Kubernetes 源码解析 - Informer

Kubernetes 源码解析 - Informer

上篇扒了 HPA 的源码,但是没深入细节,今天往细节深入。

开局先祭出一张图:

为什么要有 Informer?

Kubernetes 中的持久化数据保存在 etcd中,各个组件并不会直接访问 etcd,而是通过 api-server暴露的 RESTful 接口对集群进行访问和控制。

资源的控制器(图中右侧灰色的部分)读取数据也并不会直接从 api-server 中获取资源信息(这样会增加 api-server 的压力),而是从其“本地缓存”中读取。这个“本地缓存”只是表象的存在,加上缓存的同步逻辑就是今天要是说的Informer(灰色区域中的第一个蓝色块)所提供的功能。

从图中可以看到 Informer 的几个组件:

  • Reflector:与 api-server交互,监听资源的变更。
  • Delta FIFO Queue:增量的 FIFO 队列,保存 Reflector 监听到的资源变更(简单的封装)。
  • Indexer:Informer 的本地缓存,FIFO 队列中的数据根据不同的变更类型,在该缓存中进行操作。
    • Local Store:

上篇 提到了水平自动伸缩的控制器HorizontalController,其构造方法就需要提供 Informer

//pkg/controller/podautoscaler/horizontal.go

type HorizontalController struct {
	scaleNamespacer scaleclient.ScalesGetter
	hpaNamespacer   autoscalingclient.HorizontalPodAutoscalersGetter
	mapper          apimeta.RESTMapper
	replicaCalc   *ReplicaCalculator
	eventRecorder record.EventRecorder
	downscaleStabilisationWindow time.Duration
	hpaLister       autoscalinglisters.HorizontalPodAutoscalerLister
	hpaListerSynced cache.InformerSynced
	podLister       corelisters.PodLister
	podListerSynced cache.InformerSynced
	queue workqueue.RateLimitingInterface
	recommendations map[string][]timestampedRecommendation
}

func NewHorizontalController(
	evtNamespacer v1core.EventsGetter,
	scaleNamespacer scaleclient.ScalesGetter,
	hpaNamespacer autoscalingclient.HorizontalPodAutoscalersGetter,
	mapper apimeta.RESTMapper,
	metricsClient metricsclient.MetricsClient,
	//从HorizontalPodAutoscalerInformer 获取hpa 实例信息
	hpaInformer autoscalinginformers.HorizontalPodAutoscalerInformer,
	//从PodInformer 中获取 pod 信息
	podInformer coreinformers.PodInformer,
	resyncPeriod time.Duration,
	downscaleStabilisationWindow time.Duration,
	tolerance float64,
	cpuInitializationPeriod,
	delayOfInitialReadinessStatus time.Duration,

) *HorizontalController {
    ......
    	hpaInformer.Informer().AddEventHandlerWithResyncPeriod( //添加事件处理器
		cache.ResourceEventHandlerFuncs{
			AddFunc:    hpaController.enqueueHPA,
			UpdateFunc: hpaController.updateHPA,
			DeleteFunc: hpaController.deleteHPA,
		},
		resyncPeriod,
	)
	......
}

type HorizontalPodAutoscalerInformer interface {
	Informer() cache.SharedIndexInformer
	Lister() v1.HorizontalPodAutoscalerLister
}

HorizontalPodAutoscalerInformer的实例化方法中就出现了今天的正主cache.NewSharedIndexInformer()

//staging/src/k8s.io/client-go/informers/autoscaling/v1/horizontalpodautoscaler.go
func NewFilteredHorizontalPodAutoscalerInformer(client kubernetes.Interface, namespace string, resyncPeriod time.Duration, indexers cache.Indexers, tweakListOptions internalinterfaces.TweakListOptionsFunc) cache.SharedIndexInformer {
	return cache.NewSharedIndexInformer(
	       //用于 list 和 watch api-server 中的资源。比如用来创建 Reflector
		&cache.ListWatch{
			ListFunc: func(options metav1.ListOptions) (runtime.Object, error) {
				if tweakListOptions != nil {
					tweakListOptions(&options)
				}
				//使用 HPA API 获取 HPA资源
				return client.AutoscalingV1().HorizontalPodAutoscalers(namespace).List(options)
			},
			WatchFunc: func(options metav1.ListOptions) (watch.Interface, error) {
				if tweakListOptions != nil {
					tweakListOptions(&options)
				}
				//使用 HPA API 监控 HPA资源
				return client.AutoscalingV1().HorizontalPodAutoscalers(namespace).Watch(options)
			},
		},
		&autoscalingv1.HorizontalPodAutoscaler{},
		resyncPeriod,
		indexers,
	)
}

