Kubernetes三种探针

k8s支持存活livenessProbe和就绪readinessProbe两种探针，两种探针都支持以下三种方式

一、exec

通过执行shell命令的方式，判断退出状态码是否是0，示例：

      exec:
        command:
        - cat
        - /tmp/healthy

二、tcp

通过TCP请求的方式，是否能建立tcp连接，示例：

      tcpSocket:
        port: 8080
      initialDelaySeconds: 15
      periodSeconds: 20

三、httpGet

通过发起http请求，判断返回结果是否符合预期，示例：

...
livenessProbe:
     httpGet:
       path: /healthz
       port: 8080
       httpHeaders:
       - name: X-Custom-Header
         value: Awesome
       initialDelaySeconds: 3
       periodSeconds: 3

• initialDelaySeconds指定了容器启动后多少秒后进行探测
• periodSeconds指定每隔多少秒进行探测

Liveness 探测

Liveness 探测让用户可以自定义判断容器是否健康的条件。如果探测失败，Kubernetes 就会重启容器。还是举例说明，创建如下 Pod：启动进程首先创建文件 /tmp/healthy，30 秒后删除，在我们的设定中，如果 /tmp/healthy 文件存在，则认为容器处于正常状态，反正则发生故障。

[root@k8s-master ~]# cat liveness.yml 
apiVersion: v1
kind: Pod
metadata:
  name: liveness-pod
  namespace: default
spec:
    restartPolicy: OnFailure
    containers:
    - name: myapp
      image: busybox
      imagePullPolicy: IfNotPresent
      command: ["/bin/sh","-c","touch /tmp/healthy;sleep 30;rm -rf /tmp/healthy;sleep 3600"]
      livenessProbe:
        exec:
          command: ["test","-e","/tmp/healthy"]
        initialDelaySeconds: 5
        periodSeconds: 5

livenessProbe 部分定义如何执行 Liveness 探测：

1. 探测的方法是：通过 test命令检查 /tmp/healthy 文件是否存在。如果命令执行成功，返回值为零，Kubernetes 则认为本次 Liveness 探测成功；如果命令返回值非零，本次 Liveness 探测失败。

2. initialDelaySeconds:5 指定容器启动 5s之后开始执行 Liveness 探测，我们一般会根据应用启动的准备时间来设置。比如某个应用正常启动要花 30 秒，那么 initialDelaySeconds 的值就应该大于 30。

3. periodSeconds: 5 指定每 5 秒执行一次 Liveness 探测。Kubernetes 如果连续执行 3 次 Liveness 探测均失败，则会杀掉并重启容器。

 下面创建 Pod liveness：

[root@k8s-master ~]# kubectl apply -f liveness.yml 
pod/liveness-pod created

 从配置文件可知，最开始的 30 秒，/tmp/healthy 存在，test 命令返回 0，Liveness 探测成功，这段时间 kubectl describe pod liveness 的 Events部分会显示正常的日志。

[root@k8s-master ~]# kubectl describe pod  liveness-pod 
Events:
  Type    Reason     Age        From                 Message
  ----    ------     ----       ----                 -------
  Normal  Scheduled  <unknown>  default-scheduler    Successfully assigned default/liveness-pod to k8s-master
  Normal  Pulled     12s        kubelet, k8s-master  Container image "busybox" already present on machine
  Normal  Created    11s        kubelet, k8s-master  Created container myapp
  Normal  Started    11s        kubelet, k8s-master  Started container myapp

35 秒之后，日志会显示 /tmp/healthy 已经不存在，Liveness 探测失败。再过几十秒，几次探测都失败后，容器会被重启。

[root@k8s-master ~]# kubectl describe pod   liveness-pod 
Command:
      /bin/sh
      -c
      touch /tmp/healthy;sleep 30;rm -rf /tmp/healthy;sleep 3600
    State:          Running
      Started:      Fri, 06 Nov 2020 15:40:59 +0800
    Last State:     Terminated
      Reason:       Error
      Exit Code:    137   #可以看到退出了，OnFailure ： 容器终止运行且退出码不为0时重启
      Started:      Fri, 06 Nov 2020 15:39:45 +0800
      Finished:     Fri, 06 Nov 2020 15:40:59 +0800
    Ready:          True
    Restart Count:  1    #容器重启了1次
    Liveness:       exec [test -e /tmp/healthy] delay=5s timeout=1s period=5s #success=1 #failure=3
    Environment:    <none>
    Mounts:
      /var/run/secrets/kubernetes.io/serviceaccount from default-token-dfmvc (ro)
Events:
  Type     Reason     Age                 From                 Message
  ----     ------     ----                ----                 -------
  Normal   Scheduled  <unknown>           default-scheduler    Successfully assigned default/liveness-pod to k8s-master
  Warning  Unhealthy  55s (x3 over 65s)   kubelet, k8s-master  Liveness probe failed:
  Normal   Killing    55s                 kubelet, k8s-master  Container myapp failed liveness probe, will be restarted  #可以看到健康检查失败
  Normal   Pulled     25s (x2 over 100s)  kubelet, k8s-master  Container image "busybox" already present on machine
  Normal   Created    25s (x2 over 99s)   kubelet, k8s-master  Created container myapp
  Normal   Started    25s (x2 over 99s)   kubelet, k8s-master  Started container myapp


[root@k8s-master ~]# kubectl get pod -o wide
NAME           READY   STATUS    RESTARTS   AGE     IP            NODE         NOMINATED NODE   READINESS GATES
liveness-pod   1/1     Running   3          4m20s   10.244.0.27   k8s-master   <none>           <none>

 除了 Liveness 探测，Kubernetes Health Check 机制还包括 Readiness 探测，下篇博客见。