工作日志——k8s job
本次迭代的功能是基于k8s job实现的。与大家分享一下job相关内容。
相关资料
官方介绍：
http://kubernetes.io/docs/user-guide/jobs/
API：
http://kubernetes.io/docs/api-reference/batch/v1/operations/
分享内容
job与rc不同，包含的pod多用于执行一次性任务、批处理工作等，执行完成后pod便会停止（status.phase变为Succeeded）。这时通过kubectl get pods是看不到pod的，需要加“-a”参数。
RestartPolicy
job pod的template的RestartPolicy只能指定Never或OnFailure，当job未完成的情况下：
* 如果RestartPolicy指定Never，则job会在pod出现故障时创建新的pod，且故障pod不会消失。.status.failed加1。
* 如果RestartPolicy指定OnFailure，则job会在pod出现故障时其内部重启容器，而不是创建pod。.status.failed不变。
Never策略示例：
job.yaml如下所示，其中command故意将perl指令写错，这样pod将出现故障。
apiVersion: batch/v1
kind: Job
metadata:
name: pi
spec:
template:
metadata:
name: pi
spec:
containers:
- name: pi
image: index.tenxcloud.com/sdvdxl/perl
command: ["perl1", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: Never创建job并查看job信息：
$ kubectl create -f job.yaml
job "pi" created
$ kubectl describe job/pi
Name: pi
Namespace: default
Image(s): index.tenxcloud.com/sdvdxl/perl
Selector: controller-uid=cb44e05e-18f7-11e6-b739-0800278856e6
Parallelism: 1
Completions: 1
Start Time: Fri, 13 May 2016 10:45:22 +0000
Labels: controller-uid=cb44e05e-18f7-11e6-b739-0800278856e6,job-name=pi
Pods Statuses: 1 Running / 0 Succeeded / 41 Failed
No volumes.
Events:
FirstSeen LastSeen Count From SubobjectPath Type Reason Message
--------- -------- ----- ---- ------------- -------- ------ -------
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-fvbql
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-f8qjn
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-83rn2
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-j525m
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-msjtr
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-rnqh4
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-vvp02
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-f4rs4
1m 1m 1 {job-controller } Normal SuccessfulCreate Created pod: pi-gazi9
1m 1s 33 {job-controller } Normal SuccessfulCreate (events with common reason combined)可以发现job在不停的创建新的pod，但只有一个pod在执行任务。
$ kubectl get pods -a |grep pi|wc -l
71
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
pi-9d1iv 0/1 ContainerCreating 0 0s执行kubectl delete时可以看到删除了所有pod，耗时较长。
OnFailure策略示例：
job.yaml如下所示，其中command故意将perl指令写错，这样pod将出现故障。
apiVersion: batch/v1
kind: Job
metadata:
name: pi
spec:
template:
metadata:
name: pi
spec:
containers:
- name: pi
image: index.tenxcloud.com/sdvdxl/perl
command: ["perl1", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: OnFailure创建job并查看job信息：
$ kubectl create -f job.yaml
job "pi" created
$ kubectl describe job/pi
Name: pi
Namespace: default
Image(s): index.tenxcloud.com/sdvdxl/perl
Selector: controller-uid=36422362-18fa-11e6-b739-0800278856e6
Parallelism: 1
Completions: 1
Start Time: Fri, 13 May 2016 11:02:41 +0000
Labels: controller-uid=36422362-18fa-11e6-b739-0800278856e6,job-name=pi
Pods Statuses: 1 Running / 0 Succeeded / 0 Failed
No volumes.
Events:
FirstSeen LastSeen Count From SubobjectPath Type Reason Message
--------- -------- ----- ---- ------------- -------- ------ -------
11s 11s 1 {job-controller } Normal SuccessfulCreate Created pod: pi-2n6pa可以看到job只创建了一个pod，查看pod描述可以发现该pod一直在重启容器：
$ kubectl describe pod/pi-2n6pa
...
Events:
FirstSeen LastSeen Count From SubobjectPath Type Reason Message
--------- -------- ----- ---- ------------- -------- ------ -------
2m 2m 1 {default-scheduler } Normal Scheduled Successfully assigned pi-2n6pa to 127.0.0.1
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Normal Created Created container with docker id 310aac97cc9a
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Warning Failed Failed to start container with docker id 310aac97cc9a with error: API error (404): Container command 'perl1' not found or does not exist.
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Normal Created Created container with docker id 62cab877f860
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Warning Failed Failed to start container with docker id 62cab877f860 with error: API error (404): Container command 'perl1' not found or does not exist.
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Normal Created Created container with docker id b1cda2dbee87
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Warning Failed Failed to start container with docker id b1cda2dbee87 with error: API error (404): Container command 'perl1' not found or does not exist.
