taintManager的主要功能為：當某個node被打上NoExecute污點后，其上面的pod如果不能容忍該污點，則taintManager將會驅逐這些pod，而新建的pod也需要容忍該污點才能調度到該node上；

通過kcm啟動參數(shù)--enable-taint-manager來確定是否啟動taintManager，true時啟動（啟動參數(shù)默認值為true）；

【資料圖】

kcm啟動參數(shù)--feature-gates=TaintBasedEvictions=xxx，默認值true，配合--enable-taint-manager共同作用，兩者均為true，才會開啟污點驅逐；

當node出現(xiàn)NoExecute污點時，判斷node上的pod是否能容忍node的污點，不能容忍的pod，會被立即刪除，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，pod被刪除；

NoExecuteTaintManager結構體為taintManager的主要結構體，其主要屬性有：（1）taintEvictionQueue：不能容忍node上NoExecute的污點的pod，會被加入到該隊列中，然后pod會被刪除；（2）taintedNodes：記錄了每個node的taint；（3）nodeUpdateQueue：當node對象發(fā)生add、delete、update（新舊node對象的taint不相同）事件時，node會進入該隊列；（4）podUpdateQueue：當pod對象發(fā)生add、delete、update（新舊pod對象的NodeName或Tolerations不相同）事件時，pod會進入該隊列；（5）nodeUpdateChannels：nodeUpdateChannels即8個nodeUpdateItem類型的channel，有worker負責消費nodeUpdateQueue隊列，然后根據(jù)node name計算出index，把node放入其中1個nodeUpdateItem類型的channel中；（6）podUpdateChannels：podUpdateChannels即8個podUpdateItem類型的channel，有worker負責消費podUpdateQueue隊列，然后根據(jù)pod的node name計算出index，把pod放入其中1個podUpdateItem類型的channel中；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gotype NoExecuteTaintManager struct {client                clientset.Interfacerecorder              record.EventRecordergetPod                GetPodFuncgetNode               GetNodeFuncgetPodsAssignedToNode GetPodsByNodeNameFunctaintEvictionQueue *TimedWorkerQueue// keeps a map from nodeName to all noExecute taints on that NodetaintedNodesLock sync.MutextaintedNodes     map[string][]v1.TaintnodeUpdateChannels []chan nodeUpdateItempodUpdateChannels  []chan podUpdateItemnodeUpdateQueue workqueue.InterfacepodUpdateQueue  workqueue.Interface}

taintEvictionQueue屬性是一個TimedWorkerQueue類型的隊列，調用tc.taintEvictionQueue.AddWork，會將pod添加到該隊列中，會添加一個定時器，然后到期之后會自動執(zhí)行workFunc，初始化taintEvictionQueue時，傳入的workFunc是deletePodHandler函數(shù)，作用是刪除pod；

所以進入taintEvictionQueue中的pod，會在設置好的時間，被刪除；

pod.Spec.Tolerations配置的是pod的污點容忍信息；

// vendor/k8s.io/api/core/v1/types.gotype Toleration struct {Key string `json:"key,omitempty" protobuf:"bytes,1,opt,name=key"`Operator TolerationOperator `json:"operator,omitempty" protobuf:"bytes,2,opt,name=operator,casttype=TolerationOperator"`Value string `json:"value,omitempty" protobuf:"bytes,3,opt,name=value"`Effect TaintEffect `json:"effect,omitempty" protobuf:"bytes,4,opt,name=effect,casttype=TaintEffect"`TolerationSeconds *int64 `json:"tolerationSeconds,omitempty" protobuf:"varint,5,opt,name=tolerationSeconds"`}

Tolerations的屬性值解析如下：（1）Key：匹配node污點的Key；（2）Operator：表示Tolerations中Key與node污點的Key相同時，其Value與node污點的Value的關系，默認值Equal，代表相等，Exists則代表Tolerations中Key與node污點的Key相同即可，不用比較其Value值；（3）Value：匹配node污點的Value；（4）Effect：匹配node污點的Effect；（5）TolerationSeconds：node污點容忍時間；

