Kubernetes Informer基本原理

本文分析 k8s controller 中 informer 启动的基本流程

不论是 k8s 自身组件,还是自己编写 controller,都需要通过 apiserver 监听 etcd 事件来完成自己的控制循环逻辑。

如何高效可靠进行事件监听,k8s 客户端工具包 client-go 提供了一个通用的 informer 包,通过 informer,可以方便和高效的进行 controller 开发。

informer 包提供了如下的一些功能:

1、本地缓存(store)

2、索引机制(indexer)

3、Handler 注册功能(eventHandler)

1、informer 架构

整个 informer 机制架构如下图(图片源自 Client-go):

图片[1]-Kubernetes Informer基本原理-不念博客

可以看到这张图分为上下两个部分,上半部分由 client-go 提供,下半部分则是需要自己实现的控制循环逻辑

本文主要分析上半部分的逻辑,包括下面几个组件:

1.1、Reflector:

从图上可以看到 Reflector 是一个和 apiserver 交互的组件,通过 list 和 watch api 将资源对象压入队列

1.2、DeltaFifo:

DeltaFifo的结构体示意如下:

type DeltaFIFO struct {
  ...
  // We depend on the property that items in the s    et are in
  // the queue and vice versa, and that all Deltas in this
  // map have at least one Delta.
  items map[string]Deltas
  queue []string
  ...
}

主要分为两部分,fifo 和 delta

(1)fifo:先进先出队列

对应结构体中的 queue,结构体示例如下:

[default/centos-fd77b5886-pfrgn, xxx, xxx]

(2)delta:对应结构体中的items,存储了资源对象并且携带了资源操作类型的一个 map,结构体示例如下:

map:{"default/centos-fd77b5886-pfrgn":[{Replaced &Pod{ObjectMeta: ${pod参数}], "xxx": [{},{}]}

消费者从 queue 中 pop 出对象进行消费,并从 items 获取具体的消费操作(执行动作 Update/Deleted/Sync,和执行的对象 object spec)

1.3、Indexer:

client-go 用来存储资源对象并自带索引功能的本地存储,deltaFIFO 中 pop 出的对象将存储到 Indexer。

indexer 与 etcd 集群中的数据保持一致,从而 client-go 可以直接从本地缓存获取资源对象,减少 apiserver 和 etcd 集群的压力。

2、一个基本例子

func main() {

  stopCh := make(chan struct{})
  defer close(stopCh)
  
  // (1)New a k8s clientset
  masterUrl := "172.27.32.110:8080"
  config, err := clientcmd.BuildConfigFromFlags(masterUrl, "")
  if err != nil {
    klog.Errorf("BuildConfigFromFlags err, err: %v", err)
  }
  
  clientset, err := k.NewForConfig(config)
  if err != nil {
    klog.Errorf("Get clientset err, err: %v", err)
  }
  
  // (2)New a sharedInformers factory
  sharedInformers := informers.NewSharedInformerFactory(clientset, defaultResync)
  
  
  // (3)Register a informer
  //  f.informers[informerType] = informer,
  //  the detail for informer is build in NewFilteredPodInformer()
  podInformer := sharedInformers.Core().V1().Pods().Informer()
  
  // (4)Register event handler
  podInformer.AddEventHandler(cache.ResourceEventHandlerFuncs{
      AddFunc: func(obj interface{}) {
        mObj := obj.(v1.Object)
        klog.Infof("Get new obj: %v", mObj)
        klog.Infof("Get new obj name: %s", mObj.GetName())
      },
  })
  
  // (5)Start all informers
  sharedInformers.Start(stopCh)
  
  // (6)A cronjob for cache sync
  if !cache.WaitForCacheSync(stopCh, podInformer.HasSynced) {
    klog.Infof("Cache sync fail!")
  }
  
  // (7)Use lister
  podLister := sharedInformers.Core().V1().Pods().Lister()
  pods, err := podLister.List(labels.Everything())
  if err != nil {
    klog.Infof("err: %v", err)
  }
  klog.Infof("len(pods), %d", len(pods))
  for _, v := range pods {
    klog.Infof("pod: %s", v.Name)
  }
  
