remove most of the kata manager related code under the assumption that kata manager state is not created ever

Author: rajatchopra

controllers/object_controls.go

@@ -45,7 +45,6 @@ import (
 	"k8s.io/utils/ptr"
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/controller/controllerutil"
-	"sigs.k8s.io/yaml"
 
 	gpuv1 "github.com/NVIDIA/gpu-operator/api/nvidia/v1"
 	driverconfig "github.com/NVIDIA/gpu-operator/internal/config"
@@ -161,8 +160,6 @@ const (
 	DeviceListStrategyEnvName = "DEVICE_LIST_STRATEGY"
 	// CDIAnnotationPrefixEnvName is the name of the device-plugin envvar for configuring the CDI annotation prefix
 	CDIAnnotationPrefixEnvName = "CDI_ANNOTATION_PREFIX"
-	// KataManagerAnnotationHashKey is the annotation indicating the hash of the kata-manager configuration
-	KataManagerAnnotationHashKey = "nvidia.com/kata-manager.last-applied-hash"
 	// DefaultKataArtifactsDir is the default directory to store kata artifacts on the host
 	DefaultKataArtifactsDir = "/opt/nvidia-gpu-operator/artifacts/runtimeclasses/"
 	// PodControllerRevisionHashLabelKey is the annotation key for pod controller revision hash value
@@ -609,16 +606,6 @@ func createConfigMap(n ClusterPolicyController, configMapIdx int) (gpuv1.State,
 		}
 	}
 
-	if obj.Name == "nvidia-kata-manager-config" {
-		data, err := yaml.Marshal(config.KataManager.Config)
-		if err != nil {
-			return gpuv1.NotReady, fmt.Errorf("failed to marshal kata manager config: %v", err)
-		}
-		obj.Data = map[string]string{
-			"config.yaml": string(data),
-		}
-	}
-
 	if err := controllerutil.SetControllerReference(n.singleton, obj, n.scheme); err != nil {
 		return gpuv1.NotReady, err
 	}
@@ -1462,13 +1449,7 @@ func transformRuntimeConfigAndSocketMounts(obj *appsv1.DaemonSet, runtime string
 	}
 
 	// Handle the drop-in configs
-	// TODO: It's a bit of a hack to skip the `nvidia-kata-manager` container here.
-	// Ideally if the two projects are using the SAME API then this should be
-	// captured more rigorously.
-	// Note that we probably want to implement drop-in file support in the
-	// kata manager in any case -- in which case it will be good to use a
-	// similar implementation.
-	if dropInConfigFile != "" && container.Name != "nvidia-kata-manager" {
+	if dropInConfigFile != "" {
 		sourceConfigFileName := path.Base(dropInConfigFile)
 		sourceConfigDir := path.Dir(dropInConfigFile)
 		containerConfigDir := DefaultRuntimeDropInConfigTargetDir
@@ -5181,11 +5162,6 @@ func RuntimeClasses(n ClusterPolicyController) (gpuv1.State, error) {
 	status := gpuv1.Ready
 	state := n.idx
 
-	if n.stateNames[state] == "state-kata-manager" {
-		// Kata Manager is deprecated, no need to process anything
-		return gpuv1.Ready, nil
-	}
-
 	nvidiaRuntimeClasses := n.resources[state].RuntimeClasses
 	if n.stateNames[state] == "pre-requisites" && !n.isStateEnabled(n.stateNames[state]) {
 		err := clearRuntimeClasses(n, nvidiaRuntimeClasses)

controllers/state_manager.go

@@ -100,7 +100,6 @@ var gpuStateLabels = map[string]map[string]string{
 		"nvidia.com/gpu.deploy.sandbox-device-plugin": "true",
 		"nvidia.com/gpu.deploy.sandbox-validator":     "true",
 		"nvidia.com/gpu.deploy.vfio-manager":          "true",
-		"nvidia.com/gpu.deploy.kata-manager":          "true",
 		"nvidia.com/gpu.deploy.cc-manager":            "true",
 	},
 	gpuWorkloadConfigVMVgpu: {

controllers/transforms_test.go

@@ -511,31 +511,6 @@ func TestTransformForRuntime(t *testing.T) {
 					},
 				}),
 		},
-		// Cover the kata-manager naming case
-		{
-			description: "containerd skips drop-in for kata manager",
-			runtime:     gpuv1.Containerd,
-			input: NewDaemonset().
-				WithContainer(corev1.Container{Name: "nvidia-kata-manager"}),
-			expectedOutput: NewDaemonset().
-				WithHostPathVolume("containerd-config", filepath.Dir(DefaultContainerdConfigFile), ptr.To(corev1.HostPathDirectoryOrCreate)).
-				WithHostPathVolume("containerd-socket", filepath.Dir(DefaultContainerdSocketFile), nil).
-				WithContainer(corev1.Container{
-					Name: "nvidia-kata-manager",
-					Env: []corev1.EnvVar{
-						{Name: "RUNTIME", Value: gpuv1.Containerd.String()},
-						{Name: "CONTAINERD_RUNTIME_CLASS", Value: DefaultRuntimeClass},
-						{Name: "RUNTIME_CONFIG", Value: filepath.Join(DefaultRuntimeConfigTargetDir, filepath.Base(DefaultContainerdConfigFile))},
-						{Name: "CONTAINERD_CONFIG", Value: filepath.Join(DefaultRuntimeConfigTargetDir, filepath.Base(DefaultContainerdConfigFile))},
-						{Name: "RUNTIME_SOCKET", Value: filepath.Join(DefaultRuntimeSocketTargetDir, filepath.Base(DefaultContainerdSocketFile))},
-						{Name: "CONTAINERD_SOCKET", Value: filepath.Join(DefaultRuntimeSocketTargetDir, filepath.Base(DefaultContainerdSocketFile))},
-					},
-					VolumeMounts: []corev1.VolumeMount{
-						{Name: "containerd-config", MountPath: DefaultRuntimeConfigTargetDir},
-						{Name: "containerd-socket", MountPath: DefaultRuntimeSocketTargetDir},
-					},
-				}),
-		},
 		{
 			description: "docker",
 			runtime:     gpuv1.Docker,