feat: use official Grove 0.1.0-alpha release (#3030)

Signed-off-by: Julien Mancuso <jmancuso@nvidia.com>

deploy/cloud/helm/platform/Chart.yaml

@@ -35,11 +35,11 @@ dependencies:
     repository: "https://charts.bitnami.com/bitnami"
     condition: etcd.enabled
   - name: kai-scheduler
-    version: v0.8.4
+    version: v0.9.2
     repository: oci://ghcr.io/nvidia/kai-scheduler
     condition: kai-scheduler.enabled
   - name: grove-charts
     alias: grove
-    version: v0.0.0-6e30275
+    version: v0.1.0-alpha.1
     repository: oci://ghcr.io/nvidia/grove
     condition: grove.enabled

deploy/cloud/helm/platform/components/operator/templates/manager-rbac.yaml

@@ -119,7 +119,7 @@ rules:
 - apiGroups:
   - grove.io
   resources:
-  - podgangsets
+  - podcliquesets
   verbs:
   - create
   - delete

deploy/cloud/operator/cmd/main.go

@@ -159,7 +159,7 @@ func main() {
 	flag.StringVar(&ingressHostSuffix, "ingress-host-suffix", "",
 		"The suffix to use for the ingress host")
 	flag.DurationVar(&groveTerminationDelay, "grove-termination-delay", consts.DefaultGroveTerminationDelay,
-		"The termination delay for Grove PodGangSets")
+		"The termination delay for Grove PodCliqueSets")
 	flag.StringVar(&modelExpressURL, "model-express-url", "",
 		"URL of the Model Express server to inject into all pods")
 	flag.StringVar(&prometheusEndpoint, "prometheus-endpoint", "",

deploy/cloud/operator/config/rbac/role.yaml

@@ -110,7 +110,7 @@ rules:
 - apiGroups:
   - grove.io
   resources:
-  - podgangsets
+  - podcliquesets
   verbs:
   - create
   - delete

deploy/cloud/operator/go.mod

@@ -6,8 +6,9 @@ toolchain go1.24.3
 
 require (
 	emperror.dev/errors v0.8.1
-	github.com/NVIDIA/grove/operator/api v0.0.0-20250825164137-da01400261a6
+	github.com/NVIDIA/grove/operator/api v0.1.0-alpha.1
 	github.com/bsm/gomega v1.27.10
+	github.com/go-logr/logr v1.4.2
 	github.com/google/go-cmp v0.7.0
 	github.com/imdario/mergo v0.3.6
 	github.com/onsi/ginkgo/v2 v2.23.4
@@ -39,7 +40,6 @@ require (
 	github.com/evanphx/json-patch/v5 v5.9.11 // indirect
 	github.com/fsnotify/fsnotify v1.7.0 // indirect
 	github.com/fxamacker/cbor/v2 v2.7.0 // indirect
-	github.com/go-logr/logr v1.4.2 // indirect
 	github.com/go-logr/zapr v1.3.0 // indirect
 	github.com/go-openapi/jsonpointer v0.21.0 // indirect
 	github.com/go-openapi/jsonreference v0.21.0 // indirect

deploy/cloud/operator/go.sum

 emperror.dev/errors v0.8.1 h1:UavXZ5cSX/4u9iyvH6aDcuGkVjeexUGJ7Ij7G4VfQT0=
 emperror.dev/errors v0.8.1/go.mod h1:YcRvLPh626Ubn2xqtoprejnA5nFha+TJ+2vew48kWuE=
-github.com/NVIDIA/grove/operator/api v0.0.0-20250825164137-da01400261a6 h1:JkW8LeRVsQH/YkRTz80T/JxlDgfk0URKgTUKyYKxbso=
-github.com/NVIDIA/grove/operator/api v0.0.0-20250825164137-da01400261a6/go.mod h1:QlsR2wQLj9m/zVEqv5SsCPzyjN2ykYZ0r/NEnDf4WB4=
+github.com/NVIDIA/grove/operator/api v0.1.0-alpha.1 h1:4DE6ZGa/3muBa5gk1GtJskMVss6GjeCPpn+xTnR1h9w=
+github.com/NVIDIA/grove/operator/api v0.1.0-alpha.1/go.mod h1:QlsR2wQLj9m/zVEqv5SsCPzyjN2ykYZ0r/NEnDf4WB4=
 github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
 github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
 github.com/blang/semver/v4 v4.0.0 h1:1PFHFE6yCCTv8C1TeyNNarDzntLi7wMI5i/pzqYIsAM=

