我们前面讲了很多关于Pod的使用,但是在实际应用中,我们不会去直接创建Pod,我们一般通过Kubernetes提供的工作负载(Deployment、DeamonSet、StatefulSet、Job等)完成对一组Pod全生命周期的控制,本节开始我们来看看这些工作负载是如何使用的。
一、Deployments
Deployments可以自动部署一个容器应用的多个副本,监控其副本数量并始终维持这一数量。我们来创建一个Deployments看看:
apiVersion: apps/v1 kind: Deployment metadata: name: nginx-deployment spec: selector: matchLabels: app: nginx replicas: 3 template: metadata: labels: app: nginx spec: containers: - name: nginx image: nginx:latest ports: - containerPort: 80
我们的Deployment会创建并维持三个nginx副本,我们通过kubectl create创建此Deployment:
[root@kubevm1 workspace] kubectl create -f demo_deployment.yml [root@kubevm1 workspace] kubectl get deployments NAME READY UP-TO-DATE AVAILABLE AGE nginx-deployment 3/3 3 3 2m40s [root@kubevm1 workspace] kubectl get pods NAME READY STATUS RESTARTS AGE nginx-deployment-585449566-b6v88 1/1 Running 0 2m4s nginx-deployment-585449566-p4vb7 1/1 Running 0 2m4s nginx-deployment-585449566-s95sr 1/1 Running 0 2m4s
我们看到三个nginx的Pod豆启动完毕了,当我们手动杀掉一个Pod,Deployment会自动将其恢复:
[root@kubevm1 workspace] kubectl delete pod nginx-deployment-585449566-b6v88 [root@kubevm1 workspace] kubectl get pods NAME READY STATUS RESTARTS AGE nginx-deployment-585449566-b6v88 0/1 Terminating 0 7m23s nginx-deployment-585449566-nbrdm 0/1 ContainerCreating 0 6s nginx-deployment-585449566-p4vb7 1/1 Running 0 7m23s nginx-deployment-585449566-s95sr 1/1 Running 0 7m23s
我们前面说过Deployment是自动启动并调度Pod的,这三个Pod最终在哪些节点上运行,完全由master的Scheduler来控制,我们通过-o wide来看看这三个Pod都分布在哪里:
[root@kubevm1 workspace] kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES nginx-deployment-585449566-nbrdm 1/1 Running 0 3m12s 10.244.1.8 kubevm2 <none> <none> nginx-deployment-585449566-p4vb7 1/1 Running 0 10m 10.244.2.5 kubevm3 <none> <none> nginx-deployment-585449566-s95sr 1/1 Running 0 10m 10.244.1.7 kubevm2 <none> <none>
二、DaemonSet
DaemonSet是kubernetes 1.2新增的资源对象,它可以确保在每个Node上仅运行一份Pod的副本,DaemonSet可用于以下场景:
在每个节点上运行集群守护进程
在每个节点上运行日志收集守护进程
在每个节点上运行监控守护进程
我们创建一个DaemonSet看看效果,我们还是以Nginx为例,下面的例子定义的Daemon将会为每个Node都启动一个Nginx容器:
apiVersion: apps/v1 kind: DaemonSet metadata: name: nginx-daemonset labels: app: nginx spec: selector: matchLabels: name: nginx-daemonset template: metadata: labels: name: nginx-daemonset spec: containers: - name: nginx image: nginx:latest ports: - containerPort: 80
我们创建此DaemonSet看看:
[root@kubevm1 workspace] kubectl create -f demo_daemonset.yml daemonset.apps/nginx-daemonset created [root@kubevm1 workspace] kubectl get DaemonSet NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE nginx-daemonset 2 2 2 2 2 <none> 11m [root@kubevm1 workspace] kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES nginx-daemonset-49tcp 1/1 Running 0 14m 10.244.1.13 kubevm2 <none> <none> nginx-daemonset-w2xc2 1/1 Running 0 14m 10.244.2.11 kubevm3 <none> <none>
