今天对一个pod进行内存资源调整后, 一直卡在ContainerCreating的状态, 执行describe命令查看该 Pod 详细信息后发现如下 。

[root@master-01 ~]# kubectl describe pod busybox-deploy-b9b6d4ff9-vwfnq ...Events:Type Reason Age From Message---- ------ ---- ---- -------Normal Scheduled 40s default-scheduler Successfully assigned default/busybox-deploy-b9b6d4ff9-vwfnq to 192.168.209.132Warning FailedCreatePodSandBox 28s (x4 over 37s) kubelet, 192.168.209.132 Failed create pod sandbox: rpc error: code = Unknown desc = failed to start sandbox container for pod "busybox-deploy-b9b6d4ff9-vwfnq": Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:297: getting the final child's pid from pipe caused \"read init-p: connection reset by peer\"": unknownWarning FailedCreatePodSandBox 26s (x9 over 38s) kubelet, 192.168.209.132 Failed create pod sandbox: rpc error: code = Unknown desc = failed to start sandbox container for pod "busybox-deploy-b9b6d4ff9-vwfnq": Error response from daemon: OCI runtime create failed: container_linux.go:344: starting container process caused "process_linux.go:293: copying bootstrap data to pipe caused \"write init-p: broken pipe\"": unknownNormal SandboxChanged 26s (x12 over 37s) kubelet, 192.168.209.132 Pod sandbox changed, it will be killed and re-created.

通过查看以上 Event 信息中, 能解读到的信息极其有限 connection reset by peer: 连接被重置。 Pod sandbox changed, it will be killed and re-created: pause 容器引导的 Pod 环境被改变, 重新创建 Pod 中的 pause 引导。 copying bootstrap data to pipe caused "write init-p: broken pipe"": unknown:Google 说的docker和内核不兼容。 参考 https://github.com/docker/for-linux/issues/595