NVIDIA/k8s-device-plugin源碼分析頂原薦

Author: [email protected]

k8s-device-plugin內部實現原理圖

在Kubernetes如何通過Device Plugins來使用NVIDIA GPU中，對NVIDIA/k8s-device-plugin的工作原理進行了深入分析，爲了方便我們在這再次貼出其內部實現原理圖：

PreStartContainer和GetDevicePluginOptions兩個接口，在NVIDIA/k8s-device-plugin中可以忽略，可以認爲是空實現。我們主要關注ListAndWatch和Allocate的實現。

啓動

一切從main函數開始！核心的代碼如下：

func main() {
	log.Println("Loading NVML")
	if err := nvml.Init(); err != nil {
		select {}
	}
    ...
	log.Println("Fetching devices.")
	if len(getDevices()) == 0 {
		select {}
	}

	log.Println("Starting FS watcher.")
	watcher, err := newFSWatcher(pluginapi.DevicePluginPath)
	if err != nil {
		os.Exit(1)
	}
    ...
	log.Println("Starting OS watcher.")
	sigs := newOSWatcher(syscall.SIGHUP, syscall.SIGINT, syscall.SIGTERM, syscall.SIGQUIT)

	restart := true
	var devicePlugin *NvidiaDevicePlugin

L:
	for {
		if restart {
			if devicePlugin != nil {
				devicePlugin.Stop()
			}

			devicePlugin = NewNvidiaDevicePlugin()
			if err := devicePlugin.Serve(); err != nil {
				...
			} else {
				restart = false
			}
		}

		select {
		case event := <-watcher.Events:
			if event.Name == pluginapi.KubeletSocket && event.Op&fsnotify.Create == fsnotify.Create {
				restart = true
			}

		case err := <-watcher.Errors:

		case s := <-sigs:
			switch s {
			case syscall.SIGHUP:
				restart = true
			default:
				devicePlugin.Stop()
				break L
			}
		}
	}
}

相關說明不需多說，請參考下面的流程邏輯圖：

Serve

k8s-device-plugin啓動流程中，devicePlugin.Serve負責啓動gRPC Server Start對外提供服務，然後把自己註冊到kubelet。

// Serve starts the gRPC server and register the device plugin to Kubelet
func (m *NvidiaDevicePlugin) Serve() error {
	err := m.Start()
	if err != nil {
		log.Printf("Could not start device plugin: %s", err)
		return err
	}
	log.Println("Starting to serve on", m.socket)

	err = m.Register(pluginapi.KubeletSocket, resourceName)
	if err != nil {
		log.Printf("Could not register device plugin: %s", err)
		m.Stop()
		return err
	}
	log.Println("Registered device plugin with Kubelet")

	return nil
}

Start

Start的代碼如下：

// Start starts the gRPC server of the device plugin
func (m *NvidiaDevicePlugin) Start() error {
	err := m.cleanup()
	if err != nil {
		return err
	}

	sock, err := net.Listen("unix", m.socket)
	if err != nil {
		return err
	}

	m.server = grpc.NewServer([]grpc.ServerOption{}...)
	pluginapi.RegisterDevicePluginServer(m.server, m)

	go m.server.Serve(sock)

	// Wait for server to start by launching a blocking connexion
	conn, err := dial(m.socket, 5*time.Second)
	if err != nil {
		return err
	}
	conn.Close()

	go m.healthcheck()

	return nil
}

更加深入的代碼調用關係，這裏不多介紹，直接貼出Start的實現邏輯圖：

Start流程中負責創建nvidia.sock文件。

需要特別說明healthcheck部分：

healthcheck啓動協程對管理的devices進行健康狀態監控，一旦發現有device unhealthy，則發送到NvidiaDevicePlugin的health channel。device plugin的ListAndWatch會從health channel中獲取這些unhealthy devices，並通知到kubelet進行更新。
只監控nvmlEventTypeXidCriticalError事件，一旦監控到某個device的這個Event，就認爲該device unhealthy。關於nvmlEventTypeXidCriticalError的說明，請參考NVIDIA的nvml api文檔。
可以通過設置NVIDIA device plugin Pod內的環境變量DP_DISABLE_HEALTHCHECKS爲”all”來取消healthcheck。不設置或者設置爲其他值都會啓動healthcheck，默認部署時不設置。

Register

Start之後，接着進入Register流程，其代碼如下：

// Register registers the device plugin for the given resourceName with Kubelet.
func (m *NvidiaDevicePlugin) Register(kubeletEndpoint, resourceName string) error {
	conn, err := dial(kubeletEndpoint, 5*time.Second)
	if err != nil {
		return err
	}
	defer conn.Close()

	client := pluginapi.NewRegistrationClient(conn)
	reqt := &pluginapi.RegisterRequest{
		Version:      pluginapi.Version,
		Endpoint:     path.Base(m.socket),
		ResourceName: resourceName,
	}

	_, err = client.Register(context.Background(), reqt)
	if err != nil {
		return err
	}
	return nil
}

Register的實現流程圖如下：

註冊的Resource Name是nvidia.com/gpu
註冊的Version是v1beta1

Stop

Stop的代碼如下：

// Stop stops the gRPC server
func (m *NvidiaDevicePlugin) Stop() error {
	if m.server == nil {
		return nil
	}

	m.server.Stop()
	m.server = nil
	close(m.stop)

	return m.cleanup()
}

Stop的實現流程圖如下：

Stop流程中負責停止gRPC Server，並刪除nvidia.sock。

ListAndWatch

ListAndWatch接口主要負責監控health channel，發現有gpu變成unhealthy後，將完成的gpu list信息（ID和health狀態）發送給kubelet進行更新。