配置示例：

tolerations:- key: "key1"  operator: "Equal"  value: "value1"  effect: "NoExecute"  tolerationSeconds: 3600

上述配置表示如果該pod正在運行，同時一個匹配的污點被添加到其所在的node節(jié)點上，那么該pod還將繼續(xù)在節(jié)點上運行3600秒，然后會被驅逐（如果在此之前其匹配的node污點被刪除了，則該pod不會被驅逐）；

NewNodeLifecycleController為NodeLifecycleController的初始化函數(shù)，里面給taintManager注冊了pod與node的EventHandler，Add、Update、Delete事件都會調用taintManager的PodUpdated或NodeUpdated方法來做處理；

// pkg/controller/nodelifecycle/node_lifecycle_controller.gofunc NewNodeLifecycleController(    ...    podInformer.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{AddFunc: func(obj interface{}) {...if nc.taintManager != nil {nc.taintManager.PodUpdated(nil, pod)}},UpdateFunc: func(prev, obj interface{}) {...if nc.taintManager != nil {nc.taintManager.PodUpdated(prevPod, newPod)}},DeleteFunc: func(obj interface{}) {...if nc.taintManager != nil {nc.taintManager.PodUpdated(pod, nil)}},})    ...    if nc.runTaintManager {podGetter := func(name, namespace string) (*v1.Pod, error) { return nc.podLister.Pods(namespace).Get(name) }nodeLister := nodeInformer.Lister()nodeGetter := func(name string) (*v1.Node, error) { return nodeLister.Get(name) }nc.taintManager = scheduler.NewNoExecuteTaintManager(kubeClient, podGetter, nodeGetter, nc.getPodsAssignedToNode)nodeInformer.Informer().AddEventHandler(cache.ResourceEventHandlerFuncs{AddFunc: nodeutil.CreateAddNodeHandler(func(node *v1.Node) error {nc.taintManager.NodeUpdated(nil, node)return nil}),UpdateFunc: nodeutil.CreateUpdateNodeHandler(func(oldNode, newNode *v1.Node) error {nc.taintManager.NodeUpdated(oldNode, newNode)return nil}),DeleteFunc: nodeutil.CreateDeleteNodeHandler(func(node *v1.Node) error {nc.taintManager.NodeUpdated(node, nil)return nil}),})}...}

tc.NodeUpdated方法會判斷新舊node對象的taint是否相同，不相同則調用tc.nodeUpdateQueue.Add，將該node放入到nodeUpdateQueue隊列中；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) NodeUpdated(oldNode *v1.Node, newNode *v1.Node) {nodeName := ""oldTaints := []v1.Taint{}if oldNode != nil {nodeName = oldNode.NameoldTaints = getNoExecuteTaints(oldNode.Spec.Taints)}newTaints := []v1.Taint{}if newNode != nil {nodeName = newNode.NamenewTaints = getNoExecuteTaints(newNode.Spec.Taints)}if oldNode != nil && newNode != nil && helper.Semantic.DeepEqual(oldTaints, newTaints) {return}updateItem := nodeUpdateItem{nodeName: nodeName,}tc.nodeUpdateQueue.Add(updateItem)}

tc.PodUpdated方法會判斷新舊pod對象的NodeName或Tolerations是否相同，不相同則調用tc.podUpdateQueue.Add，將該pod放入到podUpdateQueue隊列中；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) PodUpdated(oldPod *v1.Pod, newPod *v1.Pod) {podName := ""podNamespace := ""nodeName := ""oldTolerations := []v1.Toleration{}if oldPod != nil {podName = oldPod.NamepodNamespace = oldPod.NamespacenodeName = oldPod.Spec.NodeNameoldTolerations = oldPod.Spec.Tolerations}newTolerations := []v1.Toleration{}if newPod != nil {podName = newPod.NamepodNamespace = newPod.NamespacenodeName = newPod.Spec.NodeNamenewTolerations = newPod.Spec.Tolerations}if oldPod != nil && newPod != nil && helper.Semantic.DeepEqual(oldTolerations, newTolerations) && oldPod.Spec.NodeName == newPod.Spec.NodeName {return}updateItem := podUpdateItem{podName:      podName,podNamespace: podNamespace,nodeName:     nodeName,}tc.podUpdateQueue.Add(updateItem)}