与Deployment不同的是,我们在yaml中没有指定replicas,但是DaemonSet自动为除Master节点外每一个Node都创建了一个副本,这里大家可能会问:为什么Master没有Pod副本呢?这个因为缺省状态下,DaemonSet只会在Node中创建副本,如果需要在Master中也启动Pod,则需要设置容忍度:
apiVersion: apps/v1 kind: DaemonSet metadata: name: nginx-daemonset labels: app: nginx spec: selector: matchLabels: name: nginx-daemonset template: metadata: labels: name: nginx-daemonset spec: tolerations: - key: node-role.kubernetes.io/control-plane operator: Exists effect: NoSchedule - key: node-role.kubernetes.io/master operator: Exists effect: NoSchedule containers: - name: nginx image: nginx:latest ports: - containerPort: 80
我们重新创建DaemonSet,看看效果:
[root@kubevm1 workspace] kubectl get pods -o wide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES nginx-daemonset-45z95 1/1 Running 0 46s 10.244.2.13 kubevm3 <none> <none> nginx-daemonset-9z2lt 1/1 Running 0 46s 10.244.0.5 kubevm1 <none> <none> nginx-daemonset-cjf6k 1/1 Running 0 46s 10.244.1.15 kubevm2 <none> <none>
关于容忍度的内容我们后续会详细介绍。
三、Jobs
我们使用Jobs可以定义并启动一个批处理任务,处理完成后,整个批处理任务结束。下面我们创建一个Job,此Job会输出一段文本:
apiVersion: batch/v1 kind: Job metadata: name: echo spec: template: spec: containers: - name: echo image: busybox command: ["sh", "-c", "echo this a job"] restartPolicy: Never
我们创建Job看一下效果:
[root@kubevm1 workspace] kubectl get job NAME COMPLETIONS DURATION AGE echo 1/1 14s 7h3m [root@kubevm1 workspace] kubectl get pods NAME READY STATUS RESTARTS AGE echo-c5kmf 0/1 Completed 0 7h3m [root@kubevm1 workspace] kubectl logs echo-c5kmf this a job
我们设置Job运行完毕后不会重启,所以最终其Pod状态为Complated。
上面的例子只是起了一个Pod比较简单,在实际应用中,我们往往会启动多个Pod并行处理任务,因此Job的使用有以下三种常见模式:
Non-parallel Jobs:一个Job只启动一个Pod,当Pod异常会重启Pod,一旦此Pod正常结束,Job将结束。
Parallel Jobs with a fixed completiong count:并行Job会启动多个Pod,此时需要设定Job的spec.completions为一个正数,当正常结束的Pod数达到此参数设定值后,Job结束。此外,通过设置spec.parallelism可以设置并行度即同时启动几个Job来处理工作项。
Parallel Jobs with a work queue:工作项都放在消息队列中,此模式不需要设置spec.completions参数,而Pod具有以下特性:
每个Pod都能独立判断和决定释放还有任务项需要处理;
如果某个Pod正常结束,则Job不会再启动新的Pod,并且其他Pod应该处于即将结束的状态,毕竟队列里没任务了嘛;
如果所有Pod都结束了,且至少有一个Pod成功结束,则整个Job成功结束。
限于篇幅,关于Job的后两种模式,后面会有专门讲解。
四、CronJob
CronJob其实就是具有定时功能的Job,可以设置Job定时执行或者周期性执行,CronJob使用Cron表达式配置执行周期。我们修改下上面Job定义文件:
apiVersion: batch/v1beta1 kind: CronJob metadata: name: echo spec: schedule: "*/1 * * * *" jobTemplate: spec: template: spec: containers: - name: echo image: busybox command: ["sh", "-c", "echo this a job"] restartPolicy: OnFailure
我们设置每一分钟执行一次,这里也能看出来,在Kubernetes中,CronJob最小调度间隔为分钟级。
[root@kubevm1 workspace] kubectl get cronjob NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE echo */1 * * * * False 0 29s 4m48s [root@kubevm1 workspace] kubectl get pods NAME READY STATUS RESTARTS AGE echo-1655768220-m5p9x 0/1 Completed 0 2m34s echo-1655768280-m7gsj 0/1 Completed 0 93s echo-1655768340-h2n6n 0/1 Completed 0 33s