我正在尝试在gcp计算引擎上运行启动脚本。该脚本启动，启动docker容器，并从该容器中运行命令。在启动脚本之外，整个过程没有错误，但显然启动脚本不是为此任务设计的吗？
我打电话。
gcloud compute instances create cloudml
--image-family=container-vm
--image-project=google-containers
--boot-disk-size "40"
--service-account "773889352370-compute@developer.gserviceaccount.com"
--scopes "https://www.googleapis.com/auth/cloud-platform"
--metadata-from-file startup-script=Training.sh
我可以通过串口监控启动脚本的日志。我可以看到它正确抓取脚本并开始。
第一个命令是docker run
sudo docker run -it --privileged -- gcr.io/api-project-773889352370/cloudmlengine
从日志中，它会拉动和提取
Jul 24 21:54:02 cloudml startupscript: latest: Pulling from api-project-773889352370/cloudmlengine
Jul 24 21:54:02 cloudml startupscript: 9f0706ba7422: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: ec542cd45f0c: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: 07bcbd00d9b4: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: acc6d9961302: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: ef130a4b8961: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: 5ecac320219d: Pulling fs layer
Jul 24 21:54:02 cloudml startupscript: 54110f0dc661: Pulling fs layer
并最终
downloaded newer image for gcr.io/api-project-773889352370/cloudmlengine:latest
大。该命令应该启动一个容器。我希望从容器中运行脚本的其余部分。从ssh这没问题，docker run -it给我带来了一个交互式提示，但显然启动脚本不能同样工作。下一个命令是
git clone https://github.com/bw4sz/DeepMeerkat.git
但是整个事情有点没有明显的错误
Jul 24 21:59:53 [ 422.646195] IPv6: ADDRCONF(NETDEV_UP): veth8286626: link is not ready
cloudml kernel: [ 422.639552] device veth8286626 entered promiscuous mode
Jul 24 21:59:53 cloudml kernel: [ 422.646195] IPv6: ADDRCONF(NETDEV_UP): veth8286626: link is not ready
[ 422.938314] IPv6: ADDRCONF(NETDEV_CHANGE): veth8286626: link becomes ready
[ 422.945723] docker0: port 1(veth8286626) entered forwarding state
[ 422.951977] docker0: port 1(veth8286626) entered forwarding state
[ 422.962994] IPv6: ADDRCONF(NETDEV_CHANGE): docker0: link becomes ready
Jul 24 21:59:53 cloudml kernel: [ 422.938314] IPv6: ADDRCONF(NETDEV_CHANGE): veth8286626: link becomes ready
Jul 24 21:59:53 cloudml kernel: [ 422.945723] docker0: port 1(veth8286626) entered forwarding state
Jul 24 21:59:53 cloudml kernel: [ 422.951977] docker0: port 1(veth8286626) entered forwarding state
Jul 24 21:59:53 cloudml kernel: [ 422.962994] IPv6: ADDRCONF(NETDEV_CHANGE): docker0: link becomes ready
[ 437.956027] docker0: port 1(veth8286626) entered forwarding state
Jul 24 22:00:08 cloudml kernel: [ 437.956027] docker0: port 1(veth8286626) entered forwarding state
Jul 24 22:00:29 cloudml sshd[3043]: Did not receive identification string from 139.162.75.112
Jul 24 22:02:26 cloudml sshd[3079]: Invalid user ftpuser from 128.199.72.218
Jul 24 22:02:26 cloudml sshd[3079]: input_userauth_request: invalid user ftpuser [preauth]
Jul 24 22:02:26 cloudml sshd[3079]: Connection closed by 128.199.72.218 [preauth]
Jul 24 22:06:07 cloudml sshd[3168]: Invalid user ftpuser1 from 128.199.72.218
Jul 24 22:06:07 cloudml sshd[3168]: input_userauth_request: invalid user ftpuser1 [preauth]
Jul 24 22:06:07 cloudml sshd[3168]: Connection closed by 128.199.72.218 [preauth]
它会持续一段时间。
我可以稍后ssh并尝试docker命令和下一个命令（git clone）并且它们运行没有问题，因此docker pull成功，没有错误。
Ben@cloudml:~$ sudo docker run -it --privileged -- gcr.io/api-project-773889352370/cloudmlengine
root@7d8ae18c2295:/#
root@7d8ae18c2295:/# git clone https://github.com/bw4sz/DeepMeerkat.git
Cloning into 'DeepMeerkat'...
remote: Counting objects: 768, done.
remote: Total 768 (delta 0), reused 0 (delta 0), pack-reused 768
Receiving objects: 100% (768/768), 41.38 MiB | 44.71 MiB/s, done.
Resolving deltas: 100% (517/517), done.
Checking connectivity... done.
root@7d8ae18c2295:/#
我是否会采取这种错误的方式？也许信息here这只是一次性工作，创建一个完整的kubernetes集群似乎是错误的。