Automate installing NVIDIA Container Toolkit

cmd/minikube/cmd/start.go

@@ -1285,7 +1285,7 @@ func validateFlags(cmd *cobra.Command, drvName string) {
 	}
 
 	if cmd.Flags().Changed(containerRuntime) {
-		err := validateRuntime(viper.GetString(containerRuntime))
+		err := validateRuntime(viper.GetString(containerRuntime), drvName)
 		if err != nil {
 			exit.Message(reason.Usage, "{{.err}}", out.V{"err": err})
 		}
@@ -1402,7 +1402,7 @@ func validateDiskSize(diskSize string) error {
 }
 
 // validateRuntime validates the supplied runtime
-func validateRuntime(rtime string) error {
+func validateRuntime(rtime, driverName string) error {
 	validOptions := cruntime.ValidRuntimes()
 	// `crio` is accepted as an alternative spelling to `cri-o`
 	validOptions = append(validOptions, constants.CRIO)
@@ -1431,6 +1431,11 @@ func validateRuntime(rtime string) error {
 	if !validRuntime {
 		return errors.Errorf("Invalid Container Runtime: %s. Valid runtimes are: %s", rtime, cruntime.ValidRuntimes())
 	}
+
+	if rtime == constants.NvidiaDocker && driverName != constants.Docker {
+		return errors.Errorf("The nvidia-docker container-runtime can only be run with the docker driver")
+	}
+
 	return nil
 }
 
@@ -1793,7 +1798,7 @@ func validateContainerRuntime(old *config.ClusterConfig) {
 		return
 	}
 
-	if err := validateRuntime(old.KubernetesConfig.ContainerRuntime); err != nil {
+	if err := validateRuntime(old.KubernetesConfig.ContainerRuntime, old.Driver); err != nil {
 		klog.Errorf("Error parsing old runtime %q: %v", old.KubernetesConfig.ContainerRuntime, err)
 	}
 }

pkg/drivers/kic/kic.go

@@ -90,6 +90,9 @@ func (d *Driver) Create() error {
 		APIServerPort: d.NodeConfig.APIServerPort,
 	}
 
+	if d.NodeConfig.ContainerRuntime == constants.NvidiaDocker {
+		params.GPUs = true
+	}
 	networkName := d.NodeConfig.Network
 	if networkName == "" {
 		networkName = d.NodeConfig.ClusterName

pkg/drivers/kic/oci/oci.go

@@ -190,6 +190,9 @@ func CreateContainerNode(p CreateParams) error {
 		runArgs = append(runArgs, "--network", p.Network)
 		runArgs = append(runArgs, "--ip", p.IP)
 	}
+	if p.GPUs {
+		runArgs = append(runArgs, "--gpus", "all")
+	}
 
 	memcgSwap := hasMemorySwapCgroup()
 	memcg := HasMemoryCgroup()

pkg/drivers/kic/oci/types.go

@@ -58,7 +58,8 @@ type CreateParams struct {
 	ExtraArgs     []string          // a list of any extra option to pass to oci binary during creation time, for example --expose 8080...
 	OCIBinary     string            // docker or podman
 	Network       string            // network name that the container will attach to
-	IP            string            // static IP to assign for th container in the cluster network
+	IP            string            // static IP to assign the container in the cluster network
+	GPUs          bool              // add GPU devices to the container
 }
 
 // createOpt is an option for Create

pkg/minikube/constants/constants.go

@@ -63,6 +63,8 @@ const (
 	CRIO = "crio"
 	// Docker is the default name and spelling for the docker container runtime
 	Docker = "docker"
+	// NvidiaDocker is the default name and spelling for the nvidia-docker container runtime
+	NvidiaDocker = "nvidia-docker"
 	// DefaultContainerRuntime is our default container runtime
 	DefaultContainerRuntime = ""
 

pkg/minikube/cruntime/docker.go

@@ -39,6 +39,7 @@ import (
 	"k8s.io/minikube/pkg/minikube/docker"
 	"k8s.io/minikube/pkg/minikube/download"
 	"k8s.io/minikube/pkg/minikube/image"
+	"k8s.io/minikube/pkg/minikube/out"
 	"k8s.io/minikube/pkg/minikube/style"
 	"k8s.io/minikube/pkg/minikube/sysinit"
 )
@@ -560,7 +561,10 @@ func (r *Docker) configureDocker(driver string) error {
 		},
 		StorageDriver: "overlay2",
 	}
-	if r.Type == "nvidia-docker" {
+	if r.Type == constants.NvidiaDocker {
+		if err := r.installNvidiaContainerToolkit(); err != nil {
+			return fmt.Errorf("failed installing the NVIDIA Container Toolkit: %v", err)
+		}
 		daemonConfig.DefaultRuntime = "nvidia"
 		runtimes := &dockerDaemonRuntimes{}
 		runtimes.Nvidia.Path = "/usr/bin/nvidia-container-runtime"
@@ -574,6 +578,25 @@ func (r *Docker) configureDocker(driver string) error {
 	return r.Runner.Copy(ma)
 }
 
+// installNvidiaContainerToolkit installs the NVIDIA Container Toolkit
+// https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html
+func (r *Docker) installNvidiaContainerToolkit() error {
+	out.Styled(style.Toolkit, "Installing the NVIDIA Container Toolkit...")
+	cmds := []string{
+		"curl -s -L https://nvidia.github.io/libnvidia-container/gpgkey | sudo apt-key add -",
+		"curl -s -L https://nvidia.github.io/libnvidia-container/$(. /etc/os-release;echo $ID$VERSION_ID)/libnvidia-container.list | sudo tee /etc/apt/sources.list.d/libnvidia-container.list",
+		"sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit && sudo rm -rf /var/lib/apt/lists/*o",
+	}
+
+	for _, cmd := range cmds {
+		c := exec.Command("/bin/bash", "-c", cmd)
+		if _, err := r.Runner.RunCmd(c); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
 // Preload preloads docker with k8s images:
 // 1. Copy over the preloaded tarball into the VM
 // 2. Extract the preloaded tarball to the correct directory

pkg/minikube/style/style.go

@@ -139,6 +139,7 @@ var Config = map[Enum]Options{
 	VerifyingNoLine:  {Prefix: "🤔  ", OmitNewline: true},
 	Verifying:        {Prefix: "🤔  "},
 	CNI:              {Prefix: "🔗  "},
+	Toolkit:          {Prefix: "🛠️   "},
 }
 
 // LowPrefix returns a 7-bit compatible prefix for a style

pkg/minikube/style/style_enum.go

@@ -105,4 +105,5 @@ const (
 	Warning
 	Workaround
 	CNI
+	Toolkit
 )