deploy/cloud/operator/internal/controller/dynamographdeployment_controller.go

@@ -71,7 +71,7 @@ type DynamoGraphDeploymentReconciler struct {
 // +kubebuilder:rbac:groups=nvidia.com,resources=dynamographdeployments,verbs=get;list;watch;create;update;patch;delete
 // +kubebuilder:rbac:groups=nvidia.com,resources=dynamographdeployments/status,verbs=get;update;patch
 // +kubebuilder:rbac:groups=nvidia.com,resources=dynamographdeployments/finalizers,verbs=update
-// +kubebuilder:rbac:groups=grove.io,resources=podgangsets,verbs=get;list;watch;create;update;patch;delete
+// +kubebuilder:rbac:groups=grove.io,resources=podcliquesets,verbs=get;list;watch;create;update;patch;delete
 // +kubebuilder:rbac:groups=grove.io,resources=podcliques/scale,verbs=get;update;patch
 // +kubebuilder:rbac:groups=grove.io,resources=podcliquescalinggroups/scale,verbs=get;update;patch
 // +kubebuilder:rbac:groups=scheduling.run.ai,resources=queues,verbs=get;list
@@ -258,12 +258,12 @@ func (r *DynamoGraphDeploymentReconciler) reconcileGroveScaling(ctx context.Cont
 func (r *DynamoGraphDeploymentReconciler) reconcileGroveResources(ctx context.Context, dynamoDeployment *nvidiacomv1alpha1.DynamoGraphDeployment) (State, Reason, Message, error) {
 	logger := log.FromContext(ctx)
 	// generate the dynamoComponentsDeployments from the config
-	groveGangSet, err := dynamo.GenerateGrovePodGangSet(ctx, dynamoDeployment, r.Config, r.DockerSecretRetriever)
+	groveGangSet, err := dynamo.GenerateGrovePodCliqueSet(ctx, dynamoDeployment, r.Config, r.DockerSecretRetriever)
 	if err != nil {
 		logger.Error(err, "failed to generate the Grove GangSet")
 		return "", "", "", fmt.Errorf("failed to generate the Grove GangSet: %w", err)
 	}
-	_, syncedGroveGangSet, err := commonController.SyncResource(ctx, r, dynamoDeployment, func(ctx context.Context) (*grovev1alpha1.PodGangSet, bool, error) {
+	_, syncedGroveGangSet, err := commonController.SyncResource(ctx, r, dynamoDeployment, func(ctx context.Context) (*grovev1alpha1.PodCliqueSet, bool, error) {
 		return groveGangSet, false, nil
 	})
 	if err != nil {
@@ -421,7 +421,7 @@ func (r *DynamoGraphDeploymentReconciler) SetupWithManager(mgr ctrl.Manager) err
 		})).
 		WithEventFilter(commonController.EphemeralDeploymentEventFilter(r.Config))
 	if r.Config.Grove.Enabled {
-		ctrlBuilder = ctrlBuilder.Owns(&grovev1alpha1.PodGangSet{}, builder.WithPredicates(predicate.Funcs{
+		ctrlBuilder = ctrlBuilder.Owns(&grovev1alpha1.PodCliqueSet{}, builder.WithPredicates(predicate.Funcs{
 			// ignore creation cause we don't want to be called again after we create the pod gang set
 			CreateFunc:  func(ce event.CreateEvent) bool { return false },
 			DeleteFunc:  func(de event.DeleteEvent) bool { return true },

deploy/cloud/operator/internal/controller_common/podgangset.go

@@ -6,7 +6,7 @@ import (
 	grovev1alpha1 "github.com/NVIDIA/grove/operator/api/core/v1alpha1"
 )
 