初始化

Informer

//staging/src/k8s.io/client-go/tools/cache/index.go
type Indexers map[string]IndexFunc
type IndexFunc func(obj interface{}) ([]string, error)

实例化 Indexers cache.Indexers{cache.NamespaceIndex: cache.MetaNamespaceIndexFunc}

//staging/src/k8s.io/client-go/tools/cache/shared_informer.go
// ListerWatcher 用于 list 和watch api-server 上的资源
//runtime.Object要监控的资源的运行时对象
//time.Duration同步的间隔时间
//Indexers 提供不同资源的索引数据的信息查询方法,如 namespace => MetaNamespaceIndexFunc
func NewSharedIndexInformer(lw ListerWatcher, objType runtime.Object, defaultEventHandlerResyncPeriod time.Duration, indexers Indexers) SharedIndexInformer {
	realClock := &clock.RealClock{}
	sharedIndexInformer := &sharedIndexInformer{
		processor:                       &sharedProcessor{clock: realClock},
		indexer:                         NewIndexer(DeletionHandlingMetaNamespaceKeyFunc, indexers), //初始化 Indexer
		listerWatcher:                   lw,
		objectType:                      objType,
		resyncCheckPeriod:               defaultEventHandlerResyncPeriod,
		defaultEventHandlerResyncPeriod: defaultEventHandlerResyncPeriod,
		cacheMutationDetector:           NewCacheMutationDetector(fmt.Sprintf("%T", objType)),
		clock:                           realClock,
	}
	return sharedIndexInformer
}

Indexer

Indexer提供了本地缓存的实现:计算 key 和对数据进行控制(通过调用ThreadSafeStore的接口)

type Indexer interface {
	Store
	Index(indexName string, obj interface{}) ([]interface{}, error)
	IndexKeys(indexName, indexedValue string) ([]string, error)
	ListIndexFuncValues(indexName string) []string
	ByIndex(indexName, indexedValue string) ([]interface{}, error)
	GetIndexers() Indexers
	AddIndexers(newIndexers Indexers) error
}

Indexer 的创建

//staging/src/k8s.io/client-go/tools/cache/store.go
//keyFunc:key 的生成规则
//indexers:提供了索引资源的不同信息的访问方法,如用于查询命名空间的 namespace => MetaNamespaceIndexFunc
func NewIndexer(keyFunc KeyFunc, indexers Indexers) Indexer {
	return &cache{
		cacheStorage: NewThreadSafeStore(indexers, Indices{}),
		keyFunc:      keyFunc,
	}
}

ThreadSafeStore

ThreadSafeStore提供了对存储的并发访问接口

注意事项:不能修改Get或List返回的任何内容,因为它不仅会破坏线程安全,还会破坏索引功能。

//staging/src/k8s.io/client-go/tools/cache/thread_safe_store.go
func NewThreadSafeStore(indexers Indexers, indices Indices) ThreadSafeStore {
	return &threadSafeMap{
		items:    map[string]interface{}{},
		indexers: indexers,
		indices:  indices,
	}
}
type threadSafeMap struct {
	lock  sync.RWMutex
	items map[string]interface{} //key => value
	indexers Indexers //value 的信息的访问方法
	indices Indices //索引
}

Reflector

Reflector通过 ListerWatcher(API)与api-server交互,对资源进行监控。将资源实例的创建、更新、删除等时间封装后保存在Informer的FIFO 队列中。

//staging/src/k8s.io/client-go/tools/cache/reflector.go
func NewReflector(lw ListerWatcher, expectedType interface{}, store Store, resyncPeriod time.Duration) *Reflector {
	return NewNamedReflector(naming.GetNameFromCallsite(internalPackages...), lw, expectedType, store, resyncPeriod)
}

// NewNamedReflector same as NewReflector, but with a specified name for logging
func NewNamedReflector(name string, lw ListerWatcher, expectedType interface{}, store Store, resyncPeriod time.Duration) *Reflector {
	r := &Reflector{
		name:          name,
		listerWatcher: lw,
		store:         store, //FIFO队列
		period:        time.Second,
		resyncPeriod:  resyncPeriod,
		clock:         &clock.RealClock{},
	}
	r.setExpectedType(expectedType)
	return r
}

添加同步事件监听器

通过sharedIndexInformer#AddEventHandlerWithResyncPeriod()注册事件监听器。

以前面的 HorizontalController为例,创建 informer 的时候添加了三个处理方法:AddFuncUpdateFuncDeleteFunc。这三个方法的实现是将对应的元素的 key(固定格式 namespace/name)从 workequeue中进行入队、出队的操作。(资源控制器监听了该 workqueue

