Refactor UpgradeStrategy to UpgradeSpec.Type

docs/reference/api.md

@@ -209,7 +209,7 @@ _Appears in:_
 | `serviceUnhealthySecondThreshold` _integer_ | Deprecated: This field is not used anymore. ref: https://github.com/ray-project/kuberay/issues/1685 |  |  |
 | `deploymentUnhealthySecondThreshold` _integer_ | Deprecated: This field is not used anymore. ref: https://github.com/ray-project/kuberay/issues/1685 |  |  |
 | `serveService` _[Service](https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.28/#service-v1-core)_ | ServeService is the Kubernetes service for head node and worker nodes who have healthy http proxy to serve traffics. |  |  |
-| `upgradeStrategy` _[RayServiceUpgradeStrategy](#rayserviceupgradestrategy)_ | UpgradeStrategy represents the strategy used when upgrading the RayService. Currently supports `NewCluster` and `None` |  |  |
+| `upgradeStrategy` _[RayServiceUpgradeStrategy](#rayserviceupgradestrategy)_ | UpgradeStrategy defines the scaling policy used when upgrading the RayService. |  |  |
 | `serveConfigV2` _string_ | Important: Run "make" to regenerate code after modifying this file<br />Defines the applications and deployments to deploy, should be a YAML multi-line scalar string. |  |  |
 | `rayClusterConfig` _[RayClusterSpec](#rayclusterspec)_ |  |  |  |
 | `excludeHeadPodFromServeSvc` _boolean_ | If the field is set to true, the value of the label `ray.io/serve` on the head Pod should always be false.<br />Therefore, the head Pod's endpoint will not be added to the Kubernetes Serve service. |  |  |
@@ -219,7 +219,7 @@ _Appears in:_
 
 #### RayServiceUpgradeStrategy
 
-_Underlying type:_ _string_
+
 
 
 
@@ -228,6 +228,22 @@ _Underlying type:_ _string_
 _Appears in:_
 - [RayServiceSpec](#rayservicespec)
 
+| Field | Description | Default | Validation |
+| --- | --- | --- | --- |
+| `type` _[RayServiceUpgradeType](#rayserviceupgradetype)_ | Type represents the strategy used when upgrading the RayService. Currently supports `NewCluster` and `None`. |  |  |
+
+
+#### RayServiceUpgradeType
+
+_Underlying type:_ _string_
+
+
+
+
+
+_Appears in:_
+- [RayServiceUpgradeStrategy](#rayserviceupgradestrategy)
+
 
 
 #### ScaleStrategy

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

@@ -20,13 +20,13 @@ const (
 	FailedToUpdateService            ServiceStatus = "FailedToUpdateService"
 )
 
-type RayServiceUpgradeStrategy string
+type RayServiceUpgradeType string
 
 const (
 	// During upgrade, NewCluster strategy will create new upgraded cluster and switch to it when it becomes ready
-	NewCluster RayServiceUpgradeStrategy = "NewCluster"
+	NewCluster RayServiceUpgradeType = "NewCluster"
 	// No new cluster will be created while the strategy is set to None
-	None RayServiceUpgradeStrategy = "None"
+	None RayServiceUpgradeType = "None"
 )
 
 // These statuses should match Ray Serve's application statuses
@@ -58,6 +58,11 @@ var DeploymentStatusEnum = struct {
 	UNHEALTHY: "UNHEALTHY",
 }
 
+type RayServiceUpgradeStrategy struct {
+	// Type represents the strategy used when upgrading the RayService. Currently supports `NewCluster` and `None`.
+	Type *RayServiceUpgradeType `json:"type,omitempty"`
+}
+
 // RayServiceSpec defines the desired state of RayService
 type RayServiceSpec struct {
 	// Deprecated: This field is not used anymore. ref: https://github.com/ray-project/kuberay/issues/1685
@@ -66,7 +71,7 @@ type RayServiceSpec struct {
 	DeploymentUnhealthySecondThreshold *int32 `json:"deploymentUnhealthySecondThreshold,omitempty"`
 	// ServeService is the Kubernetes service for head node and worker nodes who have healthy http proxy to serve traffics.
 	ServeService *corev1.Service `json:"serveService,omitempty"`
-	// UpgradeStrategy represents the strategy used when upgrading the RayService. Currently supports `NewCluster` and `None`
+	// UpgradeStrategy defines the scaling policy used when upgrading the RayService.
 	UpgradeStrategy *RayServiceUpgradeStrategy `json:"upgradeStrategy,omitempty"`
 	// Important: Run "make" to regenerate code after modifying this file
 	// Defines the applications and deployments to deploy, should be a YAML multi-line scalar string.