看到TaintManager的初始化方法NewNoExecuteTaintManager中，調用CreateWorkerQueue給taintEvictionQueue做了初始化；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc NewNoExecuteTaintManager(...) ... {    ...    tm.taintEvictionQueue = CreateWorkerQueue(deletePodHandler(c, tm.emitPodDeletionEvent))    ...}

CreateWorkerQueue函數(shù)初始化并返回TimedWorkerQueue結構體；

// pkg/controller/nodelifecycle/scheduler/timed_workers.gofunc CreateWorkerQueue(f func(args *WorkArgs) error) *TimedWorkerQueue {return &TimedWorkerQueue{workers:  make(map[string]*TimedWorker),workFunc: f,}}

初始化taintEvictionQueue時傳入了deletePodHandler作為隊列中元素的處理方法；deletePodHandler函數(shù)的主要邏輯是請求apiserver，刪除pod對象，所以說，被放入到taintEvictionQueue隊列中的pod，會被刪除；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc deletePodHandler(c clientset.Interface, emitEventFunc func(types.NamespacedName)) func(args *WorkArgs) error {return func(args *WorkArgs) error {ns := args.NamespacedName.Namespacename := args.NamespacedName.Nameklog.V(0).Infof("NoExecuteTaintManager is deleting Pod: %v", args.NamespacedName.String())if emitEventFunc != nil {emitEventFunc(args.NamespacedName)}var err errorfor i := 0; i < retries; i++ {err = c.CoreV1().Pods(ns).Delete(name, &metav1.DeleteOptions{})if err == nil {break}time.Sleep(10 * time.Millisecond)}return err}}

再來看一下tc.taintEvictionQueue.AddWork方法，作用是添加pod進入taintEvictionQueue隊列，即調用CreateWorker給該pod創(chuàng)建一個worker來刪除該pod；

// pkg/controller/nodelifecycle/scheduler/timed_workers.gofunc (q *TimedWorkerQueue) AddWork(args *WorkArgs, createdAt time.Time, fireAt time.Time) {key := args.KeyFromWorkArgs()klog.V(4).Infof("Adding TimedWorkerQueue item %v at %v to be fired at %v", key, createdAt, fireAt)q.Lock()defer q.Unlock()if _, exists := q.workers[key]; exists {klog.Warningf("Trying to add already existing work for %+v. Skipping.", args)return}worker := CreateWorker(args, createdAt, fireAt, q.getWrappedWorkerFunc(key))q.workers[key] = worker}

CreateWorker函數(shù)會先判斷是否應該立即執(zhí)行workFunc，是的話立即拉起一個goroutine來執(zhí)行workFunc并返回，否則定義一個timer定時器，到時間后自動拉起一個goroutine執(zhí)行workFunc；

// pkg/controller/nodelifecycle/scheduler/timed_workers.gofunc CreateWorker(args *WorkArgs, createdAt time.Time, fireAt time.Time, f func(args *WorkArgs) error) *TimedWorker {delay := fireAt.Sub(createdAt)if delay <= 0 {go f(args)return nil}timer := time.AfterFunc(delay, func() { f(args) })return &TimedWorker{WorkItem:  args,CreatedAt: createdAt,FireAt:    fireAt,Timer:     timer,}}

tc.taintEvictionQueue.AddWork方法，作用是停止對應的pod的timer，即停止執(zhí)行對應pod的workFunc（不刪除pod）；

// pkg/controller/nodelifecycle/scheduler/timed_workers.gofunc (w *TimedWorker) Cancel() {if w != nil {w.Timer.Stop()}}

nc.taintManager.Run為taintManager的啟動方法，處理邏輯都在這，主要是判斷node上的pod是否能容忍node的NoExecute污點，不能容忍的pod，會被刪除，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，被刪除；