  <- stopChan
}

上面就是一个简单的 informer 的使用例子,整个过程如上述几个步骤,着重说一下(2)、(3)、(4)、(5)四个步骤

3、流程分析

3.1、New a sharedInformers factory

sharedInformers := informers.NewSharedInformerFactory(clientset, defaultResync)

factory := &sharedInformerFactory{
  client:           client,
  namespace:        v1.NamespaceAll,
  defaultResync:    defaultResync,
  informers:        make(map[reflect.Type]cache.SharedIndexInformer),
  startedInformers: make(map[reflect.Type]bool),
  customResync:     make(map[reflect.Type]time.Duration),
}

这个过程就是创建一个 informer 的工厂 sharedInformerFactory,sharedInformerFactory 中有一个 informers 对象,里面是一个 informer 的 map,sharedInformerFactory 是为了防止过多的重复 informer 监听 apiserver,导致 apiserver 压力过大,在同一个服务中,不同的 controller 使用同一个 informer

3.2、Register a informer

这个过程主要是生成和注册 informer 到 sharedInformerFactory

podInformer := sharedInformers.Core().V1().Pods().Informer()

func (f *podInformer) Informer() cache.SharedIndexInformer {
  return f.factory.InformerFor(&corev1.Pod{}, f.defaultInformer)
}

### f.factory.InformerFor:
### 注册 informer 
func (f *sharedInformerFactory) InformerFor(obj runtime.Object, newFunc internalinterfaces.NewInformerFunc) cache.SharedIndexInformer {
  ...
  informer = newFunc(f.client, resyncPeriod)
  f.informers[informerType] = informer
  return informer
}

### f.defaultInformer:
### 生成 informer
func (f *podInformer) defaultInformer(client k.Interface, resyncPeriod time.Duration) cache.SharedIndexInformer {
  return NewFilteredPodInformer(client, f.namespace, resyncPeriod, cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc}, f.tweakListOptions)
}

func NewFilteredPodInformer(client k.Interface, namespace string, resyncPeriod time.Duration, indexers cache.Indexers, tweakListOptions internalinterfaces.TweakListOptionsFunc) cache.SharedIndexInformer {
  return cache.NewSharedIndexInformer(
    &cache.ListWatch{
    ListFunc: func(options metav1.ListOptions) (runtime.Object, error) {
      if tweakListOptions != nil {
        tweakListOptions(&options)
      }
      return client.CoreV1().Pods(namespace).List(context.TODO(), options)
    },
    WatchFunc: func(options metav1.ListOptions) (watch.Interface, error) {
      if tweakListOptions != nil {
        tweakListOptions(&options)
      }
      return client.CoreV1().Pods(namespace).Watch(context.TODO(), options)
    },
    },
    &corev1.Pod{},
    resyncPeriod,
    indexers,
  )
}

### cache.NewSharedIndexInformer:
func NewSharedIndexInformer(lw ListerWatcher, exampleObject runtime.Object, defaultEventHandlerResyncPeriod time.Duration, indexers Indexers) SharedIndexInformer {
  realClock := &clock.RealClock{}
  sharedIndexInformer := &sharedIndexInformer{
    processor:                       &sharedProcessor{clock: realClock},
    indexer:                         NewIndexer(DeletionHandlingMetaNamespaceKeyFunc, indexers),
    listerWatcher:                   lw,
    objectType:                      exampleObject,
    resyncCheckPeriod:               defaultEventHandlerResyncPeriod,
    defaultEventHandlerResyncPeriod: defaultEventHandlerResyncPeriod,
    cacheMutationDetector:           NewCacheMutationDetector(fmt.Sprintf("%T", exampleObject)),
    clock:                           realClock,
  }
  return sharedIndexInformer
}

首先通过 f.defaultInformer 方法生成 informer,然后通过 f.factory.InformerFor 方法,将 informer 注册到 sharedInformerFactory

3.3、Register event handler

这个过程展示如何注册一个回调函数,以及如何触发这个回调函数

### podInformer.AddEventHandler:
func (s *sharedIndexInformer) AddEventHandler(handler ResourceEventHandler) {
  s.AddEventHandlerWithResyncPeriod(handler, s.defaultEventHandlerResyncPeriod)
}

func (s *sharedIndexInformer) AddEventHandlerWithResyncPeriod(handler ResourceEventHandler, resyncPeriod time.Duration) {

