Created by: Daemon007

我参考paddlepaddle进阶指南在kubernetes上进行单机训练。http://doc.paddlepaddle.org/doc_cn/howto/usage/k8s/k8s_cn.html

已经将包含训练数据的mypaddle/paddle:quickstart镜像放到本地docker上，并且同时放到私有仓库registry上registry.vm-1:5000/mypaddle/paddle:quickstart

yaml文件配置如下(仅仅修改了image名称为registry.vm-1:5000/mypaddle/paddle:quickstart)：

apiVersion: batch/v1
kind: Job
metadata:
  name: quickstart
spec:
  parallelism: 1
  completions: 1
  template:
    metadata:
      name: quickstart
    spec:
      volumes:
      - name: output
        hostPath: 
          path: /home/work/paddle_output     
      containers:
      - name: pi
        image: registry.vm-1:5000/mypaddle/paddle:quickstart
        command: ["bin/bash",  "-c", "/root/paddle/demo/quick_start/train.sh"]
        volumeMounts:
        - name: output
          mountPath: /root/paddle/demo/quick_start/output
      restartPolicy: Never

查看训练结果

[root@localhost work]# kubectl get pods
NAME                   READY     STATUS             RESTARTS   AGE
private-image-test-1   0/1       CrashLoopBackOff   11         34m
[root@localhost work]# kubectl  describe pod quickstart-bgxtf
Name:		quickstart-bgxtf
Namespace:	default
Node:		vm-2/192.168.1.48
Start Time:	Sun, 16 Jul 2017 03:17:51 -0700
Labels:		controller-uid=0606a0ea-6a10-11e7-9839-000c291ffd39
		job-name=quickstart
Status:		Succeeded
IP:		10.0.1.3
Controllers:	Job/quickstart
Containers:
  pi:
    Container ID:	docker://613653ff059d67224d7d1c272150f45bb06ffb175fb59a6a93305b9591994b45
    Image:		registry.vm-1:5000/mypaddle/paddle:quickstart
    Image ID:		docker-pullable://registry.vm-1:5000/mypaddle/paddle@sha256:df130bf3ebb08d819e0899141a374b1f14e94e83142a3c3e9d3618b583add7b3
    Port:		
    Command:
      bin/bash
      -c
      /root/paddle/demo/quick_start/train.sh
    State:		Terminated
      Reason:		Completed
      Exit Code:	0
      Started:		Sun, 16 Jul 2017 03:21:45 -0700
      Finished:		Sun, 16 Jul 2017 03:21:51 -0700
    Ready:		False
    Restart Count:	0
    Volume Mounts:
      /root/paddle/demo/quick_start/output from output (rw)
    Environment Variables:	<none>
Conditions:
  Type		Status
  Initialized 	True 
  Ready 	False 
  PodScheduled 	True 
Volumes:
  output:
    Type:	HostPath (bare host directory volume)
    Path:	/home/work/paddle_output
QoS Class:	BestEffort
Tolerations:	<none>
Events:
  FirstSeen	LastSeen	Count	From			SubObjectPath		Type		Reason			Message
  ---------	--------	-----	----			-------------		--------	------			-------
  28m		28m		1	{default-scheduler }				Normal		Scheduled		Successfully assigned quickstart-bgxtf 
to vm-2  28m		28m		1	{kubelet vm-2}		spec.containers{pi}	Normal		Pulling			pulling image "registry.vm-1:5000/mypad
dle/paddle:quickstart"  28m		25m		2	{kubelet vm-2}					Warning		MissingClusterDNS	kubelet does not have ClusterDNS IP con
figured and cannot create Pod using "ClusterFirst" policy. Falling back to DNSDefault policy.  25m		25m		1	{kubelet vm-2}		spec.containers{pi}	Normal		Pulled			Successfully pulled image "registry.vm-
1:5000/mypaddle/paddle:quickstart"  25m		25m		1	{kubelet vm-2}		spec.containers{pi}	Normal		Created			Created container with docker id 613653
ff059d; Security:[seccomp=unconfined]  25m		25m		1	{kubelet vm-2}		spec.containers{pi}	Normal		Started			Started container with docker id 613653
ff059d

显示pod已经结束，但是宿主机上没有训练的结果数据。 请问这是什么原因导致的？