Add 'updateStrategy' field to RayServiceSpec, allowing users to disab…

…le/enable zero-downtime update.

docs/reference/api.md

@@ -207,6 +207,7 @@ _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` _[UpgradeStrategy](#upgradestrategy)_ | UpgradeStrategy represents the strategy used when upgrading the RayService. Currently supports `zero-downtime` |  |  |
 | `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)_ |  |  |  |
 
@@ -245,6 +246,19 @@ _Appears in:_
 | `backoffLimit` _integer_ | BackoffLimit of the submitter k8s job. |  |  |
 
 
+#### UpgradeStrategy
+
+_Underlying type:_ _string_
+
+
+
+
+
+_Appears in:_
+- [RayServiceSpec](#rayservicespec)
+
+
+
 #### UpscalingMode
 
 _Underlying type:_ _string_

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

@@ -20,6 +20,12 @@ const (
 	FailedToUpdateService            ServiceStatus = "FailedToUpdateService"
 )
 
+type UpgradeStrategy string
+
+const (
+	BlueGreenUpgrade UpgradeStrategy = "BlueGreenUpgrade"
+)
+
 // These statuses should match Ray Serve's application statuses
 // See `enum ApplicationStatus` in https://sourcegraph.com/github.com/ray-project/ray/-/blob/src/ray/protobuf/serve.proto for more details.
 var ApplicationStatusEnum = struct {
@@ -57,6 +63,8 @@ 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 `zero-downtime`
+	UpgradeStrategy UpgradeStrategy `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.
 	ServeConfigV2  string         `json:"serveConfigV2,omitempty"`

ray-operator/controllers/ray/rayservice_controller.go

@@ -424,10 +424,15 @@ func (r *RayServiceReconciler) reconcileRayCluster(ctx context.Context, rayServi
 	if clusterAction == RolloutNew {
 		// 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
 		enableZeroDowntime := true
-		if s := os.Getenv(ENABLE_ZERO_DOWNTIME); strings.ToLower(s) == "false" {
-			enableZeroDowntime = false
+		if rayServiceSpecUpgradeStrategy != "" {
+			enableZeroDowntime = rayServiceSpecUpgradeStrategy == rayv1.BlueGreenUpgrade
+		} else if zeroDowntimeEnvVar != "" {
+			enableZeroDowntime = strings.ToLower(zeroDowntimeEnvVar) == "true"
 		}
+
 		if enableZeroDowntime || !enableZeroDowntime && activeRayCluster == nil {
 			// 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

@@ -768,6 +768,7 @@ func TestReconcileRayCluster(t *testing.T) {
 		enableZeroDowntime      bool
 		shouldPrepareNewCluster bool
 		updateKubeRayVersion    bool
+		zeroDowntimeSpecTrigger bool
 	}{
 		// Test 1: Neither active nor pending clusters exist. The `markRestart` function will be called, so the `PendingServiceStatus.RayClusterName` should be set.
 		"Zero-downtime upgrade is enabled. Neither active nor pending clusters exist.": {
@@ -815,12 +816,27 @@ func TestReconcileRayCluster(t *testing.T) {
 			updateKubeRayVersion:    true,
 			kubeRayVersion:          "new-version",
 		},
+		// Test 7: Zero downtime upgrade is enabled, but is enabled through the RayServiceSpec
+		"Zero-downtime upgrade enabled. The active cluster exist. Zero-downtime upgrade is triggered through RayServiceSpec.": {
+			activeCluster:           activeCluster.DeepCopy(),
+			updateRayClusterSpec:    true,
+			enableZeroDowntime:      true,
+			shouldPrepareNewCluster: true,
+			zeroDowntimeSpecTrigger: true,
+		},
 	}
 
 	for name, tc := range tests {
 		t.Run(name, func(t *testing.T) {
 			// Enable or disable zero-downtime upgrade.
 			defer os.Unsetenv(ENABLE_ZERO_DOWNTIME)
+			if tc.enableZeroDowntime {
+				if tc.zeroDowntimeSpecTrigger {
+					rayService.Spec.UpgradeStrategy = rayv1.BlueGreenUpgrade
+				} else {
+					os.Setenv(ENABLE_ZERO_DOWNTIME, "true")
+				}
+			}
 			if !tc.enableZeroDowntime {
 				os.Setenv(ENABLE_ZERO_DOWNTIME, "false")
 			}