-func CanonicalizePodGangSet(gangSet *grovev1alpha1.PodGangSet) *grovev1alpha1.PodGangSet {
+func CanonicalizePodCliqueSet(gangSet *grovev1alpha1.PodCliqueSet) *grovev1alpha1.PodCliqueSet {
 	// sort cliques by name
 	sort.Slice(gangSet.Spec.Template.Cliques, func(i, j int) bool {
 		return gangSet.Spec.Template.Cliques[i].Name < gangSet.Spec.Template.Cliques[j].Name

deploy/cloud/operator/internal/controller_common/predicate.go

@@ -33,7 +33,7 @@ import (
 type GroveConfig struct {
 	// Enabled is automatically determined by checking if Grove CRDs are installed in the cluster
 	Enabled bool
-	// TerminationDelay configures the termination delay for Grove PodGangSets
+	// TerminationDelay configures the termination delay for Grove PodCliqueSets
 	TerminationDelay time.Duration
 }
 

deploy/cloud/operator/internal/dynamo/graph.go

@@ -317,16 +317,16 @@ type SecretsRetriever interface {
 	GetSecrets(namespace, registry string) ([]string, error)
 }
 
-// applyCliqueStartupDependencies configures StartsAfter dependencies for cliques in a PodGangSet
+// applyCliqueStartupDependencies configures StartsAfter dependencies for cliques in a PodCliqueSet
 // based on the backend framework and multinode deployment patterns.
 //
 // Rules:
 // - For VLLM and SGLang: worker cliques start after leader clique
 // - For TRTLLM: leader clique starts after worker cliques
 // - Only applies to multinode deployments (numberOfNodes > 1)
-// - Sets the PodGangSet StartupType to Explicit if any dependencies are configured
+// - Sets the PodCliqueSet StartupType to Explicit if any dependencies are configured
 func applyCliqueStartupDependencies(
-	gangSet *grovev1alpha1.PodGangSet,
+	gangSet *grovev1alpha1.PodCliqueSet,
 	roles []ServiceRole,
 	backendFramework BackendFramework,
 	numberOfNodes int32,
@@ -880,14 +880,14 @@ func GeneratePodSpecForComponent(
 	return podSpec, nil
 }
 
-// GenerateGrovePodGangSet generates a Grove PodGangSet for the given deployment, supporting both single-node and multinode cases.
-func GenerateGrovePodGangSet(
+// GenerateGrovePodCliqueSet generates a Grove PodCliqueSet for the given deployment, supporting both single-node and multinode cases.
+func GenerateGrovePodCliqueSet(
 	ctx context.Context,
 	dynamoDeployment *v1alpha1.DynamoGraphDeployment,
 	controllerConfig controller_common.Config,
 	secretsRetriever SecretsRetriever,
-) (*grovev1alpha1.PodGangSet, error) {
-	gangSet := &grovev1alpha1.PodGangSet{}
+) (*grovev1alpha1.PodCliqueSet, error) {
+	gangSet := &grovev1alpha1.PodCliqueSet{}
 	gangSet.Name = dynamoDeployment.Name
 	gangSet.Namespace = dynamoDeployment.Namespace
 	gangSet.Spec.Replicas = 1
@@ -986,7 +986,7 @@ func GenerateGrovePodGangSet(
 		gangSet.Spec.Template.PodCliqueScalingGroupConfigs = scalingGroups
 	}
 
-	return controller_common.CanonicalizePodGangSet(gangSet), nil
+	return controller_common.CanonicalizePodCliqueSet(gangSet), nil
 }
 
 func generateLabels(component *v1alpha1.DynamoComponentDeploymentOverridesSpec, dynamoDeployment *v1alpha1.DynamoGraphDeployment, componentName string) (map[string]string, error) {

deploy/cloud/operator/internal/dynamo/graph_test.go

@@ -1048,7 +1048,7 @@ func sortEnvVars(envs []corev1.EnvVar) []corev1.EnvVar {
 	return sorted
 }
 
