqual(rayClusterInstance.Status.Conditions, string(rayv1.HeadPodReady), metav1.ConditionFalse) {
+		if !meta.IsStatusConditionTrue(rayClusterInstance.Status.Conditions, string(rayv1.HeadPodReady)) {
 			logger.Info("The head Pod is not ready, requeue the resource event to avoid redundant custom resource status updates.")
 			return ctrl.Result{RequeueAfter: ServiceDefaultRequeueDuration}, false, nil
 		}