elasticsearch kubernetes pod的Back-off重启失败容器的原因是什么？

Question

当我尝试通过 kubernetes 部署运行我的 elasticsearch 容器时，我的 elasticsearch pod 会在一段时间后失败，虽然当使用 docker-compose 或 Dockerfile 直接作为 docker 容器运行时它运行得非常好。这就是我得到的结果 kubectl get pods

 NAME                  READY     STATUS    RESTARTS   AGE
es-764bd45bb6-w4ckn   0/1       Error     4          3m

下面是 kubectl describe pod 的结果

Name:           es-764bd45bb6-w4ckn
Namespace:      default
Node:           administrator-thinkpad-l480/<node_ip>
Start Time:     Thu, 30 Aug 2018 16:38:08 +0530
Labels:         io.kompose.service=es
            pod-template-hash=3206801662
Annotations:    <none>
Status:         Running
IP:             10.32.0.8
Controlled By:  ReplicaSet/es-764bd45bb6
Containers:
es:
Container ID:   docker://9be2f7d6eb5d7793908852423716152b8cefa22ee2bb06fbbe69faee6f6aa3c3
Image:          docker.elastic.co/elasticsearch/elasticsearch:6.2.4
Image ID:       docker-pullable://docker.elastic.co/elasticsearch/elasticsearch@sha256:9ae20c753f18e27d1dd167b8675ba95de20b1f1ae5999aae5077fa2daf38919e
Port:           9200/TCP
State:          Waiting
  Reason:       CrashLoopBackOff
Last State:     Terminated
  Reason:       Error
  Exit Code:    78
  Started:      Thu, 30 Aug 2018 16:42:56 +0530
  Finished:     Thu, 30 Aug 2018 16:43:07 +0530
Ready:          False
Restart Count:  5
Environment:
  ELASTICSEARCH_ADVERTISED_HOST_NAME:  es
  ES_JAVA_OPTS:                        -Xms2g -Xmx2g
  ES_HEAP_SIZE:                        2GB
Mounts:
  /var/run/secrets/kubernetes.io/serviceaccount from default-token-nhb9z (ro)
Conditions:
  Type              Status
  Initialized       True
  Ready             False
  ContainersReady   False
  PodScheduled      True
Volumes:
  default-token-nhb9z:
    Type:        Secret (a volume populated by a Secret)
    SecretName:  default-token-nhb9z
    Optional:    false
QoS Class:       BestEffort
Node-Selectors:  <none>
Tolerations:     node.kubernetes.io/not-ready:NoExecute for 300s
             node.kubernetes.io/unreachable:NoExecute for 300s
Events:
  Type     Reason     Age               From           Message
  ----     ------     ----              ----           -------
 Normal   Scheduled  6m                default-scheduler                     Successfully assigned default/es-764bd45bb6-w4ckn to administrator-thinkpad-l480
 Normal   Pulled     3m (x5 over 6m)   kubelet, administrator-thinkpad-l480  Container image "docker.elastic.co/elasticsearch/elasticsearch:6.2.4" already present on machine
 Normal   Created    3m (x5 over 6m)   kubelet, administrator-thinkpad-l480  Created container
 Normal   Started    3m (x5 over 6m)   kubelet, administrator-thinkpad-l480  Started container
 Warning  BackOff    1m (x15 over 5m)  kubelet, administrator-thinkpad-l480  Back-off restarting failed container

这是我的 elasticsearc-deployment.yaml：

 apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  annotations:
    kompose.cmd: kompose convert
    kompose.version: 1.1.0 (36652f6)
  creationTimestamp: null
  labels:
    io.kompose.service: es
  name: es
spec:
  replicas: 1
  strategy: {}
  template:
    metadata:
      creationTimestamp: null
      labels:
        io.kompose.service: es
    spec:
      containers:
      - env:
        - name: ELASTICSEARCH_ADVERTISED_HOST_NAME
          value: es
        - name: ES_JAVA_OPTS
          value: -Xms2g -Xmx2g
        - name: ES_HEAP_SIZE
          value: 2GB
        image: docker.elastic.co/elasticsearch/elasticsearch:6.2.4
        name: es
        ports:
        - containerPort: 9200
        resources: {}
      restartPolicy: Always
 status: {}

当我尝试使用 kubectl logs -f es-764bd45bb6-w4ckn 获取日志时，我得到

Error from server: Get https://<slave node ip>:10250/containerLogs/default/es-764bd45bb6-w4ckn/es?previous=true: dial tcp <slave node ip>:10250: i/o timeout

这个问题的原因和解决方案可能是什么？

原文由 Lakshya Garg 发布，翻译遵循 CC BY-SA 4.0 许可协议

Answer 1

我发现日志使用 docker logs 用于 es 容器，发现 es 没有启动，因为 vm.max_map_count 设置为非常低的值。我使用 sysctl -w vm.max_map_count=262144 将 vm.max_map_count 更改为所需的值，然后吊舱已启动。

原文由 Lakshya Garg 发布，翻译遵循 CC BY-SA 4.0 许可协议

Answer 2

我有同样的问题，这个问题可能有几个原因。在我的情况下，jar 文件丢失了。 @Lakshya 已经回答了这个问题，我想添加您可以采取的解决问题的步骤。

1. 获取 pod 状态，命令 - kubectl get pods
2. 描述 pod 以进一步查看 - kubectl describe pod “pod-name” 最后几行输出为您提供事件以及部署失败的位置
3. 获取日志以获取更多详细信息 - kubectl logs “pod-name”
4. 获取容器日志 - kubectl logs “pod-name” -c “container-name” 从 describe pod 命令的输出中获取容器名称

如果您的容器已启动，您可以使用 kubectl exec -it 命令进一步分析容器

希望它可以帮助社区成员解决未来的问题。

原文由 Pradeep 发布，翻译遵循 CC BY-SA 4.0 许可协议