Job,主要用于**负责批量处理(一次要处理指定数量任务)短暂的一次性(每个任务仅运行一次就结束)**任务。Job特点如下:
- 当Job创建的pod执行成功结束时,Job将记录成功结束的pod数量
- 当成功结束的pod达到指定的数量时,Job将完成执行
Job的资源清单文件:
apiVersion: batch/v1 # 版本号
kind: Job # 类型
metadata: # 元数据
name: # rs名称
namespace: # 所属命名空间
labels: #标签
controller: job
spec: # 详情描述
completions: 1 # 指定job需要成功运行Pods的次数。默认值: 1
parallelism: 1 # 指定job在任一时刻应该并发运行Pods的数量。默认值: 1
activeDeadlineSeconds: 30 # 指定job可运行的时间期限,超过时间还未结束,系统将会尝试进行终止。
backoffLimit: 6 # 指定job失败后进行重试的次数。默认是6
manualSelector: true # 是否可以使用selector选择器选择pod,默认是false
selector: # 选择器,通过它指定该控制器管理哪些pod
matchLabels: # Labels匹配规则
app: counter-pod
matchExpressions: # Expressions匹配规则
- {key: app, operator: In, values: [counter-pod]}
template: # 模板,当副本数量不足时,会根据下面的模板创建pod副本
metadata:
labels:
app: counter-pod
spec:
restartPolicy: Never # 重启策略只能设置为Never或者OnFailure
containers:
- name: counter
image: busybox:1.30
command: ["bin/sh","-c","for i in 9 8 7 6 5 4 3 2 1; do echo $i;sleep 2;done"]
关于重启策略设置的说明:
- 如果指定为OnFailure,则job会在pod出现故障时重启容器,而不是创建pod,failed次数不变
- 如果指定为Never,则job会在pod出现故障时创建新的pod,并且故障pod不会消失,也不会重启,failed次数加1
- 如果指定为Always的话,就意味着一直重启,意味着job任务会重复去执行了,当然不对,所以不能设置为Always
cat >test.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
name: pc-job
namespace: dev
spec:
manualSelector: true
selector:
matchLabels:
app: counter-pod
template:
metadata:
labels:
app: counter-pod
spec:
restartPolicy: Never
containers:
- name: counter
image: busybox:1.30
command: ["bin/sh","-c","for i in $(seq 1 9); do echo $i;sleep 1;done"]
EOF
# 创建job
kubectl create -f test.yaml
job.batch/pc-job created
# 查看job
kubectl get job -n dev -o wide -w
NAME COMPLETIONS DURATION AGE CONTAINERS IMAGES SELECTOR
pc-job 0/1 21s 21s counter busybox:1.30 app=counter-pod
pc-job 1/1 31s 79s counter busybox:1.30 app=counter-pod
# 通过观察pod状态可以看到,pod在运行完毕任务后,就会变成Completed状态
kubectl get pods -n dev -w
NAME READY STATUS RESTARTS AGE
pc-job-rxg96 1/1 Running 0 29s
pc-job-rxg96 0/1 Completed 0 33s
kubectl get pods -n dev -w
'''
接下来,调整下pod运行的总数量和并行数量 即:在spec下设置下面两个选项
completions: 6 指定job需要成功运行Pods的次数为6
parallelism: 3 指定job并发运行Pods的数量为3
运行job,观察效果,此时会发现,job会每次运行3个pod,总共执行了6个pod
NAME READY STATUS RESTARTS AGE
pc-job-684ft 1/1 Running 0 5s
pc-job-jhj49 1/1 Running 0 5s
pc-job-pfcvh 1/1 Running 0 5s
pc-job-684ft 0/1 Completed 0 11s
pc-job-v7rhr 0/1 Pending 0 0s
pc-job-v7rhr 0/1 Pending 0 0s
pc-job-v7rhr 0/1 ContainerCreating 0 0s
pc-job-jhj49 0/1 Completed 0 11s
pc-job-fhwf7 0/1 Pending 0 0s
pc-job-fhwf7 0/1 Pending 0 0s
pc-job-pfcvh 0/1 Completed 0 11s
pc-job-5vg2j 0/1 Pending 0 0s
pc-job-fhwf7 0/1 ContainerCreating 0 0s
pc-job-5vg2j 0/1 Pending 0 0s
pc-job-5vg2j 0/1 ContainerCreating 0 0s
pc-job-fhwf7 1/1 Running 0 2s
pc-job-v7rhr 1/1 Running 0 2s
pc-job-5vg2j 1/1 Running 0 3s
pc-job-fhwf7 0/1 Completed 0 12s
pc-job-v7rhr 0/1 Completed 0 12s
pc-job-5vg2j 0/1 Completed 0 12s
'''
# 删除job
kubectl delete -f test.yaml
'''
backoffLimit: 一旦有1个pod的c重试次数达到backoffLimit,job就会失败。
每次back-off的时间间隔会逐渐增大,第一次10s、第二次20s、第三次40s,封顶6min
activeDeadlineSeconds: 设置为多少秒,超过后所有的Pod将会终止且Job的状态将会就变成失败。
优先级高于backoffLimit,因此Job重试过一个或多个失败的Pod达到限制的时间(activeDeadlineSeconds)
后将不会部署额外的Pod,即使此时backoffLimit还没达到。
'''
cat >test.yaml <<EOF
apiVersion: batch/v1
kind: Job
metadata:
name: pc-job
namespace: dev
spec:
completions: 6
parallelism: 1
backoffLimit: 3
activeDeadlineSeconds: 20
manualSelector: true
selector:
matchLabels:
app: counter-pod
template:
metadata:
labels:
app: counter-pod
spec:
restartPolicy: Never
containers:
- name: counter
image: busybox:1.30
command: ["666"]
EOF