Kubelet VolumeManager机制流程分析

// volumeManager implements the VolumeManager interfacetype 
volumeManager struct {  
   // kubeClient is the kube API client 
   kubeClient clientset.Interface  
 
   // 实现初始化好的插件管理对象
   volumePluginMgr *volume.VolumePluginMgr  
  
   // 包含了vm想要达成的一个状态：那些卷应该被连接，那个pods会引用这些卷。从 podManager获取数据
   desiredStateOfWorld cache.DesiredStateOfWorld  
  
   // 包含了实际的vm状态。在  attach,  detach, mount, and unmount  操作的时候记录这些信息
   actualStateOfWorld cache.ActualStateOfWorld  
     
   // 异步执行 attach, detach, mount,unmount操作
   operationExecutor operationexecutor.OperationExecutor  
  
   // 运行异步定期循环，通过使用operationExecutor 来协调desiredStateOfWorld与actualStateOfWorld
   reconciler reconciler.Reconciler  
  
   //  运行异步定期循环，使用 PodManager 填充 desiredStateOfWorld 
   desiredStateOfWorldPopulator populator.DesiredStateOfWorldPopulator  
  
   // csiMigratedPluginManager keeps track of CSI migration status of plugins   
   csiMigratedPluginManager csimigration.PluginManager  
  
   // intreeToCSITranslator translates in-tree volume specs to CSI   
   intreeToCSITranslator csimigration.InTreeToCSITranslator  
}

type actualStateOfWorld struct {
	nodeName types.NodeName
	// 存储所有卷挂载情况
	attachedVolumes map[v1.UniqueVolumeName]attachedVolume
		-> mountedPods map[volumetypes.UniquePodName]mountedPod
}

type desiredStateOfWorld struct {  
	// 期望挂载的情况
   volumesToMount map[v1.UniqueVolumeName]volumeToMount
	   -> podsToMount map[types.UniquePodName]podToMount
}

func (rc *reconciler) reconcile() {  
   // Unmounts are triggered before mounts so that a volume that was  
   // referenced by a pod that was deleted and is now referenced by another   
   // pod is unmounted from the first pod before being mounted to the new  
   // pod.   
   // 遍历 node 上的 mountedVolume，检查 pod 还在不在，如果 pod 不在了就把对应 volume unmount
   rc.unmountVolumes()  
  
   // Next we mount required volumes. This function could also trigger  
   // attach if kubelet is responsible for attaching volumes.   // If underlying PVC was resized while in-use then this function also handles volume   
   // resizing.   
   // 遍历需要 mount 或者 attach 的 volume，看下是不是真的 mount 或者 attach了，没有就执行 mount 或者 attach.设备需要先 attach，因此 kubelet 这里会一直阻塞，产生一个叫做 VolumeNotAttached 的错误
   rc.mountOrAttachVolumes()  
  
   // Ensure devices that should be detached/unmounted are detached/unmounted.  
   // unmountDetachDevices：如果某个设备的多个 volume 都是 unmount 状态那就把该设备 detach 掉
   rc.unmountDetachDevices()  
  
   // After running the above operations if skippedDuringReconstruction is not empty  
   // then ensure that all volumes which were discovered and skipped during reconstruction   
   // are added to actualStateOfWorld in uncertain state.   
   if len(rc.skippedDuringReconstruction) > 0 {  
      rc.processReconstructedVolumes()  
   }  
}

wait.PollUntil(dswp.loopSleepDuration, func() (bool, error) {  
   done := sourcesReady.AllReady()  
   dswp.populatorLoop()  
	   -> {
			   dswp.findAndAddNewPods()
			   dswp.findAndRemoveDeletedPods()
			}
   return done, nil  
}, stopCh)

func (dswp *desiredStateOfWorldPopulator) findAndAddNewPods() {
	//...
	for _, pod := range dswp.podManager.GetPods() {  
	   //...
	   // 填充 desiredStateOfWorld 
	   dswp.processPodVolumes(pod, mountedVolumesForPod)
		//...  
		-> uniqueVolumeName, err := dswp.desiredStateOfWorld.AddPodToVolume(  
   uniquePodName, pod, volumeSpec, podVolume.Name, volumeGidValue, seLinuxContainerContexts[podVolume.Name])
	}
}

func (dswp *desiredStateOfWorldPopulator) findAndRemoveDeletedPods() {
	//...
	pod, podExists := dswp.podManager.GetPodByUID(volumeToMount.Pod.UID)  
	if podExists {
		// ...
		// 判断pod中容器是否都已经终止
		if !dswp.podStateProvider.ShouldPodRuntimeBeRemoved(pod.UID) {  
		   continue  
		}
		// 是否保持终止pod的卷  node上注解KeepTerminatedPodVolumesAnnotation
		if dswp.keepTerminatedPodVolumes {  
		   continue  
		}
	}
	removed := dswp.actualStateOfWorld.PodRemovedFromVolume(volumeToMount.PodName, volumeToMount.VolumeName)
	if removed && podExists {
		continue
	}
	// 删除期望的状态，以及相关记录
	dswp.desiredStateOfWorld.DeletePodFromVolume(  
volumeToMount.PodName, volumeToMount.VolumeName)  
	dswp.deleteProcessedPod(volumeToMount.PodName)
}

func (oe *operationExecutor) MountVolume(...) error {
	-> func (og *operationGenerator) GenerateDetachVolumeFunc(...)(...) {
		volumePlugin, err :=  
	   og.volumePluginMgr.FindPluginBySpec(volumeToMount.VolumeSpec)
		...
		mountVolumeFunc := func() volumetypes.OperationContext {
		   // 获取挂载卷插件
			volumePlugin, err := og.volumePluginMgr.FindPluginBySpec(volumeToMount.VolumeSpec)
			...
			// 检查node亲和性  
			affinityErr := checkNodeAffinity(og, volumeToMount)
			// 获取mounter，所有的挂载类型都拥有
			volumeMounter, newMounterErr := volumePlugin.NewMounter(...)
			//获取attacher，所有的卷都需要挂载以及接触挂载，但是不是所有的都需要attacher  
			attachableVolumePlugin, _ := og.volumePluginMgr.FindAttachablePluginBySpec(volumeToMount.VolumeSpec)
			...
			deviceMountableVolumePlugin, _ := og.volumePluginMgr.FindDeviceMountablePluginBySpec(volumeToMount.VolumeSpec)
			// 核心 设置挂载文件（创建路径并且写入内容）
			mountErr := volumeMounter.SetUp(...)
				// setup失败后会删除卷
				-> func (b *configMapVolumeMounter) SetUpAt(dir string, mounterArgs volume.MounterArgs) error {
					...
					// 获取configmap资源，如果不是可选项找不到的时候会报configmap not fount错误
					configMap, err := b.getConfigMap(b.pod.Namespace, b.source.Name)
					// 拼装payload  map[string]volumeutil.FileProjection对象（文件名和文件投影信息）,但没有指定mode的时候采用默认模式
					payload, err := MakePayload(b.source.Items, configMap, b.source.DefaultMode, optional)
					// 创建挂载点，获取当前卷的所有挂载路径
					// 1. 不能再rootfs之外创建目录
					// 2. 检查每个装载点，看看它是否嵌套在这个卷下面
					volumeutil.MakeNestedMountpoints(b.volName, dir, b.pod)
				}
		}
	}
}