// ListAndWatch lists devices and update that list according to the health status
func (m *NvidiaDevicePlugin) ListAndWatch(e *pluginapi.Empty, s pluginapi.DevicePlugin_ListAndWatchServer) error {
	s.Send(&pluginapi.ListAndWatchResponse{Devices: m.devs})

	for {
		select {
		case <-m.stop:
			return nil
		case d := <-m.health:
			// FIXME: there is no way to recover from the Unhealthy state.
			d.Health = pluginapi.Unhealthy
			s.Send(&pluginapi.ListAndWatchResponse{Devices: m.devs})
		}
	}
}

ListAndWatch的實現流程圖如下：

Allocate

Allocate負責接口kubelet爲Container請求分配gpu的請求，請求的結構體如下：

// - Allocate is expected to be called during pod creation since allocation
//   failures for any container would result in pod startup failure.
// - Allocate allows kubelet to exposes additional artifacts in a pod's
//   environment as directed by the plugin.
// - Allocate allows Device Plugin to run device specific operations on
//   the Devices requested
type AllocateRequest struct {
	ContainerRequests []*ContainerAllocateRequest `protobuf:"bytes,1,rep,name=container_requests,json=containerRequests" json:"container_requests,omitempty"`
}

type ContainerAllocateRequest struct {
	DevicesIDs []string `protobuf:"bytes,1,rep,name=devicesIDs" json:"devicesIDs,omitempty"`
}

device plugin Allocate的Response結構體定義如下：

// AllocateResponse includes the artifacts that needs to be injected into
// a container for accessing 'deviceIDs' that were mentioned as part of
// 'AllocateRequest'.
// Failure Handling:
// if Kubelet sends an allocation request for dev1 and dev2.
// Allocation on dev1 succeeds but allocation on dev2 fails.
// The Device plugin should send a ListAndWatch update and fail the
// Allocation request
type AllocateResponse struct {
	ContainerResponses []*ContainerAllocateResponse `protobuf:"bytes,1,rep,name=container_responses,json=containerResponses" json:"container_responses,omitempty"`
}

type ContainerAllocateResponse struct {
	// List of environment variable to be set in the container to access one of more devices.
	Envs map[string]string `protobuf:"bytes,1,rep,name=envs" json:"envs,omitempty" protobuf_key:"bytes,1,opt,name=key,proto3" protobuf_val:"bytes,2,opt,name=value,proto3"`
	// Mounts for the container.
	Mounts []*Mount `protobuf:"bytes,2,rep,name=mounts" json:"mounts,omitempty"`
	// Devices for the container.
	Devices []*DeviceSpec `protobuf:"bytes,3,rep,name=devices" json:"devices,omitempty"`
	// Container annotations to pass to the container runtime
	Annotations map[string]string `protobuf:"bytes,4,rep,name=annotations" json:"annotations,omitempty" protobuf_key:"bytes,1,opt,name=key,proto3" protobuf_val:"bytes,2,opt,name=value,proto3"`
}

Allocate的代碼實現如下：

// Allocate which return list of devices.
func (m *NvidiaDevicePlugin) Allocate(ctx context.Context, reqs *pluginapi.AllocateRequest) (*pluginapi.AllocateResponse, error) {
	devs := m.devs
	responses := pluginapi.AllocateResponse{}
	for _, req := range reqs.ContainerRequests {
		response := pluginapi.ContainerAllocateResponse{
			Envs: map[string]string{
				"NVIDIA_VISIBLE_DEVICES": strings.Join(req.DevicesIDs, ","),
			},
		}

		for _, id := range req.DevicesIDs {
			if !deviceExists(devs, id) {
				return nil, fmt.Errorf("invalid allocation request: unknown device: %s", id)
			}
		}

		responses.ContainerResponses = append(responses.ContainerResponses, &response)
	}

	return &responses, nil
}

下面是其實現邏輯圖：

Allocate中會遍歷ContainerRequests，將DeviceIDs封裝到ContainerAllocateResponse的Envs:NVIDIA_VISIBLE_DEVICES中，格式爲：”${ID_1},${ID_2},...”
除此之外，並沒有封裝Mounts, Devices, Annotations。

總結

NVIDIA/k8s-device-plugin的代碼中，依賴於nvidia-docker代碼庫，存在很多golang調用C庫的地方，還需要大家自行到[nvml api文檔](https://docs.nvidia.com/deploy/nvml-api）中查看相關C函數聲明。上一篇博客介紹了 Kubernetes如何通過Device Plugins來使用NVIDIA GPU，這篇博客介紹NVIDIA/k8s-device-plugin的代碼實現流程，下一篇博客我覺得還有必要對kubelet device plugin manger進行代碼分析，如此才能完整的理解整個交互細節。

NVIDIA/k8s-device-plugin源碼分析頂原薦

k8s-device-plugin內部實現原理圖

啓動

Serve

Start

Register

Stop

ListAndWatch

Allocate

總結

使用neovim打造go ide(支持代碼跳轉, 代碼補全, 實時語法檢查)

挑戰程序設計競賽 2.3章習題 poj 3046 Ant Counting

Shell/Python中的用戶名獲取

SkyDNS2源碼分析

Kubernetes DNS Service技術研究

Kubernetes ReplicationController源碼分析

kube-proxy源碼分析

從源碼解析kube-scheduler默認的配置

https://yachay.unat.edu.pe/blog/index.php?comment_area=format_blog&comment_component=blog&comment_co

linux以太網驅動總結

NVIDIA/k8s-device-plugin源碼分析 頂 原 薦

k8s-device-plugin內部實現原理圖

啓動

Serve

Start

Register

Stop

ListAndWatch

Allocate

總結

NVIDIA/k8s-device-plugin源碼分析頂原薦