欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

kubernetes Event 源码解析

程序员文章站 2022-05-22 08:02:41
...

本文来自于【阿里云官方镜像站:https://developer.aliyun.com/mirror/?utm_content=g_1000307095 】

原文链接:https://developer.aliyun.com/article/748756?spm=a2c6h.12873581.0.0.54c47e46mLfYep

镜像下载、域名解析、时间同步请点击 阿里巴巴开源镜像站

一、Event定义

其实event也是一个资源对象,并且通过apiserver将event存储在etcd中,所以我们也可以通过 kubectl get event 命令查看对应的event对象。
以下是一个event的yaml文件:

apiVersion: v1count: 1eventTime: nullfirstTimestamp: "2020-03-02T13:08:22Z"involvedObject:  apiVersion: v1  kind: Pod  name: example-foo-d75d8587c-xsf64  namespace: default  resourceVersion: "429837"  uid: ce611c62-6c1a-4bd8-9029-136a1adf7de4kind: EventlastTimestamp: "2020-03-02T13:08:22Z"message: Pod sandbox changed, it will be killed and re-created.metadata:  creationTimestamp: "2020-03-02T13:08:30Z"  name: example-foo-d75d8587c-xsf64.15f87ea1df862b64  namespace: default  resourceVersion: "479466"  selfLink: /api/v1/namespaces/default/events/example-foo-d75d8587c-xsf64.15f87ea1df862b64  uid: 9fe6f72a-341d-4c49-960b-e185982d331areason: SandboxChangedreportingComponent: ""reportingInstance: ""source:  component: kubelet  host: minikubetype: Normal

主要字段说明:

involvedObject: 触发event的资源类型lastTimestamp:最后一次触发的时间message:事件说明metadata :event的元信息,name,namespace等reason:event的原因source:上报事件的来源,比如kubelet中的某个节点type:事件类型,Normal或Warning

event字段定义可以看这里:types.go#L5078
接下来我们来看看,整个event是如何下入的。

二、写入事件

1、这里以kubelet为例,看看是如何进行事件写入的
2、文中代码以Kubernetes 1.17.3为例进行分析

先以一幅图来看下整个的处理流程

创建操作事件的客户端:
kubelet/app/server.go#L461

// makeEventRecorder sets up kubeDeps.Recorder if it's nil. It's a no-op otherwise.func makeEventRecorder(kubeDeps *kubelet.Dependencies, nodeName types.NodeName) {    if kubeDeps.Recorder != nil {        return
   }    //事件广播
   eventBroadcaster := record.NewBroadcaster()    //创建EventRecorder
   kubeDeps.Recorder = eventBroadcaster.NewRecorder(legacyscheme.Scheme, v1.EventSource{Component: componentKubelet, Host: string(nodeName)})    //发送event至log输出
   eventBroadcaster.StartLogging(klog.V(3).Infof)    if kubeDeps.EventClient != nil {
       klog.V(4).Infof("Sending events to api server.")        //发送event至apiserver
       eventBroadcaster.StartRecordingToSink(&v1core.EventSinkImpl{Interface: kubeDeps.EventClient.Events("")})
   } else {
       klog.Warning("No api server defined - no events will be sent to API server.")
   }
}

通过 makeEventRecorder 创建了 EventRecorder 实例,这是一个事件广播器,通过它提供了StartLogging和StartRecordingToSink两个事件处理函数,分别将event发送给log和apiserver。
NewRecorder创建了 EventRecorder 的实例,它提供了 Event ,Eventf 等方法供事件记录。

EventBroadcaster

我们来看下EventBroadcaster接口定义:event.go#L113

// EventBroadcaster knows how to receive events and send them to any EventSink, watcher, or log.type EventBroadcaster interface {    //
   StartEventWatcher(eventHandler func(*v1.Event)) watch.Interface    StartRecordingToSink(sink EventSink) watch.Interface    StartLogging(logf func(format string, args ...interface{})) watch.Interface    NewRecorder(scheme *runtime.Scheme, source v1.EventSource) EventRecorder    Shutdown()}

