Kubernetes Healthcheck

• Kubernetes Healthcheck
  • startupProbe

Kubernetes Healthcheck

对 Pod 的健康状态检查可以通过两类探针来检查：LivenessProbe 和 ReadinessProbe。

• LivenessProbe ：用于判断容器是否存活，如果 LivenessProbe 探针探测到容器不健康，则 kubelet 将杀掉容器，并根据容器的启动策略做相应的处理，如果一个容器不包括 LivenessProbe 探针，kubelet 认为该容器的 LivenessProbe 探针返回的值永远是 Success。
• ReadinessProbe：用于判断容器是否启动完成（ready 状态），可以接受请求，如果 ReadinessProbe 探针检测到失败，则 Pod 的状态则被更改。

1）ExecAction ：容器内不执行一个命令，如果该命令的返回码为 0 ，则表示容器健康,15秒探测一次，超时1秒后会重启服务。

kind: ... spec:   containers:   - name: ..     livenessProbe:      exec:      command:       - cat       - /tmp/health       initialDelaySeconds: 15       timeoutSeconds: 1

2）TCPSocketAction：通过容器的 IP 地址和端口号执行 TCP 检查，如果能够建立 TCP 连接，则表明容器健康。

... spec:   containers:   - name:   ...     livenessProbe:       tcpSocket:         port: 80       initialDelaySeconds: 30       timeoutSeconds: 1

3）HTTPGETAction：通过容器的 IP 地址，端口及路径调用 HTTP GET 方法，如果响应的状态码大于等于 200 且小于 400，则任务健康,访问的路径是 localhost:80/_status/healthz

... spec:   containers:   - ports:   ...     livenessProbe:       httpGet:         path: /_status/healthz         port: 80       initialDelaySeconds: 30       timeoutSeconds: 1

它们的含义分别如下：

• initialDelaySeconds：启动容器后进行首次健康检查的等待时间，单位为 s 。

• timeutSeconds：健康检查发送请求后等待相应的超时时间，单位为 s。当超时发生时，kubeler 会认为容器已经无法提供服务，将会重启该容器。

• initialDelaySeconds：检查开始执行的时间，以容器启动完成为起点计算

• periodSeconds：检查执行的周期，默认为10秒，最小为1秒

• timeoutSeconds：检查超时的时间，默认为1秒，最小为1秒

• successThreshold：从上次检查失败后重新认定检查成功的检查次数阈值（必须是连续成功），默认为1

• failureThreshold：从上次检查成功后认定检查失败的检查次数阈值（必须是连续失败），默认为1

• httpGet的属性

  • host：主机名或IP

  • scheme：链接类型，HTTP或HTTPS，默认为HTTP

  • path：请求路径

  • httpHeaders：自定义请求头

  • port：请求端口

定义说明如下：

$ kubectl explain pod.spec.containers.livenessProbe KIND:     Pod VERSION:  v1 RESOURCE: livenessProbe <Object> DESCRIPTION:      Periodic probe of container liveness. Container will be restarted if the      probe fails. Cannot be updated. More info:      https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle#container-probes      Probe describes a health check to be performed against a container to      determine whether it is alive or ready to receive traffic. FIELDS:    exec <Object>      One and only one of the following should be specified. Exec specifies the      action to take.    failureThreshold <integer>      Minimum consecutive failures for the probe to be considered failed after      having succeeded. Defaults to 3. Minimum value is 1.    httpGet  <Object>      HTTPGet specifies the http request to perform.    initialDelaySeconds  <integer>      Number of seconds after the container has started before liveness probes      are initiated. More info:      https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle#container-probes    periodSeconds    <integer>      How often (in seconds) to perform the probe. Default to 10 seconds. Minimum      value is 1.    successThreshold <integer>      Minimum consecutive successes for the probe to be considered successful      after having failed. Defaults to 1. Must be 1 for liveness. Minimum value      is 1.    tcpSocket    <Object>      TCPSocket specifies an action involving a TCP port. TCP hooks not yet      supported    timeoutSeconds   <integer>      Number of seconds after which the probe times out. Defaults to 1 second.      Minimum value is 1. More info:      https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle#container-probes

startupProbe

新增 startupProbe, 主要解决 livenessProbe 启动时，如果无法正常启动或服务启动时间较长引起的容器重新启动的问题。

apiVersion: apps/v1 kind: Deployment metadata:   name: busybox-lifecycles-nginx-sleep spec:   replicas: 2   selector:     matchLabels:       app: busybox-lifecycles-nginx-sleep       env-o: nginx-sleep   template:     metadata:       labels:         app: busybox-lifecycles-nginx-sleep         env-o: nginx-sleep       annotations:         consul.hashicorp.com/connect-inject: "true"     spec:       terminationGracePeriodSeconds: 120       containers:       - image: slzcc/terminal-ctl:ubuntu-20.04         imagePullPolicy: Always         command:           - nginx           - -g           - daemon off;         name: busybox         livenessProbe:           exec:             command:             - /bin/bash             - -c             - nc -z -v -n 127.0.0.1 80           failureThreshold: 10           initialDelaySeconds: 5           periodSeconds: 10           successThreshold: 1           timeoutSeconds: 5         readinessProbe:           exec:             command:             - /bin/bash             - -c             - nc -z -v -n 127.0.0.1 80           initialDelaySeconds: 5           periodSeconds: 10         startupProbe:           httpGet:             path: /             port: 801           failureThreshold: 3           periodSeconds: 10       restartPolicy: Always --- apiVersion: v1 kind: Service metadata:   name: busybox-lifecycles-nginx-sleep   labels:     app: busybox-lifecycles-nginx-sleep spec:   ports:    - name: http      port: 80      targetPort: 80      protocol: TCP   selector:     app: busybox-lifecycles-nginx-sleep