kubelet组件
## kubelet
每个节点上都运行一个kubelet服务进程，默认监听10250端口。
●接收并执行master发来的指令;
●管理Pod及Pod中的容器;
●每个kubelet进程会在API Server上注册节点自身信息，定期向master节点汇报节点的资源使用情况，并通过cAdvisor监控节点和容器的资源。
### kubelet架构
### kubelet 管理pod的核心流程
### 节点管理
节点管理主要是节点自注册和节点状态更新:
●kubelet 可以通过设置启动参数`--register-node`来确定是否向API Server注册自己;
●如果kubelet没有选择自注册模式，则需要用户自己配置Node资源信息，同时需要告知kubelet集群上的API Server的位置;
●kubelet 在启动时通过API Server 注册节点信息，并定时向 API Server 发送节点新消息，API Server在接收到新消息后，将信息写入etcd。
### pod管理
获取Pod清单:
●文件:启动参数--config指定的配置目录下的文件(默认/etc/Kubernetes/manifests/)。 该文件每20秒重新检查一次(可配置)
●HTTP endpoint (URL) :启动参数--manifest-url设置。每20秒检查- 次这个端点(可配置)。
●API Server: 通过API Server监听etcd目录，同步Pod清单。
●HTTP Server: kubelet侦听HTTP请求，并响应简单的API以提交新的Pod清单。
### pod启动流程
### kubelet启动pod的流程
监控
## kubelet 代码走读
```golang
server.go:NewKubeletCommand()-->>
kubeletDeps, err := UnsecuredDependencies(kubeletServer, utilfeature.DefaultFeatureGate)-->>
plugins, err := ProbeVolumePlugins(featureGate)
server.go:Run(kubeletServer, kubeletDeps, utilfeature.DefaultFeatureGate, stopCh)-->>
run(s, kubeDeps, featureGate, stopCh)-->>
kubeDeps.ContainerManager, err = cm.NewContainerManager() // init runtime service(CRI), -container-runtime=remote --runtime-request-timeout=15m --container-runtime-endpoint=unix:///var/containerd/containerd.sock
kubelet.PreInitRuntimeService()-->>
remote.NewRemoteRuntimeService(remoteRuntimeEndpoint, kubeCfg.RuntimeRequestTimeout.Duration)
RunKubelet(s, kubeDeps, s.RunOnce)-->>
createAndInitKubelet()-->>
kubelet.NewMainKubelet()-->>
makePodSourceConfig(kubeCfg, kubeDeps, nodeName, bootstrapCheckpointPath)-->>
updatechannel = cfg.Channel(kubetypes.ApiserverSource)
klet := &Kubelet{}
//*******init volume plugins
runtime, err := kuberuntime.NewKubeGenericRuntimeManager()
NewInitializedVolumePluginMgr()-->>
kvh.volumePluginMgr.InitPlugins(plugins, prober, kvh)-->>
startKubelet()-->>
k.Run(podCfg.Updates())-->>
//*******run volume manager, and reconcile function would attach volume for attachable plugin and mount volume
go kl.volumeManager.Run(kl.sourcesReady, wait.NeverStop)-->>
vm.desiredStateOfWorldPopulator.Run(sourcesReady, stopCh)-->>
populatorLoop-->>
dswp.findAndAddNewPods()-->>
dswp.processPodVolumes(pod, mountedVolumesForPod, processedVolumesForFSResize)-->>
mounts, devices := util.GetPodVolumeNames(pod)
dswp.createVolumeSpec(podVolume, pod.Name, pod.Namespace, mounts, devices)
dswp.desiredStateOfWorld.AddPodToVolume(uniquePodName, pod, volumeSpec, podVolume.Name, volumeGidValue)-->>
dsw.volumesToMount[volumeName] = volumeToMount{}
vm.reconciler.Run(stopCh)-->>
reconciliationLoopFunc() -->> // reconcile every 100 ms
mountAttachVolumes-->>
rc.desiredStateOfWorld.GetVolumesToMount()
rc.operationExecutor.AttachVolume()-->> // attachable plugin, e.g. CSI plugin
operationGenerator.GenerateAttachVolumeFunc(volumeToAttach, actualStateOfWorld).Run()
rc.operationExecutor.MountVolume()-->> // volume need to mount, like ceph, configmap, emptyDir
oe.operationGenerator.GenerateMapVolumeFunc().Run()-->>
volumePlugin, err := og.volumePluginMgr.FindPluginBySpec(volumeToMount.VolumeSpec)
volumePlugin.NewMounter()
volumeMounter.SetUp()
kl.syncLoop(updates, kl)-->>
kl.syncLoopIteration(updates, handler, syncTicker.C, housekeepingTicker.C, plegCh)-->>
//*******handle pod creation event
handler.HandlePodAdditions(u.Pods)-->>
kl.podManager.AddPod(pod)
kl.canAdmitPod(activePods, pod) // check admit, if admit check fail, it will error out
kl.dispatchWork(pod, kubetypes.SyncPodCreate, mirrorPod, start)-->>
kl.podWorkers.UpdatePod()-->>
p.managePodLoop(podUpdates)-->>
p.syncPodFn()-->>
kubelet.go:syncPod()-->>
runnable := kl.canRunPod(pod) // check soft admin, pod will be pending if check fails
kl.runtimeState.networkErrors() // check network plugin status
kl.containerManager.UpdateQOSCgroups()
kl.makePodDataDirs(pod)
kl.volumeManager.WaitForAttachAndMount(pod)
(kubeGenericRuntimeManager)kl.containerRuntime.SyncPod()-->>
m.computePodActions(pod, podStatus) // create sandbox container?
