Pod Lifecycle

Linux补充知识

SIGTERM vs SIGKILL: What’s the Difference?

https://linuxhandbook.com/sigterm-vs-sigkill/

Pod Lifecycle

参考 https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle/

pod phase

init-container

Container state

  • waiting

  • Running

  • Termination

    • Process is terminated/crashed

    • Pod is deleted

    • Node failure or maintenance

    • Evicteed due to lack of resources

Stopping/Terminating Pods

kubectl delete pod <name>

当执行删除的时候，

  • API server会设置一个timer（grace period Timer），默认是30秒

  • 同时pod状态改成Terminating

  • Pod所在node上的kubelet收到命令，会给Pod里的container发送SIGTERM信号，然后等等container退出

  • 如果container在timer到时之前退出了，那么pod信息同时会被API server从存储中删除

  • 如果container没有在timer到时之前退出，则kubelet会发送SIGKILL信息到pod里的容器，强制杀死容器， 最后API server更新存储etcd

grace period timer是可以修改的

$ kubectl delete pod <name> --grace-period=<seconds>

apiVersion: v1
kind: Pod
metadata:
  name: web
spec:
  terminationGracePeriodSeconds: 10
  containers:
  - image: nginx
    name: web

或者也可以直接就强制删除pod （SIGKILL）

$ kubectl delete pod <name> --grace-period=0 --force

Persistency of Pod

Pod本身不会重新自动部署

如果一个pod停止了，它不会重启动，只有可能创建一个新的

如果有配置需要持久化怎么办？

  • Pod Manifests, secrets and ConfigMaps

  • environment variables

数据持久？

  • PersistentVolume

  • PersistentVolumeClaim

Container Restart Policy

The spec of a Pod has a restartPolicy field with possible values Always, OnFailure, and Never. The default value is Always.

vagrant@k8s-master:~$ kubectl run web --image nginx --dry-run=client -o yaml
apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: null
  labels:
    run: web
  name: web
spec:
  containers:
  - image: nginx
    name: web
    resources: {}
  dnsPolicy: ClusterFirst
  restartPolicy: Always
status: {}