具体实现是通过 eventBroadcasterImpl struct来实现了各个方法。
其中StartLogging 和 StartRecordingToSink 其实就是完成了对事件的消费,EventRecorder实现对事件的写入,中间通过channel实现了生产者消费者模型。

EventRecorder

我们先来看下EventRecorder 接口定义:event.go#L88,提供了一下4个方法

// EventRecorder knows how to record events on behalf of an EventSource.type EventRecorder interface {    // Event constructs an event from the given information and puts it in the queue for sending.
   // 'object' is the object this event is about. Event will make a reference-- or you may also
   // pass a reference to the object directly.
   // 'type' of this event, and can be one of Normal, Warning. New types could be added in future
   // 'reason' is the reason this event is generated. 'reason' should be short and unique; it
   // should be in UpperCamelCase format (starting with a capital letter). "reason" will be used
   // to automate handling of events, so imagine people writing switch statements to handle them.
   // You want to make that easy.
   // 'message' is intended to be human readable.
   //
   // The resulting event will be created in the same namespace as the reference object.
   Event(object runtime.Object, eventtype, reason, message string)
   // Eventf is just like Event, but with Sprintf for the message field.
   Eventf(object runtime.Object, eventtype, reason, messageFmt string, args ...interface{})
   // PastEventf is just like Eventf, but with an option to specify the event's 'timestamp' field.
   PastEventf(object runtime.Object, timestamp metav1.Time, eventtype, reason, messageFmt string, args ...interface{})
   // AnnotatedEventf is just like eventf, but with annotations attached
   AnnotatedEventf(object runtime.Object, annotations map[string]string, eventtype, reason, messageFmt string, args ...interface{})}

主要参数说明:

object 对应event资源定义中的 involvedObjecteventtype 对应event资源定义中的type,可选Normal,Warning.reason :事件原因message :事件消息

我们来看下当我们调用 Event(object runtime.Object, eventtype, reason, message string) 的整个过程。
发现最终都调用到了 generateEvent 方法:event.go#L316

func (recorder *recorderImpl) generateEvent(object runtime.Object, annotations map[string]string, timestamp metav1.Time, eventtype, reason, message string) {
   .....
   event := recorder.makeEvent(ref, annotations, eventtype, reason, message)
   event.Source = recorder.source    go func() {        // NOTE: events should be a non-blocking operation
       defer utilruntime.HandleCrash()
       recorder.Action(watch.Added, event)
   }()
}

最终事件在一个 goroutine 中通过调用 recorder.Action 进入处理,这里保证了每次调用event方法都是非阻塞的。
其中 makeEvent 的作用主要是构造了一个event对象,事件name根据InvolvedObject中的name加上时间戳生成:

注意看:对于一些非namespace资源产生的event,event的namespace是default

func (recorder *recorderImpl) makeEvent(ref *v1.ObjectReference, annotations map[string]string, eventtype, reason, message string) *v1.Event {
   t := metav1.Time{Time: recorder.clock.Now()}
   namespace := ref.Namespace
   if namespace == "" {
       namespace = metav1.NamespaceDefault
   }
   return &v1.Event{
       ObjectMeta: metav1.ObjectMeta{
           Name:        fmt.Sprintf("%v.%x", ref.Name, t.UnixNano()),
           Namespace:   namespace,
           Annotations: annotations,
       },
       InvolvedObject: *ref,
       Reason:         reason,
       Message:        message,
       FirstTimestamp: t,
       LastTimestamp:  t,
       Count:          1,
       Type:           eventtype,
   }}

进一步跟踪Action方法,apimachinery/blob/master/pkg/watch/mux.go#L188:23

// Action distributes the given event among all watchers.func (m *Broadcaster) Action(action EventType, obj runtime.Object) {
   m.incoming <- Event{action, obj}
}