m.createPodSandbox(pod, podContainerChanges.Attempt)-->>
m.osInterface.MkdirAll(podSandboxConfig.LogDirectory, 0755)
//*******calling CRI
m.runtimeService.RunPodSandbox(podSandboxConfig, runtimeHandler)-->>// k8s.io/cri-api/pkg/apis/runtime/v1alpha2/api.pb.go
c.cc.Invoke(ctx, "/runtime.v1alpha2.RuntimeService/RunPodSandbox") // call remote runtime service which is served by containerd
start("init container", containerStartSpec(container))-->>
startContainer()
start("container", containerStartSpec(&pod.Spec.Containers[idx]))-->>
startContainer()-->>
m.imagePuller.EnsureImageExists(pod, container, pullSecrets, podSandboxConfig)-->>
c.cc.Invoke(ctx, "/runtime.v1alpha2.ImageService/PullImage"
m.runtimeService.CreateContainer(podSandboxID, containerConfig, podSandboxConfig)-->>
c.cc.Invoke(ctx, "/runtime.v1alpha2.RuntimeService/CreateContainer")
m.internalLifecycle.PreStartContainer(pod, container, containerID) // set cpu set
m.runtimeService.StartContainer(containerID)-->>
c.cc.Invoke(ctx, "/runtime.v1alpha2.RuntimeService/StartContainer")
containerd
server.go:service.Register(grpcServer)-->>
runtime.RegisterRuntimeServiceServer(s, instrumented)-->>
s.RegisterService(&_RuntimeService_serviceDesc, srv)-->>
Handler: _RuntimeService_RunPodSandbox_Handler, // api.pb.go
srv.(RuntimeServiceServer).RunPodSandbox(ctx, in)-->> //"/runtime.v1alpha2.RuntimeService/RunPodSandbox"
RunPodSandbox()-->> // pkg/server/sandbox_run.go
sandboxstore.NewSandbox()
c.ensureImageExists()
c.getSandboxRuntime(config, r.GetRuntimeHandler())
netns.NewNetNS()
c.setupPodNetwork(ctx, &sandbox)-->>
c.netPlugin.Setup(ctx, id, path, opts...)-->>
network.Attach(ctx, ns)-->>
n.cni.AddNetworkList(ctx, n.config, ns.config(n.ifName))-->>
c.addNetwork(ctx, list.Name, list.CNIVersion, net, result, rt)-->>// for each network plugin
c.exec.FindInPath(net.Network.Type, c.Path)
buildOneConfig(name, cniVersion, net, prevResult, rt)
invoke.ExecPluginWithResult(ctx, pluginPath, newConf.Bytes, c.args("ADD", rt), c.exec)
c.client.NewContainer(ctx, id, opts...)
c.os.MkdirAll(sandboxRootDir, 0755)
c.os.MkdirAll(volatileSandboxRootDir, 0755)
c.setupSandboxFiles(id, config)
container.NewTask(ctx, containerdio.NullIO, taskOpts...)
task.Start(ctx)-->>
c.client.TaskService().Create(ctx, request)-->>
s.local.Create(ctx, r)-->>
l.getRuntime(container.Runtime.Name)
rtime.Create(ctx, r.ContainerID, opts)-->>
b := shimBinary(ctx, bundle, opts.Runtime, m.containerdAddress, m.containerdTTRPCAddress, m.events, m.tasks)
b.Start()
shim.Create(ctx, opts)-->>
c.client.Call(ctx, "containerd.task.v2.Task", "Create", req, &resp)
runtime.RegisterImageServiceServer(s, instrumented)
```