k8s的job是用来执行一次性任务的一类资源，相关的还有cronjob，用于执行以下周期性任务。
部署job之后，k8s会起对应pod，当pod的状态为finished之后，job的状态会更新为complete，即这个job任务已经执行完成，pod不在系统中继续运行。
相对于ReplicaSet、ReplicationController等controller而言，k8s的job是当pod完成某项任务之后，退出对应的pod，而ReplicaSet、ReplicationController是保证了k8s集群环境中始终保持有对应数量的pod在运行。因此可以说job是ReplicaSet、ReplicationController等controller等的补充。
下面通过几个例子来看看k8s job机制的使用：
1. 创建一个简单的job，用于计算pi的值，精度为小数点后2000位，对应yaml：
apiVersion: batch/v1
kind: Job
metadata:
name: pi
spec:
template:
spec:
containers:
- name: pi
image: perl
command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: Never
backoffLimit: 4此时k8s会在node上起一个pod，当pod完成计算任务之后，状态会更新为complete，而job的状态也是complete
上述例子中，当一个pod完成任务之后，job的状态即为complete，也就是completions的值达到1时则认为job已经完成。其实我们通过指定job的completions值，来决定job完成的条件。
2. 创建一个job任务，当completions值达到3时job才完成，并使用并发数量为2，即同一时刻该job有对应两个pod在工作，eg：
[root@calico learn]# cat job-parallel.yaml
apiVersion: batch/v1
kind: Job
metadata:
name: pi
spec:
parallelism: 2
completions: 3
template:
spec:
containers:
- name: pi
image: perl
command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: Never
backoffLimit: 4此时查看pod的状态：
#先创建2个pod用于执行任务
[root@calico learn]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pi-2xlxl 0/1 ContainerCreating 0 23s <none> calico-node1 <none> <none>
pi-7wcbj 0/1 ContainerCreating 0 23s <none> calico-node2 <none> <none>
...
#等一会之后，2个pod完成，再起第3个pod来完成第三个completion目标
[root@calico learn]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pi-2xlxl 0/1 Completed 0 7m14s 192.168.63.150 calico-node1 <none> <none>
pi-7wcbj 0/1 Completed 0 7m14s 192.168.186.85 calico-node2 <none> <none>
pi-996x8 1/1 Running 0 3m8s 192.168.63.151 calico-node1 <none> <none>
#查看此时job的状态，completions为2/3
[root@calico learn]# kubectl get job -o wide
NAME COMPLETIONS DURATION AGE CONTAINERS IMAGES SELECTOR
pi 2/3 7m26s 7m27s pi perl controller-uid=9736793a-5075-11e9-a970-5254000ebe60
...
#此时3个pod都完成任务，状态都为complete
[root@calico learn]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pi-2xlxl 0/1 Completed 0 8m37s 192.168.63.150 calico-node1 <none> <none>
pi-7wcbj 0/1 Completed 0 8m37s 192.168.186.85 calico-node2 <none> <none>
pi-996x8 0/1 Completed 0 4m31s 192.168.63.151 calico-node1 <none> <none>
#查看此时job的状态，completions为3/3
[root@calico learn]# kubectl get job -o wide
NAME COMPLETIONS DURATION AGE CONTAINERS IMAGES SELECTOR
pi 3/3 8m38s 9m27s pi perl controller-uid=9736793a-5075-11e9-a970-5254000ebe60在job的yaml中，有个参数为backoffLimit。该参数为job在执行过程中，指定任务执行失败重试的次数，但达到backoffLimit后任务仍未成功，则job的状态会更新为failed。
除了backoffLimit用于限制job执行的次数，job的timeout机制也用于限制job执行的时间，当job超过timeout时间，则job的状态也会更新为failed。注意，当job指定了timeout之后，backoffLimit参数就不生效了。
3. 创建一个带timeout的job，通过在yaml中指定activeDeadlineSeconds：
[root@calico learn]# cat job-timeout.yaml
apiVersion: batch/v1
kind: Job
metadata:
name: pi-timeout
