1、问题

pod启动或者通过ReplicaSet等控制器启动pod后，pod的状态一瞬间呈现Completed状态，随后一直显示CrashLoopBackOff状态，导致pod一直重启失败，如下所示

[root@k8s-master01 sc_work]# kubectl create -f replica_set_example.yaml 
replicaset.extensions/nginx-relica created
[root@k8s-master01 sc_work]# kubectl get pod
NAME                 READY   STATUS      RESTARTS   AGE
nginx-relica-d6fhn   0/1     Completed   0          3s
nginx-relica-gqsbn   0/1     Completed   0          3s
nginx-relica-nvjk5   0/1     Completed   1          3s
[root@k8s-master01 sc_work]# kubectl get pod
NAME                 READY   STATUS             RESTARTS   AGE
nginx-relica-d6fhn   0/1     CrashLoopBackOff   1          6s
nginx-relica-gqsbn   0/1     CrashLoopBackOff   1          6s
nginx-relica-nvjk5   0/1     CrashLoopBackOff   1          6s

所用的yaml如下所示：

apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
  name: nginx-relica
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx-label
  template:
    metadata:
      labels:
        app: nginx-label
    spec:
      containers:
      - name: nginx-pod
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80
        command: ["/bin/sh", "-c", "echo hello world > /usr/share/nginx/html/hello.html"]

2、原因

当ReplicaSet控制创建3个pod后，3个pod中分别启动了容器nginx-pod，启动容器后，以非常短的时间执行 ["/bin/sh", “-c”, “echo hello world > /usr/share/nginx/html/hello.html”]后，容器就要结束了，退出码为0，pod状态为completed状态。但ReplicaSet设置的副本数为3，导致3个pod要重启，但pod中容器只执行一个command命令就结束，导致pod的生命周期非常短，k8s都来不及启动就消亡了，导致pod出现CrashLoopBackOff状态。

3、解决办法

延长pod的声明周期，比如yaml改成如下形式，在command中增加sleep延长pod的运行时间。

apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
  name: nginx-relica
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx-label
  template:
    metadata:
      labels:
        app: nginx-label
    spec:
      containers:
      - name: nginx-pod
        image: nginx
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 80
        #command中增加sleep，延长pod运行时间
        command: ["/bin/sh", "-c", "echo hello world > /usr/share/nginx/html/hello.html; sleep 60"]

删除原来的ReplicaSet后，再重新执行ReplicaSet，pod正常运行

[root@k8s-master01 sc_work]# kubectl create -f replica_set_example.yaml 
replicaset.extensions/nginx-relica created
[root@k8s-master01 sc_work]# kubectl get pod
NAME                 READY   STATUS    RESTARTS   AGE
nginx-relica-4vjl9   1/1     Running   0          4s
nginx-relica-5l7w2   1/1     Running   0          4s
nginx-relica-t6qdc   1/1     Running   0          4s
[root@k8s-master01 sc_work]# kubectl get pod
NAME                 READY   STATUS    RESTARTS   AGE
nginx-relica-4vjl9   1/1     Running   0          16s
nginx-relica-5l7w2   1/1     Running   0          16s
nginx-relica-t6qdc   1/1     Running   0          16s