  ...
  listener := newProcessListener(handler, resyncPeriod, determineResyncPeriod(resyncPeriod, s.resyncCheckPeriod), s.clock.Now(),  initialBufferSize)
  if !s.started {
    s.processor.addListener(listener)
    return
  }
  ...
}

### s.processor.addListener(listener):
func (p *sharedProcessor) addListener(listener *processorListener) {
  p.addListenerLocked(listener)
  if p.listenersStarted {
    p.wg.Start(listener.run)
    p.wg.Start(listener.pop)
  }
}

### listener.run:
func (p *processorListener) run() {
  // this call blocks until the channel is closed.  When a panic happens during the notification
  // we will catch it, **the offending item will be skipped!**, and after a short delay (one second)
  // the next notification will be attempted.  This is usually better than the alternative of never
  // delivering again.
  stopCh := make(chan struct{})
  wait.Until(func() {
    for next := range p.nextCh {
      switch notification := next.(type) {        // 通过next结构体本身的类型来判断事件类型
      case updateNotification:
        p.handler.OnUpdate(notification.oldObj, notification.newObj)
      case addNotification:
        p.handler.OnAdd(notification.newObj)
      case deleteNotification:
        p.handler.OnDelete(notification.oldObj)
      default:
        utilruntime.HandleError(fmt.Errorf("unrecognized notification: %T", next))
      }
    }
    // the only way to get here is if the p.nextCh is empty and closed
    close(stopCh)
  }, 1*time.Second, stopCh)
}

### listener.pop:
func (p *processorListener) pop() {

  var nextCh chan<- interface{}
  var notification interface{}
  for {
    select {
    case nextCh <- notification:
      // Notification dispatched
      var ok bool
      notification, ok = p.pendingNotifications.ReadOne()
      if !ok { // Nothing to pop
        nextCh = nil // Disable this select case
      }
    case notificationToAdd, ok := <-p.addCh:
      if !ok {
        return
      }
      if notification == nil { // No notification to pop (and pendingNotifications is empty)
        // Optimize the case - skip adding to pendingNotifications
        notification = notificationToAdd
        nextCh = p.nextCh
      } else { // There is already a notification waiting to be dispatched
        p.pendingNotifications.WriteOne(notificationToAdd)
      }
    }
  }
}

这个过程总结就是:

(1)AddEventHandler 到 sharedProcessor,注册事件回调函数到 sharedProcessor

(2)listener pop 方法里会监听 p.addCh,通过 nextCh = p.nextCh 将 addCh 将事件传递给 p.nextCh

(3)listener run 方法里会监听 p.nextCh,收到信号之后,判断是属于什么类型的方法,并且执行前面注册的 Handler

所以后面需要关注当资源对象发生变更时,是如何将变更信号给 p.addCh,进一步触发回调函数的

3.4、Start all informers

通过 sharedInformers.Start(stopCh)启动所有的 informer,代码如下:

// Start initializes all requested informers.
func (f *sharedInformerFactory) Start(stopCh <-chan struct{}) {
  for informerType, informer := range f.informers {
    if !f.startedInformers[informerType] {
      go informer.Run(stopCh)
      f.startedInformers[informerType] = true
    }
  }
}

我们的例子中其实就只启动了 PodInformer,接下来看到 podInformer 的 Run 方法做了什么

### go informer.Run(stopCh):

func (s *sharedIndexInformer) Run(stopCh <-chan struct{}){
  defer utilruntime.HandleCrash()

  fifo := NewDeltaFIFOWithOptions(DeltaFIFOOptions{   // Deltafifo
    KnownObjects:          s.indexer,
    EmitDeltaTypeReplaced: true,
  })
  cfg := &Config{
    Queue:            fifo,         // Deltafifo
    ListerWatcher:    s.listerWatcher,  // listerWatcher
    ObjectType:       s.objectType,
    FullResyncPeriod: s.resyncCheckPeriod,
    RetryOnError:     false,
    ShouldResync:     s.processor.shouldResync,
    // HandleDeltas, added to process, and done in processloop
    Process:           s.HandleDeltas,
    WatchErrorHandler: s.watchErrorHandler,
  }

  func() {
    ...
    s.controller = New(cfg)
    ...
  }
  
  s.controller.Run(stopCh)
}
### s.controller.Run(stopCh)
func (c *controller) Run(stopCh <-chan struct{}) {

  r := NewReflector(
    c.config.ListerWatcher,
    c.config.ObjectType,
    c.config.Queue,
    c.config.FullResyncPeriod,
  )
  c.reflector = r

