hadoop - Docker 容器独立运行但在 kubernetes 中失败

Question

我有 docker 容器(Hadoop 安装 https://github.com/kiwenlau/hadoop-cluster-docker)，我可以使用 sudo docker run -itd -p 50070:50070 -p 8088:8088 --name hadoop-master kiwenlau/hadoop:1.0 命令没有任何问题，但是当尝试将相同的图像部署到 kubernetes 时，pod 无法启动。为了创建部署，我使用了 kubectl run hadoop-master --image=kiwenlau/hadoop:1.0 --port=8088 --port=50070 命令

这里是describe pod命令日志

Events:
  FirstSeen     LastSeen        Count   From                                                    SubObjectPath                   Type            Reason          Message
  ---------     --------        -----   ----                                                    -------------                   --------        ------          -------
  6m            6m              1       default-scheduler                                                                       Normal          Scheduled       Successfully assigned hadoop-master-2828539450-rnwsd to gke-mtd-cluster-default-pool-6b97d4d0-hcbt
  6m            6m              1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal          Created         Created container with id 1560ff87e0e7357c76cec89f5f429e0b9b5fc51523c79e4e2c12df1834d7dd75
  6m            6m              1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal          Started         Started container with id 1560ff87e0e7357c76cec89f5f429e0b9b5fc51523c79e4e2c12df1834d7dd75
  6m            6m              1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal          Created         Created container with id c939d3336687a33e69d37aa73177e673fd56d766cb499a4235e89d554d233c37
  6m            6m              1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal          Started         Started container with id c939d3336687a33e69d37aa73177e673fd56d766cb499a4235e89d554d233c37
  6m            6m              2       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning         FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 10s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

  6m    6m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Created         Created container with id 7d1c67686c039e459ee0ea3936eedb4996a5201f6a1fec02ac98d219bb07745f
  6m    6m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Started         Started container with id 7d1c67686c039e459ee0ea3936eedb4996a5201f6a1fec02ac98d219bb07745f
  6m    6m      2       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 20s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

  5m    5m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Started         Started container with id a8879a2c794b3e62f788ad56e403cb619644e9219b2c092e760ddeba506b2e44
  5m    5m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Created         Created container with id a8879a2c794b3e62f788ad56e403cb619644e9219b2c092e760ddeba506b2e44
  5m    5m      3       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 40s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

  5m    5m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Created         Created container with id 8907cdf19c51b87cea6e1e611649e874db2c21f47234df54bd9f27515cee0a0e
  5m    5m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Started         Started container with id 8907cdf19c51b87cea6e1e611649e874db2c21f47234df54bd9f27515cee0a0e
  5m    3m      7       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 1m20s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

  3m    3m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Created         Created container with id 294072caea596b47324914a235c1882dbc521cc355644a1e25ebf06f0e04301f
  3m    3m      1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Started         Started container with id 294072caea596b47324914a235c1882dbc521cc355644a1e25ebf06f0e04301f
  3m    1m      12      kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 2m40s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

  6m    50s     7       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Pulled          Container image "kiwenlau/hadoop:1.0" already present on machine
  50s   50s     1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Created         Created container with id 7da7508ac864d04d47639b0d2c374a27c3e8a3351e13a2564e57453cf857426d
  50s   50s     1       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Normal  Started         Started container with id 7da7508ac864d04d47639b0d2c374a27c3e8a3351e13a2564e57453cf857426d
  6m    0s      31      kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt     spec.containers{hadoop-master}  Warning BackOff         Back-off restarting failed container
  49s   0s      5       kubelet, gke-mtd-cluster-default-pool-6b97d4d0-hcbt                                     Warning FailedSync      Error syncing pod, skipping: failed to "StartContainer" for "hadoop-master" with CrashLoopBackOff: "Back-off 5m0s restarting failed container=hadoop-master pod=hadoop-master-2828539450-rnwsd_default(562dae39-a757-11e7-a5a3-42010a8401c6)"

kubectl 日志输出:

kubectl logs hadoop-master-2828539450-rnwsd
 * Starting OpenBSD Secure Shell server sshd
   ...done.

请注意，docker 容器本身并未启动 hadoop。为了启动它，我必须连接到容器并手动启动 hadoop，但是我现在希望能够在 K8s 中简单地运行容器。

谢谢

Answer 1

kubernetes 中的等效命令是

kubectl run -it hadoop-master --image=kiwenlau/hadoop:1.0 --port=8088 --port=50070

其实不是kubernetes的问题，Dockerfile有问题

A docker container exits when its main process finishes.

使用 CMD [ "sh", "-c", "service ssh start; bash"]，SSH 服务在后台启动，然后容器在作业完成后停止。

它应该是一个始终在前台运行的可执行脚本/程序，如~/start-hadoop.sh。

仅供引用，通常我们不需要通过 ssh 连接到容器，因为 docker exec -it some_container bash 应该足够了。