主要邏輯：（1）創(chuàng)建8個類型為nodeUpdateItem的channel（緩沖區(qū)大小10），并賦值給tc.nodeUpdateChannels；創(chuàng)建8個類型為podUpdateItem的channel（緩沖區(qū)大小1），并賦值給podUpdateChannels；

（2）消費tc.nodeUpdateQueue隊列，根據(jù)node name計算hash，將node放入對應的tc.nodeUpdateChannels[hash]中；

（3）消費tc.podUpdateQueue隊列，根據(jù)pod的node name計算hash，將node放入對應的tc.podUpdateChannels[hash]中；

（4）啟動8個goroutine，調用tc.worker對其中一個tc.nodeUpdateChannels與tc.podUpdateChannels做處理，判斷node上的pod是否能容忍node的NoExecute污點，不能容忍的pod，會被刪除，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，被刪除；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) Run(stopCh <-chan struct{}) {klog.V(0).Infof("Starting NoExecuteTaintManager")for i := 0; i < UpdateWorkerSize; i++ {tc.nodeUpdateChannels = append(tc.nodeUpdateChannels, make(chan nodeUpdateItem, NodeUpdateChannelSize))tc.podUpdateChannels = append(tc.podUpdateChannels, make(chan podUpdateItem, podUpdateChannelSize))}// Functions that are responsible for taking work items out of the workqueues and putting them// into channels.go func(stopCh <-chan struct{}) {for {item, shutdown := tc.nodeUpdateQueue.Get()if shutdown {break}nodeUpdate := item.(nodeUpdateItem)hash := hash(nodeUpdate.nodeName, UpdateWorkerSize)select {case <-stopCh:tc.nodeUpdateQueue.Done(item)returncase tc.nodeUpdateChannels[hash] <- nodeUpdate:// tc.nodeUpdateQueue.Done is called by the nodeUpdateChannels worker}}}(stopCh)go func(stopCh <-chan struct{}) {for {item, shutdown := tc.podUpdateQueue.Get()if shutdown {break}// The fact that pods are processed by the same worker as nodes is used to avoid races// between node worker setting tc.taintedNodes and pod worker reading this to decide// whether to delete pod.// It"s possible that even without this assumption this code is still correct.podUpdate := item.(podUpdateItem)hash := hash(podUpdate.nodeName, UpdateWorkerSize)select {case <-stopCh:tc.podUpdateQueue.Done(item)returncase tc.podUpdateChannels[hash] <- podUpdate:// tc.podUpdateQueue.Done is called by the podUpdateChannels worker}}}(stopCh)wg := sync.WaitGroup{}wg.Add(UpdateWorkerSize)for i := 0; i < UpdateWorkerSize; i++ {go tc.worker(i, wg.Done, stopCh)}wg.Wait()}

tc.worker方法負責消費nodeUpdateChannels和podUpdateChannels，分別調用tc.handleNodeUpdate和tc.handlePodUpdate方法做進一步處理；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) worker(worker int, done func(), stopCh <-chan struct{}) {defer done()// When processing events we want to prioritize Node updates over Pod updates,// as NodeUpdates that interest NoExecuteTaintManager should be handled as soon as possible -// we don"t want user (or system) to wait until PodUpdate queue is drained before it can// start evicting Pods from tainted Nodes.for {select {case <-stopCh:returncase nodeUpdate := <-tc.nodeUpdateChannels[worker]:tc.handleNodeUpdate(nodeUpdate)tc.nodeUpdateQueue.Done(nodeUpdate)case podUpdate := <-tc.podUpdateChannels[worker]:// If we found a Pod update we need to empty Node queue first.priority:for {select {case nodeUpdate := <-tc.nodeUpdateChannels[worker]:tc.handleNodeUpdate(nodeUpdate)tc.nodeUpdateQueue.Done(nodeUpdate)default:break priority}}// After Node queue is emptied we process podUpdate.tc.handlePodUpdate(podUpdate)tc.podUpdateQueue.Done(podUpdate)}}}

tc.handleNodeUpdate方法主要是判斷node上的pod是否能容忍node的NoExecute污點，不能容忍的pod，會被刪除，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，被刪除；