spec:
parallelism: 2
completions: 3
backoffLimit: 5
activeDeadlineSeconds: 100
template:
spec:
containers:
- name: pi
image: perl
command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: Never显然该job执行时间会超过100s，因此job会超时失败。
#查看job状态
[root@calico learn]# kubectl get job -o wide
NAME COMPLETIONS DURATION AGE CONTAINERS IMAGES SELECTOR
pi-timeout 0/3 109s 109s pi perl controller-uid=e34174a9-5077-11e9-a970-5254000ebe60
#查看pod状态
[root@calico learn]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
pi-timeout-5rkvs 1/1 Terminating 0 113s 192.168.186.86 calico-node2 <none> <none>
pi-timeout-6cj5k 1/1 Terminating 0 113s 192.168.63.152 calico-node1 <none> <none>可以看到，超过100s之后，job并没有完成，此时会把pod删掉。注意，job的删除是级联的（cascading），因此pod和job都会被删掉。
此时查看job的详细信息，可以看到pod状态是0 Running / 0 Succeeded / 2 Failed：
[root@calico learn]# kubectl describe job pi-timeout
Name: pi-timeout
Namespace: default
Selector: controller-uid=e34174a9-5077-11e9-a970-5254000ebe60
Labels: controller-uid=e34174a9-5077-11e9-a970-5254000ebe60
job-name=pi-timeout
Annotations: <none>
Parallelism: 2
Completions: 3
Start Time: Wed, 27 Mar 2019 06:05:45 -0400
Active Deadline Seconds: 100s
Pods Statuses: 0 Running / 0 Succeeded / 2 Failed
Pod Template:
Labels: controller-uid=e34174a9-5077-11e9-a970-5254000ebe60
job-name=pi-timeout
Containers:
pi:
Image: perl
Port: <none>
Host Port: <none>
Command:
perl
-Mbignum=bpi
-wle
print bpi(2000)
Environment: <none>
Mounts: <none>
Volumes: <none>
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal SuccessfulCreate 2m5s job-controller Created pod: pi-timeout-5rkvs
Normal SuccessfulCreate 2m5s job-controller Created pod: pi-timeout-6cj5k
Normal SuccessfulDelete 25s job-controller Deleted pod: pi-timeout-5rkvs
Normal SuccessfulDelete 25s job-controller Deleted pod: pi-timeout-6cj5k
Warning DeadlineExceeded 25s job-controller Job was active longer than specified deadline在上面的例子中，当pod finished之后（finished指pod的状态是complete或者failed），pod仍然运行在系统上，这会占用系统的资源，对系统造成一定的负担。因此，在k8s 1.12版本之后，引入ttl controller，对应finished状态的pod进行清理。（目前ttl controller也只是对job资源有清理，后续会对其他资源也支持）
当在job的yaml设置ttl后，ttl controller会在pod状态更新为finished之后，经过ttl秒删掉把pod删掉。注意，删除是把pod，job都删掉了
- ttl为0时，finished后马上删掉
- ttl不为n，经过n秒之后删除finished的pod
- ttl不设置，不会删除finished的pod，job
示例：
apiVersion: batch/v1
kind: Job
metadata:
name: pi-with-ttl
spec:
ttlSecondsAfterFinished: 100
template:
spec:
containers:
- name: pi
image: perl
command: ["perl", "-Mbignum=bpi", "-wle", "print bpi(2000)"]
restartPolicy: Never
4. 前面提到，k8s job的删除是级联的，即删掉job之后，也会对应的把pod删掉。那么当我们有时不希望把pod删掉，只对job进行操作呢。比如，pod正在跑现网上的任务，而我们又需要修改job模板，比如job的名字等等，此时我们可以用这个命令：
kubectl delete jobs/old --cascade=false此命令不会删除pod，只会删除对应的job
5. job的RestartPolicy
job支持的RestartPolicy有两种：OnFailure或Never。而在k8s中，RestartPolicy默认是Always。因此部署job时，一定要指定job的RestartPolicy，否则会有问题。