运行

controller-manager

在通过InformerFactory创建Informer完成后,都会将新建的 Informer加入到InformerFactory的一个map中。

controller-manager在完成所有的控制器(各种Controller,包括 CRD)后,会调用InformerFactory#Start()来启动InformerFactorymap中的所有 Informer(调用Informer#Run()方法)

sharedIndexInformer#Run()

//staging/src/k8s.io/client-go/tools/cache/shared_informer.go
func (s *sharedIndexInformer) Run(stopCh <-chan struct{}) {
	defer utilruntime.HandleCrash()
      //创建一个增量的 FIFO队列:DeltaFIFO
	fifo := NewDeltaFIFO(MetaNamespaceKeyFunc, s.indexer)
	cfg := &Config{
		Queue:            fifo,
		ListerWatcher:    s.listerWatcher,
		ObjectType:       s.objectType,
		FullResyncPeriod: s.resyncCheckPeriod,
		RetryOnError:     false,
		ShouldResync:     s.processor.shouldResync,

		Process: s.HandleDeltas,
	}
     //启动前的初始化,创建 Controller
	func() {
		s.startedLock.Lock()
		defer s.startedLock.Unlock()

		s.controller = New(cfg)
		s.controller.(*controller).clock = s.clock
		s.started = true
	}()
	processorStopCh := make(chan struct{})
	var wg wait.Group
	defer wg.Wait()              // Wait for Processor to stop
	defer close(processorStopCh) // Tell Processor to stop
	wg.StartWithChannel(processorStopCh, s.cacheMutationDetector.Run)
	wg.StartWithChannel(processorStopCh, s.processor.run)
     //退出时的状态清理
	defer func() {
		s.startedLock.Lock()
		defer s.startedLock.Unlock()
		s.stopped = true // Don't want any new listeners
	}()
	//实行控制逻辑
	s.controller.Run(stopCh)
}

controller#Run()

//staging/src/k8s.io/client-go/tools/cache/controller.go
func (c *controller) Run(stopCh <-chan struct{}) {
	defer utilruntime.HandleCrash()
	go func() {
		<-stopCh
		c.config.Queue.Close()
	}()
	//创建一个 Reflector,用于从 api-server list 和 watch 资源
	r := NewReflector(
		c.config.ListerWatcher,
		c.config.ObjectType,
		c.config.Queue,
		c.config.FullResyncPeriod,
	)
	r.ShouldResync = c.config.ShouldResync
	r.clock = c.clock
      //为 controller 指定 Reflector
	c.reflectorMutex.Lock()
	c.reflector = r
	c.reflectorMutex.Unlock()
	var wg wait.Group
	defer wg.Wait()
     //执行Reflector#Run():会启动一个goroutine开始监控资源,将 watch 到的数据写入到queue(FIFO 队列)中
	wg.StartWithChannel(stopCh, r.Run)
      //持续从 queue(FIFO 队列) 获取数据并进行处理,处理的逻辑在sharedIndexInformer#HandleDeltas()
	wait.Until(c.processLoop, time.Second, stopCh)
}

sharedIndexInformer#HandleDeltas()

//staging/src/k8s.io/client-go/tools/cache/shared_informer.go
func (s *sharedIndexInformer) HandleDeltas(obj interface{}) error {
	s.blockDeltas.Lock()
	defer s.blockDeltas.Unlock()

	// from oldest to newest
	for _, d := range obj.(Deltas) { //循环处理 FIFO 队列中取出的资源实例
		switch d.Type {
		case Sync, Added, Updated: //同步(后面详细解读)、新增、更新事件
			isSync := d.Type == Sync
			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 { //如果 indexer 中已经存在,更掉用 update 方法进行更新
					return err
				}
				//更新成功后发送“更新”通知:包含了新、旧资源实例
				s.processor.distribute(updateNotification{oldObj: old, newObj: d.Object}, isSync)
			} else {
			      //如果 indexer 中没有该资源实例,则放入 indexer 中
				if err := s.indexer.Add(d.Object); err != nil {
					return err
				}
				//添加成功后,发送“新增”通知:包含了新加的资源实例
				s.processor.distribute(addNotification{newObj: d.Object}, isSync)
			}
		case Deleted: //删除事件
			if err := s.indexer.Delete(d.Object); err != nil {//从 indexer 中删除
				return err
			}
			//删除成功后,发送“删除通知”:包含了删除的资源实例
			s.processor.distribute(deleteNotification{oldObj: d.Object}, false)
		}
	}
	return nil
}

总结

Informer 的实现不算复杂,却在 Kubernetes 中很常见,每种资源的控制也都通过 Informer 来获取api-server的资源实例的变更。

(转载本站文章请注明作者和出处乱世浮生,请勿用于任何商业用途)

comments powered by Disqus