主要邏輯：（1）從informer本地緩存中獲取node對象；（2）從node.Spec.Taints中獲取NoExecute的taints；（3）將該node的NoExecute的taints更新到tc.taintedNodes中；（4）調用tc.getPodsAssignedToNode，獲取該node上的所有pod，如果pod數(shù)量為0，直接return；（5）如果node的NoExecute的taints數(shù)量為0，則遍歷該node上所有pod，調用tc.cancelWorkWithEvent，將該pod從taintEvictionQueue隊列中移除，然后直接return；（6）遍歷該node上所有pod，調用tc.processPodOnNode，對pod做進一步處理；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) handleNodeUpdate(nodeUpdate nodeUpdateItem) {node, err := tc.getNode(nodeUpdate.nodeName)if err != nil {if apierrors.IsNotFound(err) {// Deleteklog.V(4).Infof("Noticed node deletion: %#v", nodeUpdate.nodeName)tc.taintedNodesLock.Lock()defer tc.taintedNodesLock.Unlock()delete(tc.taintedNodes, nodeUpdate.nodeName)return}utilruntime.HandleError(fmt.Errorf("cannot get node %s: %v", nodeUpdate.nodeName, err))return}// Create or Updateklog.V(4).Infof("Noticed node update: %#v", nodeUpdate)taints := getNoExecuteTaints(node.Spec.Taints)func() {tc.taintedNodesLock.Lock()defer tc.taintedNodesLock.Unlock()klog.V(4).Infof("Updating known taints on node %v: %v", node.Name, taints)if len(taints) == 0 {delete(tc.taintedNodes, node.Name)} else {tc.taintedNodes[node.Name] = taints}}()// This is critical that we update tc.taintedNodes before we call getPodsAssignedToNode:// getPodsAssignedToNode can be delayed as long as all future updates to pods will call// tc.PodUpdated which will use tc.taintedNodes to potentially delete delayed pods.pods, err := tc.getPodsAssignedToNode(node.Name)if err != nil {klog.Errorf(err.Error())return}if len(pods) == 0 {return}// Short circuit, to make this controller a bit faster.if len(taints) == 0 {klog.V(4).Infof("All taints were removed from the Node %v. Cancelling all evictions...", node.Name)for i := range pods {tc.cancelWorkWithEvent(types.NamespacedName{Namespace: pods[i].Namespace, Name: pods[i].Name})}return}now := time.Now()for _, pod := range pods {podNamespacedName := types.NamespacedName{Namespace: pod.Namespace, Name: pod.Name}tc.processPodOnNode(podNamespacedName, node.Name, pod.Spec.Tolerations, taints, now)}}

tc.processPodOnNode方法主要作用是判斷pod是否能容忍node上所有的NoExecute的污點，如果不能，則將該pod加到taintEvictionQueue隊列中，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，加到taintEvictionQueue隊列中；