2m 2m 1 {kubelet 127.0.0.1} Warning FailedSync Error syncing pod, skipping: failed to "StartContainer" for "pi" with CrashLoopBackOff: "Back-off 20s restarting failed container=pi pod=pi-2n6pa_default(3643a8cc-18fa-11e6-b739-0800278856e6)"
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Warning Failed Failed to start container with docker id fc88d454f0c2 with error: API error (404): Container command 'perl1' not found or does not exist.
2m 2m 1 {kubelet 127.0.0.1} spec.containers{pi} Normal Created Created container with docker id fc88d454f0c2
1m 1m 3 {kubelet 127.0.0.1} Warning FailedSync Error syncing pod, skipping: failed to "StartContainer" for "pi" with CrashLoopBackOff: "Back-off 40s restarting failed container=pi pod=pi-2n6pa_default(3643a8cc-18fa-11e6-b739-0800278856e6)"
2m 1m 5 {kubelet 127.0.0.1} spec.containers{pi} Normal Pulling pulling image "index.tenxcloud.com/sdvdxl/perl"
2m 1m 5 {kubelet 127.0.0.1} spec.containers{pi} Normal Pulled Successfully pulled image "index.tenxcloud.com/sdvdxl/perl"
1m 1m 1 {kubelet 127.0.0.1} spec.containers{pi} Warning Failed Failed to start container with docker id 6a6464d1f31b with error: API error (404): Container command 'perl1' not found or does not exist.
2m 1m 5 {kubelet 127.0.0.1} Warning FailedSync Error syncing pod, skipping: failed to "StartContainer" for "pi" with RunContainerError: "runContainer: API error (404): Container command 'perl1' not found or does not exist.\n"
1m 1m 1 {kubelet 127.0.0.1} spec.containers{pi} Normal Created Created container with docker id 6a6464d1f31b
2m 3s 12 {kubelet 127.0.0.1} spec.containers{pi} Warning BackOff Back-off restarting failed docker container
1m 3s 8 {kubelet 127.0.0.1} Warning FailedSync Error syncing pod, skipping: failed to "StartContainer" for "pi" with CrashLoopBackOff: "Back-off 1m20s restarting failed container=pi pod=pi-2n6pa_default(3643a8cc-18fa-11e6-b739-0800278856e6)"设置超时
job执行超时时间可以通过spec.activeDeadlineSeconds来设置，超过指定时间未完成的job会以DeadlineExceeded原因停止：
$ kubectl describe job pi
Name: pi
Namespace: default
Image(s): index.tenxcloud.com/sdvdxl/perl
Selector: controller-uid=8c172a2d-18fb-11e6-b739-0800278856e6
Parallelism: 1
Completions: 1
Start Time: Fri, 13 May 2016 11:12:14 +0000
Active Deadline Seconds: 1s
Labels: controller-uid=8c172a2d-18fb-11e6-b739-0800278856e6,job-name=pi
Pods Statuses: 0 Running / 0 Succeeded / 1 Failed
No volumes.
Events:
FirstSeen LastSeen Count From SubobjectPath Type Reason Message
--------- -------- ----- ---- ------------- -------- ------ -------
7s 7s 1 {job-controller } Normal SuccessfulCreate Created pod: pi-eq3yt
4s 4s 1 {job-controller } Normal SuccessfulDelete Deleted pod: pi-eq3yt
4s 4s 1 {job-controller } Normal DeadlineExceeded Job was active longer than specified deadline删除job
通过API删除job时其对应的pod并不会自动被删除，需要手动调用删除pod的API，删除时指定query参数labelSelector=job-name=
停止job
设置.spec.parallelism为0可以将job停止。此时job不会创建Pod，且会删除所有已运行的Pod（运行结束的不会被删除）。
pod selector
job同样可以指定selector来关联pod。需要注意的是job目前可以使用两个API组来操作，batch/v1和extensions/v1beta1。当用户需要自定义selector时，使用两种API组时定义的参数有所差异。
- 使用batch/v1时，用户需要将jod的spec.manualSelector设置为true，才可以定制selector。默认为false。
- 使用extensions/v1beta1时，用户不需要额外的操作。因为extensions/v1beta1的spec.autoSelector默认为false，该项与batch/v1的spec.manualSelector含义正好相反。换句话说，使用extensions/v1beta1时，用户不想定制selector时，需要手动将spec.autoSelector设置为true。
多容器
如果Job中定义了多个容器，则Job的状态将根据所有容器的执行状态来变化。如果Job定义的容器中存在http server、mysql等长期的容器和一些批处理容器，则Job状态不会发生变化（因为长期运行的容器不会主动结束）。此时可以通过Pod的.status.containerStatuses获取指定容器的运行状态。