前言
容器分为两类
-
在线业务 - 服务类容器
需要持续不中断的提供服务,容器需要一直运行
-
离线业务 - 工作类容器
一般执行一次性任务,比如统计日志数据等,运行完成后容器即可关闭
更新历史
- 20200618 - 初稿 - 左程立
- 原文地址 - https://blog.zuolinux.com/2020/06/18/controller-job-cronjob.html
Job
cat job.yaml
apiVersion: batch/v1
kind: Job
metadata:
name: job-test
spec:
template:
metadata:
name: job-test
spec:
containers:
- name: test-job
image: busybox
command: ["echo", "test job!"]
restartPolicy: Never
参数
restartPolicy 只能是 Never 或者 onFailure
启动
[root@master01 ~]# kubectl apply -f job.yaml
job.batch/job-test created
运行状态
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-t2gbw 0/1 ContainerCreating 0 12s
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-t2gbw 0/1 Completed 0 24s
查看运行结果
[root@master01 ~]# kubectl logs job-test-t2gbw
test job!
Job 执行失败如何处理
将 command 胡乱设置导致 job 无法成功启动
command: ["echo123", "test job!"]
restartPolicy: Never
当 restartPolicy: Never 时,可以看到 k8s 在不断的开新的 pod
但不会让它一直开下去,默认参数 spec.backoffLimit: 6 会进行阻止
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-2zrt4 0/1 ContainerCannotRun 0 87s
job-test-5z884 0/1 ContainerCannotRun 0 77s
job-test-cxgmm 0/1 ContainerCannotRun 0 107s
job-test-nt9gt 0/1 ContainerCannotRun 0 57s
job-test-wxv7l 0/1 ContainerCreating 0 17s
当 restartPolicy: OnFailure 时,k8s 不开新 pod,只会不断重启 pod
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-bd695 0/1 CrashLoopBackOff 4 3m15s
如果 job 没有报错,但是它一直不结束怎么办?
参数 spec.activeDeadlineSeconds: 100 会在 100s 后将 Job 中所有 Pod 进行关闭
此时 Pod 关闭状态为 reason: DeadlineExceeded
并行 Job
cat job.yaml
apiVersion: batch/v1
kind: Job
metadata:
name: job-test
spec:
parallelism: 2
......
参数
parallelism: 2 仅允许最多两个 Pod 同时运行
查看运行情况
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-6ttz9 0/1 ContainerCreating 0 8s
job-test-mknjc 0/1 ContainerCreating 0 8s
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-6ttz9 0/1 Completed 0 68s
job-test-mknjc 0/1 Completed 0 68s
[root@master01 ~]# kubectl get job
NAME COMPLETIONS DURATION AGE
job-test 0/1 of 2 10s 10s
[root@master01 ~]# kubectl get job
NAME COMPLETIONS DURATION AGE
job-test 2/1 of 2 19s 66s
增加参数 completions
apiVersion: batch/v1
kind: Job
metadata:
name: job-test
spec:
parallelism: 2
completions: 8
参数
completions: 8 表示至少成功运行 8 个 pod
如果不是8个成功,那么会根据 restartPolicy 的策略进行处理
可以认为是一种检查机制
查看运行情况
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-6xwpt 0/1 ContainerCreating 0 8s
job-test-grk4q 0/1 ContainerCreating 0 8s
[root@master01 ~]# kubectl get job
NAME COMPLETIONS DURATION AGE
job-test 0/8 11s 11s
[root@master01 ~]# kubectl get job
NAME COMPLETIONS DURATION AGE
job-test 6/8 76s 76s
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
job-test-2jxqx 0/1 Completed 0 28s
job-test-6xwpt 0/1 Completed 0 90s
job-test-gndq7 0/1 Completed 0 70s
job-test-grk4q 0/1 Completed 0 90s
job-test-n96k6 0/1 Completed 0 70s
job-test-np4n7 0/1 Completed 0 49s
job-test-scc9c 0/1 Completed 0 28s
job-test-zcnbp 0/1 Completed 0 49s
[root@master01 ~]# kubectl get job
NAME COMPLETIONS DURATION AGE
job-test 8/8 82s 91s
CronJob
CronJob 会定时 create 一个 job 对象
cat cronjob.yaml
apiVersion: batch/v1beta1
kind: CronJob
metadata:
name: hello
spec:
schedule: "*/1 * * * *"
jobTemplate:
spec:
template:
spec:
containers:
- name: hello
image: busybox
command: ["echo","test cron job!"]
restartPolicy: OnFailure
查看运行结果
[root@master01 ~]# kubectl get cronjob
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
hello */1 * * * * False 0 67s 3m45s
[root@master01 ~]# kubectl get jobs
NAME COMPLETIONS DURATION AGE
hello-1595319480 1/1 27s 3m23s
hello-1595319540 1/1 19s 2m23s
hello-1595319600 1/1 24s 83s
hello-1595319660 0/1 22s 22s
[root@master01 ~]# kubectl get pod
NAME READY STATUS RESTARTS AGE
hello-1595319480-2kl8h 0/1 Completed 0 3m21s
hello-1595319540-gwthv 0/1 Completed 0 2m21s
hello-1595319600-54548 0/1 Completed 0 81s
hello-1595319660-8xqn6 0/1 ContainerCreating 0 20s
可以看到,就是把前面说过的 Job 对象进行定时的运行。
结束语
Job/CronJob 控制器可以进行离线业务的处理,不需要长期运行一个 Pod。
这类控制器管理的 Pod 不对外提供持续在线服务,任务运行完成后,即会将容器进行关闭,可以用于一些定时数据处理类的工作。