将event写入到了一个channel里面。
注意:
这个Action方式是apimachinery包中的方法,因为实现的sturt recorderImpl
将 *watch.Broadcaster 作为一个匿名struct,并且在 NewRecorder 进行 Broadcaster 赋值,这个Broadcaster其实就是 eventBroadcasterImpl 中的Broadcaster。
到此,基本清楚了event最终被写入到了 Broadcaster 中的 incoming channel中,下面看下是怎么进行消费的。

三、消费事件

在 makeEventRecorder 调用的 StartLogging 和 StartRecordingToSink 其实就是完成了对事件的消费。

StartLogging直接将event输出到日志StartRecordingToSink将事件写入到apiserver

两个方法内部都调用了 StartEventWatcher 方法,并且传入一个 eventHandler 方法对event进行处理

func (e *eventBroadcasterImpl) StartEventWatcher(eventHandler func(*v1.Event)) watch.Interface {
   watcher := e.Watch()    go func() {        defer utilruntime.HandleCrash()        for watchEvent := range watcher.ResultChan() {
           event, ok := watchEvent.Object.(*v1.Event)            if !ok {                // This is all local, so there's no reason this should
               // ever happen.
               continue
           }
           eventHandler(event)
       }
   }()    return watcher
}

其中 watcher.ResultChan 方法就拿到了事件,这里是在一个goroutine中通过func (m *Broadcaster) loop() ==>func (m *Broadcaster) distribute(event Event) 方法调用将event又写入了broadcasterWatcher.result
主要看下 StartRecordingToSink 提供的的eventHandler, recordToSink 方法:

func recordToSink(sink EventSink, event *v1.Event, eventCorrelator *EventCorrelator, sleepDuration time.Duration) {    // Make a copy before modification, because there could be multiple listeners.
   // Events are safe to copy like this.
   eventCopy := *event
   event = &eventCopy
   result, err := eventCorrelator.EventCorrelate(event)    if err != nil {
       utilruntime.HandleError(err)
   }    if result.Skip {        return
   }
   tries := 0
   for {        if recordEvent(sink, result.Event, result.Patch, result.Event.Count > 1, eventCorrelator) {            break
       }
       tries++        if tries >= maxTriesPerEvent {
           klog.Errorf("Unable to write event '%#v' (retry limit exceeded!)", event)            break
       }        // Randomize the first sleep so that various clients won't all be
       // synced up if the master goes down.
       // 第一次重试增加随机性,防止 apiserver 重启的时候所有的事件都在同一时间发送事件
       if tries == 1 {            time.Sleep(time.Duration(float64(sleepDuration) * rand.Float64()))
       } else {            time.Sleep(sleepDuration)
       }
   }
}

其中event被经过了一个 eventCorrelator.EventCorrelate(event) 方法做预处理,主要是聚合相同的事件(避免产生的事件过多,增加 etcd 和 apiserver 的压力,也会导致查看 pod 事件很不清晰)
下面一个for循环就是在进行重试,最大重试次数是12次,调用 recordEvent 方法才真正将event写入到了apiserver。

事件处理

我们来看下EventCorrelate方法:

// EventCorrelate filters, aggregates, counts, and de-duplicates all incoming eventsfunc (c *EventCorrelator) EventCorrelate(newEvent *v1.Event) (*EventCorrelateResult, error) {    if newEvent == nil {        return nil, fmt.Errorf("event is nil")
   }
   aggregateEvent, ckey := c.aggregator.EventAggregate(newEvent)
   observedEvent, patch, err := c.logger.eventObserve(aggregateEvent, ckey)    if c.filterFunc(observedEvent) {        return &EventCorrelateResult{Skip: true}, nil
   }    return &EventCorrelateResult{Event: observedEvent, Patch: patch}, err
}

分别调用了 aggregator.EventAggregate , logger.eventObserve , filterFunc 三个方法,分别作用是:

1、aggregator.EventAggregate:聚合event,如果在最近 10 分钟出现过 10 个相似的事件(除了 message 和时间戳之外其他关键字段都相同的事件),aggregator 会把它们的 message 设置为 (combined from similar events)+event.Message
2、logger.eventObserve:它会把相同的事件以及包含 aggregator 被聚合了的相似的事件,通过增加 Count 字段来记录事件发生了多少次。
3、filterFunc: 这里实现了一个基于令牌桶的限流算法,如果超过设定的速率则丢弃,保证了apiserver的安全。