ray-operator/controllers/ray/rayservice_controller.go

@@ -247,10 +247,12 @@ func validateRayServiceSpec(rayService *rayv1.RayService) error {
 	if headSvc := rayService.Spec.RayClusterSpec.HeadGroupSpec.HeadService; headSvc != nil && headSvc.Name != "" {
 		return fmt.Errorf("spec.rayClusterConfig.headGroupSpec.headService.metadata.name should not be set")
 	}
-
-	if upgradeStrategy := rayService.Spec.UpgradeStrategy; upgradeStrategy != nil {
-		if *upgradeStrategy != rayv1.NewCluster && *upgradeStrategy != rayv1.None {
-			return fmt.Errorf("spec.UpgradeStrategy value %s is invalid, valid options are %s or %s", *upgradeStrategy, rayv1.NewCluster, rayv1.None)
+	if rayService.Spec.UpgradeSpec == nil {
+		return nil
+	}
+	if upgradeType := rayService.Spec.UpgradeSpec.Type; upgradeType != nil {
+		if *upgradeType != rayv1.NewCluster && *upgradeType != rayv1.None {
+			return fmt.Errorf("Spec.UpgradeSpec.Type value %s is invalid, valid options are %s or %s", *upgradeType, rayv1.NewCluster, rayv1.None)
 		}
 	}
 	return nil
@@ -430,24 +432,27 @@ func (r *RayServiceReconciler) reconcileRayCluster(ctx context.Context, rayServi
 		// For LLM serving, some users might not have sufficient GPU resources to run two RayClusters simultaneously.
 		// Therefore, KubeRay offers ENABLE_ZERO_DOWNTIME as a feature flag for zero-downtime upgrades.
 		zeroDowntimeEnvVar := os.Getenv(ENABLE_ZERO_DOWNTIME)
-		rayServiceSpecUpgradeStrategy := rayServiceInstance.Spec.UpgradeStrategy
-		// There are two ways to enable zero downtime upgrade. Through ENABLE_ZERO_DOWNTIME env var or setting Spec.UpgradeStrategy.
+		var rayServiceSpecUpgradeType *rayv1.RayServiceUpgradeStrategy
+		if rayServiceInstance.Spec.UpgradeSpec != nil {
+			rayServiceSpecUpgradeType = rayServiceInstance.Spec.UpgradeSpec.Type
+		}
+		// There are two ways to enable zero downtime upgrade. Through ENABLE_ZERO_DOWNTIME env var or setting Spec.UpgradeSpec.Type.
 		// If no fields are set, zero downtime upgrade by default is enabled.
-		// Spec.UpgradeStrategy takes precedence over ENABLE_ZERO_DOWNTIME.
+		// Spec.UpgradeSpec.Type takes precedence over ENABLE_ZERO_DOWNTIME.
 		enableZeroDowntime := true
-		strategyMessage := ""
+		typeMessage := ""
 		if zeroDowntimeEnvVar != "" {
 			enableZeroDowntime = strings.ToLower(zeroDowntimeEnvVar) != "false"
-			strategyMessage = fmt.Sprintf("ENABLE_ZERO_DOWNTIME environmental variable is set to %q", strings.ToLower(zeroDowntimeEnvVar))
+			typeMessage = fmt.Sprintf("ENABLE_ZERO_DOWNTIME environmental variable is set to %q", strings.ToLower(zeroDowntimeEnvVar))
 		}
-		if rayServiceSpecUpgradeStrategy != nil {
-			enableZeroDowntime = *rayServiceSpecUpgradeStrategy == rayv1.NewCluster
-			strategyMessage = fmt.Sprintf("Upgrade Strategy is set to %q", *rayServiceSpecUpgradeStrategy)
+		if rayServiceSpecUpgradeType != nil {
+			enableZeroDowntime = *rayServiceSpecUpgradeType == rayv1.NewCluster
+			typeMessage = fmt.Sprintf("Upgrade Type is set to %q", *rayServiceSpecUpgradeType)
 		}
 
 		if enableZeroDowntime || !enableZeroDowntime && activeRayCluster == nil {
 			if enableZeroDowntime && activeRayCluster != nil {
-				r.Recorder.Event(rayServiceInstance, "Normal", "UpgradeStrategy", strategyMessage)
+				r.Recorder.Event(rayServiceInstance, "Normal", "UpgradeSpec.Type", typeMessage)
 			}
 			// Add a pending cluster name. In the next reconcile loop, shouldPrepareNewRayCluster will return DoNothing and we will
 			// actually create the pending RayCluster instance.

ray-operator/controllers/ray/rayservice_controller_unit_test.go

@@ -50,10 +50,12 @@ func TestValidateRayServiceSpec(t *testing.T) {
 	var upgradeStrat rayv1.RayServiceUpgradeStrategy = "invalidStrategy"
 	err = validateRayServiceSpec(&rayv1.RayService{
 		Spec: rayv1.RayServiceSpec{
-			UpgradeStrategy: &upgradeStrat,
+			UpgradeSpec: &rayv1.RayServiceUpgradeSpec{
+				Type: &upgradeStrat,
+			},
 		},
 	})
-	assert.Error(t, err, "spec.UpgradeStrategy is invalid")
+	assert.Error(t, err, "spec.UpgradeSpec.Type is invalid")
 }
 
 func TestGenerateHashWithoutReplicasAndWorkersToDelete(t *testing.T) {
@@ -888,8 +890,9 @@ func TestReconcileRayCluster(t *testing.T) {
 				Recorder: record.NewFakeRecorder(1),
 			}
 			service := rayService.DeepCopy()
+			service.Spec.UpgradeSpec = &rayv1.RayServiceUpgradeSpec{}
 			if tc.rayServiceUpgradeStrategy != "" {
-				service.Spec.UpgradeStrategy = &tc.rayServiceUpgradeStrategy
+				service.Spec.UpgradeSpec.Type = &tc.rayServiceUpgradeStrategy
 			}
 			if tc.updateRayClusterSpec {
 				service.Spec.RayClusterSpec.RayVersion = "new-version"