  // Run reflector
  wg.StartWithChannel(stopCh, r.Run)  

  // Run processLoop, pop from deltafifo and do ProcessFunc,
  // ProcessFunc is the s.HandleDeltas before
  wait.Until(c.processLoop, time.Second, stopCh)
}

可以看到上面的逻辑首先生成一个 DeltaFifo,然后接下来的逻辑分为两块,生产和消费:

(1)生产—r.Run:

主要的逻辑就是利用 list and watch 将资源对象包括操作类型压入队列 DeltaFifo

#### r.Run:

func (r *Reflector) Run(stopCh <-chan struct{}) {
// 执行listAndWatch
if err := r.ListAndWatch(stopCh);
}

// 执行ListAndWatch流程
func (r *Reflector)ListAndWatch(stopCh <-chan struct{}) error{
  // 1、list:
  // (1)、list pods, 实际调用的是podInformer里的ListFunc方法,
  // client.CoreV1().Pods(namespace).List(context.TODO(), options)
  
  r.listerWatcher.List(opts)
  // (2)、获取资源版本号,用于watch
  resourceVersion = listMetaInterface.GetResourceVersion()
  
  //  (3)、数据转换,转换成列表
  items, err := meta.ExtractList(list)
  
  // (4)、将资源列表中的资源对象和版本号存储到DeltaFifo中
  r.syncWith(items, resourceVersion);
  
  // 2、watch,无限循环去watch apiserver,当watch到事件的时候,执行watchHandler将event事件压入fifo
  for {
    // (1)、watch pods, 实际调用的是podInformer里的WatchFunc方法,
    // client.CoreV1().Pods(namespace).Watch(context.TODO(), options)
    w, err := r.listerWatcher.Watch(options)
    
    // (2)、watchHandler
    // watchHandler watches pod,更新DeltaFifo信息,并且更新resourceVersion
    if err := r.watchHandler(start, w, &resourceVersion, resyncerrc, stopCh);
  }
}

### r.watchHandler
// watchHandler watches w and keeps *resourceVersion up to date.
func (r *Reflector) watchHandler(start time.Time, w watch.Interface, resourceVersion *string, errc chan error, stopCh <-chan struct{}) error {
    ...
loop:
  for {
    select {
    case event, ok := <-w.ResultChan():
      newResourceVersion := meta.GetResourceVersion()
      switch event.Type {
      case watch.Added:
        err := r.store.Add(event.Object)    // Add event to srore, store的具体方法在fifo中
        if err != nil {
            utilruntime.HandleError(fmt.Errorf("%s: unable to add watch event object (%#v) to store: %v", r.name, event.Object, err))
        }
      ...
      }
      *resourceVersion = newResourceVersion
      r.setLastSyncResourceVersion(newResourceVersion)
      eventCount++
    }
  }
  ...
}

### r.store.Add:
## 即为deltaFifo的add方法:

func (f *DeltaFIFO) Add(obj interface{}) error {
  ...
  return f.queueActionLocked(Added, obj)
  ...
}

func (f *DeltaFIFO) queueActionLocked(actionType DeltaType, obj interface{}) error {
  id, err := f.KeyOf(obj)
  if err != nil {
    return KeyError{obj, err}
  }
  newDeltas := append(f.items[id], Delta{actionType, obj})
  newDeltas = dedupDeltas(newDeltas)
  if len(newDeltas) > 0 {
    if _, exists := f.items[id]; !exists {
      f.queue = append(f.queue, id)
    }

    f.items[id] = newDeltas
    f.cond.Broadcast()          // 通知所有阻塞住的消费者
  }
  ...
  return nil
}
(2)消费—c.processLoop:

消费逻辑就是从 DeltaFifo pop 出对象,然后做两件事情:(1)触发前面注册的 eventhandler (2)更新本地索引缓存 indexer,保持数据和 etcd 一致

func (c *controller) processLoop() {
  for {
    obj, err := c.config.Queue.Pop(PopProcessFunc(c.config.Process))
  }
}

