fix: add label to persist DGD name on downstream pods (#2729)

deploy/cloud/operator/internal/consts/consts.go

@@ -23,6 +23,7 @@ const (
 
 	KubeAnnotationEnableGrove = "nvidia.com/enable-grove"
 
+	KubeLabelDynamoGraphDeploymentName  = "nvidia.com/dynamo-graph-deployment-name"
 	KubeLabelDynamoComponent            = "nvidia.com/dynamo-component"
 	KubeLabelDynamoNamespace            = "nvidia.com/dynamo-namespace"
 	KubeLabelDynamoDeploymentTargetType = "nvidia.com/dynamo-deployment-target-type"

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

@@ -1146,6 +1146,9 @@ func (r *DynamoComponentDeploymentReconciler) generatePodTemplateSpec(ctx contex
 		podLabels[commonconsts.KubeLabelMetricsEnabled] = commonconsts.KubeLabelValueTrue
 	}
 
+	// Add label for the dynamo graph deployment on the pods themselves
+	podLabels[commonconsts.KubeLabelDynamoGraphDeploymentName] = opt.dynamoComponentDeployment.Spec.Labels[commonconsts.KubeLabelDynamoGraphDeploymentName]
+
 	// Add component type label if specified
 	if opt.dynamoComponentDeployment.Spec.ComponentType != "" {
 		podLabels[commonconsts.KubeLabelDynamoComponentType] = opt.dynamoComponentDeployment.Spec.ComponentType

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

@@ -782,7 +782,8 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									commonconsts.KubeLabelMetricsEnabled: commonconsts.KubeLabelValueTrue,
 									"role":                               "leader",
 									"nvidia.com/label1":                  "label1",
-									commonconsts.KubeLabelDynamoComponentType: commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoGraphDeploymentName: "",
 								},
 								Annotations: map[string]string{
 									"scheduling.k8s.io/group-name": "test-lws-deploy-0",
@@ -891,7 +892,8 @@ func TestDynamoComponentDeploymentReconciler_generateLeaderWorkerSet(t *testing.
 									commonconsts.KubeLabelMetricsEnabled: commonconsts.KubeLabelValueTrue,
 									"role":                               "worker",
 									"nvidia.com/label1":                  "label1",
-									commonconsts.KubeLabelDynamoComponentType: commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoComponentType:       commonconsts.ComponentTypeWorker,
+									commonconsts.KubeLabelDynamoGraphDeploymentName: "",
 								},
 								Annotations: map[string]string{
 									"scheduling.k8s.io/group-name": "test-lws-deploy-0",

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

@@ -138,6 +138,7 @@ func GenerateDynamoComponentsDeployments(ctx context.Context, parentDynamoGraphD
 		deployment.Labels = labels
 		labels[commonconsts.KubeLabelDynamoComponent] = componentName
 		labels[commonconsts.KubeLabelDynamoNamespace] = graphDynamoNamespace
+		labels[commonconsts.KubeLabelDynamoGraphDeploymentName] = parentDynamoGraphDeployment.Name
 
 		// Propagate metrics annotation from parent deployment if present
 		if parentDynamoGraphDeployment.Annotations != nil {
@@ -960,6 +961,7 @@ func GenerateGrovePodGangSet(
 func generateLabels(component *v1alpha1.DynamoComponentDeploymentOverridesSpec, dynamoDeployment *v1alpha1.DynamoGraphDeployment, componentName string) (map[string]string, error) {
 	labels := make(map[string]string)
 	labels[commonconsts.KubeLabelDynamoSelector] = GetDynamoComponentName(dynamoDeployment, componentName)
+	labels[commonconsts.KubeLabelDynamoGraphDeploymentName] = dynamoDeployment.Name
 	if component.ComponentType != "" {
 		labels[commonconsts.KubeLabelDynamoComponentType] = component.ComponentType
 	}