Set cdi.enabled to true by default

Signed-off-by: Christopher Desiniotis <[email protected]>

Author: cdesiniotis

api/nvidia/v1/clusterpolicy_types.go

@@ -1660,7 +1660,7 @@ type VGPUDevicesConfigSpec struct {
 type CDIConfigSpec struct {
 	// Enabled indicates whether the Container Device Interface (CDI) should be used as the mechanism for making GPUs accessible to containers.
 	// +kubebuilder:validation:Optional
-	// +kubebuilder:default=false
+	// +kubebuilder:default=true
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.displayName="Enable CDI as the mechanism for making GPUs accessible to containers"
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.x-descriptors="urn:alm:descriptor:com.tectonic.ui:booleanSwitch"
@@ -2070,7 +2070,7 @@ func (l *DriverLicensingConfigSpec) IsNLSEnabled() bool {
 // providing GPU access to containers
 func (c *CDIConfigSpec) IsEnabled() bool {
 	if c.Enabled == nil {
-		return false
+		return true
 	}
 	return *c.Enabled
 }

bundle/manifests/gpu-operator-certified.clusterserviceversion.yaml

@@ -34,6 +34,9 @@ metadata:
               "initContainer": {
               }
             },
+            "cdi": {
+              "enabled": true
+            },
             "sandboxWorkloads": {
               "enabled": false,
               "defaultWorkload": "container"

bundle/manifests/nvidia.com_clusterpolicies.yaml

@@ -141,7 +141,7 @@ spec:
                       for making GPUs accessible to containers.'
                     type: boolean
                   enabled:
-                    default: false
+                    default: true
                     description: Enabled indicates whether the Container Device Interface
                       (CDI) should be used as the mechanism for making GPUs accessible
                       to containers.

config/crd/bases/nvidia.com_clusterpolicies.yaml

@@ -141,7 +141,7 @@ spec:
                       for making GPUs accessible to containers.'
                     type: boolean
                   enabled:
-                    default: false
+                    default: true
                     description: Enabled indicates whether the Container Device Interface
                       (CDI) should be used as the mechanism for making GPUs accessible
                       to containers.

controllers/transforms_test.go

@@ -643,6 +643,9 @@ func TestTransformToolkit(t *testing.T) {
 						},
 					},
 					Env: []corev1.EnvVar{
+						{Name: CDIEnabledEnvName, Value: "true"},
+						{Name: NvidiaRuntimeSetAsDefaultEnvName, Value: "false"},
+						{Name: NvidiaCtrRuntimeModeEnvName, Value: "cdi"},
 						{Name: "foo", Value: "bar"},
 						{Name: "RUNTIME", Value: "containerd"},
 						{Name: "CONTAINERD_RUNTIME_CLASS", Value: "nvidia"},
@@ -713,6 +716,9 @@ func TestTransformToolkit(t *testing.T) {
 						},
 					},
 					Env: []corev1.EnvVar{
+						{Name: CDIEnabledEnvName, Value: "true"},
+						{Name: NvidiaRuntimeSetAsDefaultEnvName, Value: "false"},
+						{Name: NvidiaCtrRuntimeModeEnvName, Value: "cdi"},
 						{Name: "CONTAINERD_CONFIG", Value: "/runtime/config-dir/config.toml"},
 						{Name: "CONTAINERD_SOCKET", Value: "/runtime/sock-dir/containerd.sock"},
 						{Name: "CONTAINERD_RUNTIME_CLASS", Value: "nvidia"},
@@ -770,6 +776,10 @@ func TestTransformDevicePlugin(t *testing.T) {
 						{Name: "foo", Value: "bar"},
 					},
 				},
+				Toolkit: gpuv1.ToolkitSpec{
+					Enabled:    newBoolPtr(true),
+					InstallDir: "/path/to/install",
+				},
 			},
 			expectedDs: NewDaemonset().WithContainer(corev1.Container{
 				Name:            "nvidia-device-plugin",
@@ -778,6 +788,10 @@ func TestTransformDevicePlugin(t *testing.T) {
 				Args:            []string{"--fail-on-init-error=false"},
 				Env: []corev1.EnvVar{
 					{Name: "NVIDIA_MIG_MONITOR_DEVICES", Value: "all"},
+					{Name: CDIEnabledEnvName, Value: "true"},
+					{Name: DeviceListStrategyEnvName, Value: "cdi-annotations,cdi-cri"},
+					{Name: CDIAnnotationPrefixEnvName, Value: "cdi.k8s.io/"},
+					{Name: NvidiaCDIHookPathEnvName, Value: "/path/to/install/toolkit/nvidia-cdi-hook"},
 					{Name: "foo", Value: "bar"},
 				},
 			}).WithContainer(corev1.Container{Name: "dummy"}).WithPullSecret("pull-secret").WithRuntimeClassName("nvidia"),
@@ -867,13 +881,19 @@ func TestTransformMigManager(t *testing.T) {
 						{Name: "foo", Value: "bar"},
 					},
 				},
+				Toolkit: gpuv1.ToolkitSpec{
+					Enabled:    newBoolPtr(true),
+					InstallDir: "/path/to/install",
+				},
 			},
 			expectedDs: NewDaemonset().WithContainer(corev1.Container{
 				Name:            "mig-manager",
 				Image:           "nvcr.io/nvidia/cloud-native/mig-manager:v1.0.0",
 				ImagePullPolicy: corev1.PullIfNotPresent,
 				Args:            []string{"--test-flag"},
 				Env: []corev1.EnvVar{
+					{Name: CDIEnabledEnvName, Value: "true"},
+					{Name: NvidiaCDIHookPathEnvName, Value: "/path/to/install/toolkit/nvidia-cdi-hook"},
 					{Name: "foo", Value: "bar"},
 				},
 			}).WithPullSecret("pull-secret").WithRuntimeClassName("nvidia"),

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

@@ -141,7 +141,7 @@ spec:
                       for making GPUs accessible to containers.'
                     type: boolean
                   enabled:
-                    default: false
+                    default: true
                     description: Enabled indicates whether the Container Device Interface
                       (CDI) should be used as the mechanism for making GPUs accessible
                       to containers.

deployments/gpu-operator/values.yaml

@@ -13,7 +13,7 @@ psa:
   enabled: false
 
 cdi:
-  enabled: false
+  enabled: true
 
 sandboxWorkloads:
   enabled: false