### Queue.Pop:
## Queue.Pop是一个带有处理函数的pod方法,首先先看Pod逻辑,即为deltaFifo的pop方法:
func (f *DeltaFIFO) Pop(process PopProcessFunc) (interface{}, error) {
  for {                       // 无限循环
    for len(f.queue) == 0 {
      f.cond.Wait()       // 阻塞直到生产端broadcast方法通知
    }
    id := f.queue[0]
    item, ok := f.items[id]
    delete(f.items, id)
    err := process(item)        // 执行处理方法
    if e, ok := err.(ErrRequeue); ok {
      f.addIfNotPresent(id, item)     // 如果处理失败的重新加入到fifo中重新处理
      err = e.Err
    }
    return item, err
  }
}

### c.config.Process:
## c.config.Process是在初始化controller的时候赋值的,即为前面的s.HandleDeltas

### s.HandleDeltas:
func (s *sharedIndexInformer) HandleDeltas(obj interface{}) error {
  s.blockDeltas.Lock()
  defer s.blockDeltas.Unlock()
  // from oldest to newest
  for _, d := range obj.(Deltas) {
    switch d.Type {
    case Sync, Replaced, Added, Updated:
      s.cacheMutationDetector.AddObject(d.Object)
        if old, exists, err := s.indexer.Get(d.Object); err == nil && exists {
          if err := s.indexer.Update(d.Object); err != nil {
            return err
          }
          isSync := false
          switch {
          case d.Type == Sync:
            // Sync events are only propagated to listeners that requested resync
            isSync = true
          case d.Type == Replaced:
            if accessor, err := meta.Accessor(d.Object); err == nil {
                if oldAccessor, err := meta.Accessor(old); err == nil {
                  // Replaced events that didn't change resourceVersion are treated as resync events
                  // and only propagated to listeners that requested resync
                  isSync = accessor.GetResourceVersion() == oldAccessor.GetResourceVersion()
                }
            }
          }
          s.processor.distribute(updateNotification{oldObj: old, newObj: d.Object}, isSync)
        } else {
          if err := s.indexer.Add(d.Object); err != nil {
            return err
          }
          s.processor.distribute(addNotification{newObj: d.Object}, false)
        }
    case Deleted:
      if err := s.indexer.Delete(d.Object); err != nil {
        return err
      }
      s.processor.distribute(deleteNotification{oldObj: d.Object}, false)
    }
  }
  return nil
}

可以看到上面主要执行两部分逻辑:

s.processor.distribute
#### s.processor.distribute:
### 例如新增通知:s.processor.distribute(addNotification{newObj: d.Object}, false)
### 其中addNotification就是add类型的通知,后面会通过notification结构体的类型来执行不同的eventHandler

func (p *sharedProcessor) distribute(obj interface{}, sync bool) {
  p.listenersLock.RLock()
  defer p.listenersLock.RUnlock()
  
  if sync {
    for _, listener := range p.syncingListeners {
      listener.add(obj)
    }
  } else {
    for _, listener := range p.listeners {
      listener.add(obj)
    }
  }
}

func (p *processorListener) add(notification interface{}) {
  p.addCh <- notification     // 新增notification到addCh
}

这里 p.addCh 对应到前面说的关注对象 p.addCh,processorListener 收到 addCh 信号之后传递给 nextCh,然后通过 notification 结构体的类型来执行不同的 eventHandler

s.indexer 的增删改:

这个就是本地数据的缓存和索引,自定义控制逻辑里面会通过 indexer 获取操作对象的具体参数,这里就不展开细讲了。

4、总结

至此一个 informer 的 client-go 部分的流程就走完了,可以看到启动 informer 主要流程就是:

1、Reflector ListAndWatch:

(1)通过一个 reflector run 起来一个带有 list 和 watch api 的 client

(2)list 到的 pod 列表通过 DeltaFifo 存储,并更新最新的 ResourceVersion

(3)继续监听 pod,监听到的 pod 操作事件继续存储到 DeltaFifo 中

2、DeltaFifo 生产和消费:

(1)生产:list and watch 到的事件生产压入队列 DeltaFifo

(2)消费:执行注册的 eventHandler,并更新本地 indexer

所以 informer 本质其实就是一个通过 deltaFifo 建立生产消费机制,并且带有本地缓存和索引,以及可以注册回调事件的 apiServer 的客户端库。

© 版权声明
THE END