-func TestGenerateGrovePodGangSet(t *testing.T) {
+func TestGenerateGrovePodCliqueSet(t *testing.T) {
 	type args struct {
 		ctx              context.Context
 		dynamoDeployment *v1alpha1.DynamoGraphDeployment
@@ -1057,11 +1057,11 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 	tests := []struct {
 		name    string
 		args    args
-		want    *grovev1alpha1.PodGangSet
+		want    *grovev1alpha1.PodCliqueSet
 		wantErr bool
 	}{
 		{
-			name: "test_generate_grove_pod_gang_set_single_node",
+			name: "test_generate_grove_pod_clique_set_single_node",
 			args: args{
 				ctx: context.Background(),
 				controllerConfig: controller_common.Config{
@@ -1220,14 +1220,14 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 					},
 				},
 			},
-			want: &grovev1alpha1.PodGangSet{
+			want: &grovev1alpha1.PodCliqueSet{
 				ObjectMeta: metav1.ObjectMeta{
 					Name:      "test-dynamo-graph-deployment",
 					Namespace: "test-namespace",
 				},
-				Spec: grovev1alpha1.PodGangSetSpec{
+				Spec: grovev1alpha1.PodCliqueSetSpec{
 					Replicas: 1,
-					Template: grovev1alpha1.PodGangSetTemplateSpec{
+					Template: grovev1alpha1.PodCliqueSetTemplateSpec{
 						StartupType: ptr.To(grovev1alpha1.CliqueStartupTypeAnyOrder),
 						HeadlessServiceConfig: &grovev1alpha1.HeadlessServiceConfig{
 							PublishNotReadyAddresses: true,
@@ -1737,14 +1737,14 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 					},
 				},
 			},
-			want: &grovev1alpha1.PodGangSet{
+			want: &grovev1alpha1.PodCliqueSet{
 				ObjectMeta: metav1.ObjectMeta{
 					Name:      "test-dynamo-graph-deployment",
 					Namespace: "test-namespace",
 				},
-				Spec: grovev1alpha1.PodGangSetSpec{
+				Spec: grovev1alpha1.PodCliqueSetSpec{
 					Replicas: 1,
-					Template: grovev1alpha1.PodGangSetTemplateSpec{
+					Template: grovev1alpha1.PodCliqueSetTemplateSpec{
 						HeadlessServiceConfig: &grovev1alpha1.HeadlessServiceConfig{
 							PublishNotReadyAddresses: true,
 						},
@@ -2533,14 +2533,14 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 					},
 				},
 			},
-			want: &grovev1alpha1.PodGangSet{
+			want: &grovev1alpha1.PodCliqueSet{
 				ObjectMeta: metav1.ObjectMeta{
 					Name:      "test-dynamo-graph-deployment",
 					Namespace: "test-namespace",
 				},
-				Spec: grovev1alpha1.PodGangSetSpec{
+				Spec: grovev1alpha1.PodCliqueSetSpec{
 					Replicas: 1,
-					Template: grovev1alpha1.PodGangSetTemplateSpec{
+					Template: grovev1alpha1.PodCliqueSetTemplateSpec{
 						StartupType: ptr.To(grovev1alpha1.CliqueStartupTypeAnyOrder),
 						HeadlessServiceConfig: &grovev1alpha1.HeadlessServiceConfig{
 							PublishNotReadyAddresses: true,
@@ -3099,9 +3099,9 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			got, err := GenerateGrovePodGangSet(tt.args.ctx, tt.args.dynamoDeployment, tt.args.controllerConfig, nil)
+			got, err := GenerateGrovePodCliqueSet(tt.args.ctx, tt.args.dynamoDeployment, tt.args.controllerConfig, nil)
 			if (err != nil) != tt.wantErr {
-				t.Errorf("GenerateGrovePodGangSet() error = %v, wantErr %v", err, tt.wantErr)
+				t.Errorf("GenerateGrovePodCliqueSet() error = %v, wantErr %v", err, tt.wantErr)
 				return
 			}
 			sort.Slice(got.Spec.Template.Cliques, func(i, j int) bool {
@@ -3124,7 +3124,7 @@ func TestGenerateGrovePodGangSet(t *testing.T) {
 			}
 
