Add annotations support to DCGM Exporter service configuration

api/nvidia/v1/clusterpolicy_types.go

@@ -946,6 +946,12 @@ type DCGMExporterServiceConfig struct {
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.displayName="Internal Traffic Policy for the DCGM Exporter K8s Service"
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.x-descriptors="urn:alm:descriptor:com.tectonic.ui:text"
 	InternalTrafficPolicy *corev1.ServiceInternalTrafficPolicy `json:"internalTrafficPolicy,omitempty"`
+
+	// Annotations to be added to the DCGM Exporter service
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.displayName="Annotations for the DCGM Exporter K8s Service"
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.x-descriptors="urn:alm:descriptor:com.tectonic.ui:text"
+	Annotations map[string]string `json:"annotations,omitempty"`
 }
 
 // DCGMExporterServiceMonitorConfig defines configuration options for the ServiceMonitor

bundle/manifests/nvidia.com_clusterpolicies.yaml

@@ -385,6 +385,12 @@ spec:
                     description: 'Optional: Service configuration for NVIDIA DCGM
                       Exporter'
                     properties:
+                      annotations:
+                        additionalProperties:
+                          type: string
+                        description: Annotations to be added to the DCGM Exporter
+                          service
+                        type: object
                       internalTrafficPolicy:
                         description: InternalTrafficPolicy describes how nodes distribute
                           service traffic they receive on the ClusterIP.

config/crd/bases/nvidia.com_clusterpolicies.yaml

@@ -385,6 +385,12 @@ spec:
                     description: 'Optional: Service configuration for NVIDIA DCGM
                       Exporter'
                     properties:
+                      annotations:
+                        additionalProperties:
+                          type: string
+                        description: Annotations to be added to the DCGM Exporter
+                          service
+                        type: object
                       internalTrafficPolicy:
                         description: InternalTrafficPolicy describes how nodes distribute
                           service traffic they receive on the ClusterIP.

controllers/object_controls.go

@@ -998,6 +998,16 @@ func TransformDCGMExporterService(obj *corev1.Service, config *gpuv1.ClusterPoli
 		if serviceConfig.InternalTrafficPolicy != nil {
 			obj.Spec.InternalTrafficPolicy = serviceConfig.InternalTrafficPolicy
 		}
+
+		// set annotations if specified for dcgm-exporter service
+		if len(serviceConfig.Annotations) > 0 {
+			if obj.ObjectMeta.Annotations == nil {
+				obj.ObjectMeta.Annotations = make(map[string]string)
+			}
+			for annKey, annValue := range serviceConfig.Annotations {
+				obj.ObjectMeta.Annotations[annKey] = annValue
+			}
+		}
 	}
 	return nil
 }

controllers/transforms_test.go

@@ -1600,6 +1600,124 @@ func TestTransformNodeStatusExporter(t *testing.T) {
 	}
 }
 
+func TestTransformDCGMExporterService(t *testing.T) {
+	testCases := []struct {
+		description     string
+		service         *corev1.Service
+		cpSpec          *gpuv1.ClusterPolicySpec
+		expectedService *corev1.Service
+	}{
+		{
+			description: "service without annotations",
+			service: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape": "true",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeClusterIP,
+				},
+			},
+			cpSpec: &gpuv1.ClusterPolicySpec{
+				DCGMExporter: gpuv1.DCGMExporterSpec{
+					ServiceSpec: &gpuv1.DCGMExporterServiceConfig{
+						Type: corev1.ServiceTypeNodePort,
+					},
+				},
+			},
+			expectedService: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape": "true",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeNodePort,
+				},
+			},
+		},
+		{
+			description: "service with custom annotations",
+			service: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape": "true",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeClusterIP,
+				},
+			},
+			cpSpec: &gpuv1.ClusterPolicySpec{
+				DCGMExporter: gpuv1.DCGMExporterSpec{
+					ServiceSpec: &gpuv1.DCGMExporterServiceConfig{
+						Type: corev1.ServiceTypeNodePort,
+						Annotations: map[string]string{
+							"custom.annotation/key": "custom-value",
+							"another.annotation":    "another-value",
+						},
+					},
+				},
+			},
+			expectedService: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape":  "true",
+						"custom.annotation/key": "custom-value",
+						"another.annotation":    "another-value",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeNodePort,
+				},
+			},
+		},
+		{
+			description: "service with nil service spec",
+			service: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape": "true",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeClusterIP,
+				},
+			},
+			cpSpec: &gpuv1.ClusterPolicySpec{
+				DCGMExporter: gpuv1.DCGMExporterSpec{
+					ServiceSpec: nil,
+				},
+			},
+			expectedService: &corev1.Service{
+				ObjectMeta: metav1.ObjectMeta{
+					Name: "nvidia-dcgm-exporter",
+					Annotations: map[string]string{
+						"prometheus.io/scrape": "true",
+					},
+				},
+				Spec: corev1.ServiceSpec{
+					Type: corev1.ServiceTypeClusterIP,
+				},
+			},
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.description, func(t *testing.T) {
+			err := TransformDCGMExporterService(tc.service, tc.cpSpec)
+			require.NoError(t, err)
+			require.EqualValues(t, tc.expectedService, tc.service)
+		})
+	}
+}
+
 func TestTransformDriver(t *testing.T) {
 	initMockK8sClients()
 	testCases := []struct {

deployments/gpu-operator/crds/nvidia.com_clusterpolicies.yaml

@@ -385,6 +385,12 @@ spec:
                     description: 'Optional: Service configuration for NVIDIA DCGM
                       Exporter'
                     properties:
+                      annotations:
+                        additionalProperties:
+                          type: string
+                        description: Annotations to be added to the DCGM Exporter
+                          service
+                        type: object
                       internalTrafficPolicy:
                         description: InternalTrafficPolicy describes how nodes distribute
                           service traffic they receive on the ClusterIP.

deployments/gpu-operator/values.yaml

@@ -295,6 +295,7 @@ dcgmExporter:
   resources: {}
   service:
     internalTrafficPolicy: Cluster
+    annotations: {}
   serviceMonitor:
     enabled: false
     interval: 15s