我们主要来看下aggregator.EventAggregate方法:

func (e *EventAggregator) EventAggregate(newEvent *v1.Event) (*v1.Event, string) {
   now := metav1.NewTime(e.clock.Now())    var record aggregateRecord    // eventKey is the full cache key for this event
   //eventKey 是将除了时间戳外所有字段结合在一起
   eventKey := getEventKey(newEvent)    // aggregateKey is for the aggregate event, if one is needed.
   //aggregateKey 是除了message和时间戳外的字段结合在一起,localKey 是message
   aggregateKey, localKey := e.keyFunc(newEvent)    // Do we have a record of similar events in our cache?
   e.Lock()    defer e.Unlock()    //从cache中根据aggregateKey查询是否存在,如果是相同或者相类似的事件会被放入cache中
   value, found := e.cache.Get(aggregateKey)    if found {
       record = value.(aggregateRecord)
   }    //判断上次事件产生的时间是否超过10分钟,如何操作则重新生成一个localKeys集合(集合中存放message)
   maxInterval := time.Duration(e.maxIntervalInSeconds) * time.Second
   interval := now.Time.Sub(record.lastTimestamp.Time)    if interval > maxInterval {
       record = aggregateRecord{localKeys: sets.NewString()}
   }    // Write the new event into the aggregation record and put it on the cache
   //将locakKey也就是message放入集合中,如果message相同就是覆盖了
   record.localKeys.Insert(localKey)
   record.lastTimestamp = now
   e.cache.Add(aggregateKey, record)    // If we are not yet over the threshold for unique events, don't correlate them
   //判断localKeys集合中存放的类似事件是否超过10个,
   if uint(record.localKeys.Len()) < e.maxEvents {        return newEvent, eventKey
   }    // do not grow our local key set any larger than max
   record.localKeys.PopAny()    // create a new aggregate event, and return the aggregateKey as the cache key
   // (so that it can be overwritten.)
   eventCopy := &v1.Event{
       ObjectMeta: metav1.ObjectMeta{
           Name:      fmt.Sprintf("%v.%x", newEvent.InvolvedObject.Name, now.UnixNano()),
           Namespace: newEvent.Namespace,
       },
       Count:          1,
       FirstTimestamp: now,
       InvolvedObject: newEvent.InvolvedObject,
       LastTimestamp:  now,        //这里会对message加个前缀:(combined from similar events):
       Message:        e.messageFunc(newEvent),
       Type:           newEvent.Type,
       Reason:         newEvent.Reason,
       Source:         newEvent.Source,
   }    return eventCopy, aggregateKey
}

aggregator.EventAggregate方法中其实就是判断了通过cache和localKeys判断事件是否相似,如果最近 10 分钟出现过 10 个相似的事件就合并并加上前缀,后续通过logger.eventObserve方法进行count累加,如果message也相同,肯定就是直接count++。

四、总结

event处理的整个流程基本就是这样,我们可以概括为以下几点,也可以结合文中的图对比一起来看:

1、创建 EventRecorder 对象,通过其提供的 Event 等方法,创建好event对象
2、将创建出来的对象发送给 EventBroadcaster 中的channel中
3、EventBroadcaster 通过后台运行的goroutine,从管道中取出事件,并广播给提前注册好的handler处理
4、当输出log的handler收到事件就直接打印事件
5、当 EventSink handler收到处理事件就通过预处理之后将事件发送给apiserver
6、其中预处理包含三个动作,1、限流 2、聚合 3、计数
7、apiserver收到事件处理之后就存储在etcd中

回顾event的整个流程,可以看到event并不是保证100%事件写入(从预处理的过程来看),这样做是为了后端服务etcd的可用性,因为event事件在整个集群中产生是非常频繁的,尤其在服务不稳定的时候,而相比Deployment,Pod等其他资源,又没那么的重要。所以这里做了个取舍。