主要邏輯：（1）如果node的NoExecute的taints數(shù)量為0，則調用tc.cancelWorkWithEvent，將該pod從taintEvictionQueue隊列中移除；（2）調用v1helper.GetMatchingTolerations，判斷pod是否容忍node上所有的NoExecute的taints，以及獲取能容忍taints的容忍列表；（3）如果不能容忍所有污點，則調用tc.taintEvictionQueue.AddWork，將該pod加到taintEvictionQueue隊列中；（4）如果能容忍所有污點，則等待所有污點的容忍時間里最小值后，再調用tc.taintEvictionQueue.AddWork，將該pod加到taintEvictionQueue隊列中；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) processPodOnNode(podNamespacedName types.NamespacedName,nodeName string,tolerations []v1.Toleration,taints []v1.Taint,now time.Time,) {if len(taints) == 0 {tc.cancelWorkWithEvent(podNamespacedName)}allTolerated, usedTolerations := v1helper.GetMatchingTolerations(taints, tolerations)if !allTolerated {klog.V(2).Infof("Not all taints are tolerated after update for Pod %v on %v", podNamespacedName.String(), nodeName)// We"re canceling scheduled work (if any), as we"re going to delete the Pod right away.tc.cancelWorkWithEvent(podNamespacedName)tc.taintEvictionQueue.AddWork(NewWorkArgs(podNamespacedName.Name, podNamespacedName.Namespace), time.Now(), time.Now())return}minTolerationTime := getMinTolerationTime(usedTolerations)// getMinTolerationTime returns negative value to denote infinite toleration.if minTolerationTime < 0 {klog.V(4).Infof("New tolerations for %v tolerate forever. Scheduled deletion won"t be cancelled if already scheduled.", podNamespacedName.String())return}startTime := nowtriggerTime := startTime.Add(minTolerationTime)scheduledEviction := tc.taintEvictionQueue.GetWorkerUnsafe(podNamespacedName.String())if scheduledEviction != nil {startTime = scheduledEviction.CreatedAtif startTime.Add(minTolerationTime).Before(triggerTime) {return}tc.cancelWorkWithEvent(podNamespacedName)}tc.taintEvictionQueue.AddWork(NewWorkArgs(podNamespacedName.Name, podNamespacedName.Namespace), startTime, triggerTime)}

tc.handlePodUpdate方法最終也是調用了tc.processPodOnNode對pod做進一步處理；

tc.processPodOnNode方法在上面已經分析過了，這里不再進行分析；

主要邏輯：（1）從informer本地緩存中獲取pod對象；（2）獲取pod的node name，如果為空，直接return；（3）根據(jù)node name從tc.taintedNodes中獲取node的污點，如果污點為空，直接return；（4）調用tc.processPodOnNode對pod做進一步處理；

// pkg/controller/nodelifecycle/scheduler/taint_manager.gofunc (tc *NoExecuteTaintManager) handlePodUpdate(podUpdate podUpdateItem) {pod, err := tc.getPod(podUpdate.podName, podUpdate.podNamespace)if err != nil {if apierrors.IsNotFound(err) {// DeletepodNamespacedName := types.NamespacedName{Namespace: podUpdate.podNamespace, Name: podUpdate.podName}klog.V(4).Infof("Noticed pod deletion: %#v", podNamespacedName)tc.cancelWorkWithEvent(podNamespacedName)return}utilruntime.HandleError(fmt.Errorf("could not get pod %s/%s: %v", podUpdate.podName, podUpdate.podNamespace, err))return}// We key the workqueue and shard workers by nodeName. If we don"t match the current state we should not be the one processing the current object.if pod.Spec.NodeName != podUpdate.nodeName {return}// Create or UpdatepodNamespacedName := types.NamespacedName{Namespace: pod.Namespace, Name: pod.Name}klog.V(4).Infof("Noticed pod update: %#v", podNamespacedName)nodeName := pod.Spec.NodeNameif nodeName == "" {return}taints, ok := func() ([]v1.Taint, bool) {tc.taintedNodesLock.Lock()defer tc.taintedNodesLock.Unlock()taints, ok := tc.taintedNodes[nodeName]return taints, ok}()// It"s possible that Node was deleted, or Taints were removed before, which triggered// eviction cancelling if it was needed.if !ok {return}tc.processPodOnNode(podNamespacedName, nodeName, pod.Spec.Tolerations, taints, time.Now())}

taintManager的主要功能為：當某個node被打上NoExecute污點后，其上面的pod如果不能容忍該污點，則taintManager將會驅逐這些pod，而新建的pod也需要容忍該污點才能調度到該node上；

通過kcm啟動參數(shù)--enable-taint-manager來確定是否啟動taintManager，true時啟動（啟動參數(shù)默認值為true）；

kcm啟動參數(shù)--feature-gates=TaintBasedEvictions=xxx，默認值true，配合--enable-taint-manager共同作用，兩者均為true，才會開啟污點驅逐；

當node出現(xiàn)NoExecute污點時，判斷node上的pod是否能容忍node的污點，不能容忍的pod，會被立即刪除，能容忍所有污點的pod，則等待所有污點的容忍時間里最小值后，pod被刪除；