 			if diff := cmp.Diff(got, tt.want); diff != "" {
-				t.Errorf("GenerateGrovePodGangSet() mismatch (-want +got):\n%s", diff)
+				t.Errorf("GenerateGrovePodCliqueSet() mismatch (-want +got):\n%s", diff)
 			}
 		})
 	}
@@ -4072,10 +4072,10 @@ func XTestApplyCliqueStartupDependencies(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			// Create a PodGangSet with cliques matching the roles
-			gangSet := &grovev1alpha1.PodGangSet{
-				Spec: grovev1alpha1.PodGangSetSpec{
-					Template: grovev1alpha1.PodGangSetTemplateSpec{
+			// Create a PodCliqueSet with cliques matching the roles
+			gangSet := &grovev1alpha1.PodCliqueSet{
+				Spec: grovev1alpha1.PodCliqueSetSpec{
+					Template: grovev1alpha1.PodCliqueSetTemplateSpec{
 						Cliques: []*grovev1alpha1.PodCliqueTemplateSpec{},
 					},
 				},
@@ -4234,7 +4234,7 @@ func XTestGetCliqueStartupDependencies(t *testing.T) {
 
 // deactivated for now.
 // TODO: reactivate this when we have a better way to handle the readiness probe for the leader.
-func XTestGenerateGrovePodGangSet_StartsAfterDependencies(t *testing.T) {
+func XTestGenerateGrovePodCliqueSet_StartsAfterDependencies(t *testing.T) {
 	secretsRetriever := &mockSecretsRetriever{}
 
 	tests := []struct {
@@ -4301,9 +4301,9 @@ func XTestGenerateGrovePodGangSet_StartsAfterDependencies(t *testing.T) {
 				NatsAddress: "nats-address",
 			}
 
-			got, err := GenerateGrovePodGangSet(context.Background(), dynamoDeployment, controllerConfig, secretsRetriever)
+			got, err := GenerateGrovePodCliqueSet(context.Background(), dynamoDeployment, controllerConfig, secretsRetriever)
 			if err != nil {
-				t.Errorf("GenerateGrovePodGangSet() error = %v", err)
+				t.Errorf("GenerateGrovePodCliqueSet() error = %v", err)
 				return
 			}
 

deploy/helm/README.md

@@ -38,7 +38,7 @@ helm upgrade --install dynamo-graph ./deploy/helm/chart -n dynamo-cloud -f ./com
 
 ### Installation using Grove
 
-Same example as above, but using Grove PodGangSet resources.
+Same example as above, but using Grove PodCliqueSet resources.
 
 ```bash
 helm upgrade --install dynamo-graph ./deploy/helm/chart -n dynamo-cloud -f ./components/backends/vllm/deploy/agg.yaml --set deploymentType=grove
@@ -72,10 +72,10 @@ The following table shows which deployment features are supported by the **Helm
 | Feature | Helm Chart | Operator | Description |
 |---------|------------|----------|-------------|
 | **Singlenode** (k8sDeployments) | ✅ Supported | ✅ Supported | Single-node deployments using standard Kubernetes Deployments |
-| **Singlenode** (Grove PodGangSet) | ✅ Supported | ✅ Supported | Single-node deployments using Grove PodGangSet resources |
-| **Multinode** (Grove PodGangSet and LWS) | ❌ Not Supported | ✅ Supported | Multi-node deployments requiring Grove PodGangSet and LeaderWorkerSet (LWS) |
+| **Singlenode** (Grove PodCliqueSet) | ✅ Supported | ✅ Supported | Single-node deployments using Grove PodCliqueSet resources |
+| **Multinode** (Grove PodCliqueSet and LWS) | ❌ Not Supported | ✅ Supported | Multi-node deployments requiring Grove PodCliqueSet and LeaderWorkerSet (LWS) |
 
 **Key Differences:**
-- **Helm Chart**: Best for simple single-node deployments and quick testing. Supports both basic Kubernetes deployments and Grove PodGangSet resources.
+- **Helm Chart**: Best for simple single-node deployments and quick testing. Supports both basic Kubernetes deployments and Grove PodCliqueSet resources.
 - **Operator**: Required for advanced multi-node deployments. Provides full feature support including complex distributed inference configurations.