Kubernetes【K8S】(四)：资源控制器

什么是控制器

​ Kubernetes中内建了很多controller(控制器)，这些相当于一个状态机，用来控制Pod的具体状态和行为。

控制器类型

• ReplicationController和ReplicaSet
• Deployment
• DaemonSet
• StateFulSet
• Job/CronJob
• Horizontal Pod Autoscaling

命令式编程和声明式编程

• 命令式编程：命令“机器”如何去做事情(how)，这样不管你想要的是什么(what)，它都会按照你的命令实现。
• 声明式编程：告诉“机器”你想要的是什么(what)，让机器想出如何去做(how)

声明式推荐用apply创建，命令式推荐用create创建。Deployment属于声明式，RelicaSet属于命令式。替换创建不是不可以，只是推荐属于最优方式。

Pod分类

• 自主式Pod： Pod退出、删除，不会被自动创建新副本。
• 控制器管理的Pod： 在控制器的生命周期里，始终要维持Pod的副本数目。

ReplicationController和ReplicaSet

​ ReplicationController(RC)用来确保容器应用的副本数始终保持在用户定义的副本数，即如果有容器异常退出，会自动创建新的Pod来替代；而如果异常多出来的容器也会自动回收。

​ 在新版本的Kubernetes中建议使用ReplicaSet来取代ReplicationController。ReplicaSet和ReplicationController没有本质区别，ReplicaSet新增支持集合式的selector。

编写rs.yaml

apiVersion: extensions/v1beta1
kind: ReplicaSet
metadata:
  name: frontend
spec:
  replicas: 3
  selector:
    matchLabels:
      tier: frontend
  template:
    metadata:
      name: frontend
      labels:
        tier: frontend
    spec:
      containers:
        - name: myapp
          image: chinda.com/library/myapp:v1
          env:
            - name: GET_HOSTS_FROM
              value: dns
          ports:
            - containerPort: 80

创建RS

[root@k8s-master01 ~]# kubectl create -f rs.yaml 
[root@k8s-master01 ~]# kubectl get pod
NAME             READY   STATUS    RESTARTS   AGE
frontend-gzjh7   1/1     Running   0          5m58s
frontend-jnwtp   1/1     Running   0          5m58s
frontend-ktmdc   1/1     Running   0          5m58s
# 可以显示labels
[root@k8s-master01 ~]# kubectl get pod --show-labels
NAME             READY   STATUS    RESTARTS   AGE     LABELS
frontend-gzjh7   1/1     Running   0          3m45s   tier=frontend
frontend-jnwtp   1/1     Running   0          3m45s   tier=frontend
frontend-ktmdc   1/1     Running   0          3m45s   tier=frontend
# 修改labels
[root@k8s-master01 ~]# kubectl label pod frontend-gzjh7 tier=frontend1
error: 'tier' already has a value (frontend), and --overwrite is false
# 因为--overwrite 是 false
[root@k8s-master01 ~]# kubectl get pod --show-labels
NAME             READY   STATUS    RESTARTS   AGE     LABELS
frontend-gtpc7   1/1     Running   0          21s     tier=frontend
frontend-gzjh7   1/1     Running   0          9m13s   tier=frontend1
frontend-jnwtp   1/1     Running   0          9m13s   tier=frontend
frontend-ktmdc   1/1     Running   0          9m13s   tier=frontend
# 删除rs看下，修改过label的没有随着rs删除跟着销毁，是因为现在这个pod已经不属于rs管理范围，变成自主式的pod
[root@k8s-master01 ~]# kubectl delete rs --all
replicaset.extensions "frontend" deleted
[root@k8s-master01 ~]# kubectl get pod
NAME             READY   STATUS    RESTARTS   AGE
frontend-gzjh7   1/1     Running   0          11m

RS与Deployment的关联

Deployment

​ Deployement为Pod和ReplicaSet提供了一个声明式定义(Declarative)方法，用来替代以前的ReplicationController来方便管理应用。

​ 典型的应用场景：

• 定义Deployment来创建Pod和ReplicaSet
• 滚动升级和回滚应用
• 扩容和缩容
• 暂停和继续Deployment

编写deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: myapp
  labels:
    app: myapp
spec:
  replicas: 3
  template:
    metadata:
      name: myapp
      labels:
        app: myapp
    spec:
      containers:
        - name: myapp
          image: chinda.com/library/myapp:v1
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 80
      restartPolicy: Always
  selector:
    matchLabels:
      app: myapp

创建Deployment

# --record参数可以记录命令，可以很方便的查看每次revision的变化
[root@k8s-master01 ~]# kubectl apply -f deployment.yaml --record
deployment.apps/myapp created
# 查看pods
[root@k8s-master01 ~]# kubectl get pods -o wide
NAME                    READY   STATUS    RESTARTS   AGE     IP            NODE         NOMINATED NODE   READINESS GATES
myapp-cd7f9754d-5qbgh   1/1     Running   0          4m37s   10.244.1.29   k8s-node01   <none>           <none>
myapp-cd7f9754d-7pr52   1/1     Running   0          4m37s   10.244.2.21   k8s-node02   <none>           <none>
myapp-cd7f9754d-ccrbq   1/1     Running   0          4m37s   10.244.1.30   k8s-node01   <none>           <none>
# 查看rs
[root@k8s-master01 ~]# kubectl get rs
NAME              DESIRED   CURRENT   READY   AGE
myapp-cd7f9754d   3         3         3       5m12s
# 查看deployment
[root@k8s-master01 ~]# kubectl get deployment
NAME    READY   UP-TO-DATE   AVAILABLE   AGE
myapp   3/3     3            3           5m43s

扩容

# 扩容
[root@k8s-master01 ~]# kubectl scale deployment myapp --replicas=10
deployment.extensions/myapp scaled
[root@k8s-master01 ~]# kubectl get pods
NAME                    READY   STATUS    RESTARTS   AGE
myapp-cd7f9754d-2gdvk   1/1     Running   0          60s
myapp-cd7f9754d-5qbgh   1/1     Running   0          9m8s
myapp-cd7f9754d-66lqp   1/1     Running   0          60s
myapp-cd7f9754d-7nnfj   1/1     Running   0          60s
myapp-cd7f9754d-7pr52   1/1     Running   0          9m8s
myapp-cd7f9754d-7vt7p   1/1     Running   0          60s
myapp-cd7f9754d-8jsvj   1/1     Running   0          60s
myapp-cd7f9754d-ccrbq   1/1     Running   0          9m8s
myapp-cd7f9754d-hm89p   1/1     Running   0          60s
myapp-cd7f9754d-qtx2d   1/1     Running   0          60s
# 注意：扩容前后rs的名称没有变动
[root@k8s-master01 ~]# kubectl get rs
NAME              DESIRED   CURRENT   READY   AGE
myapp-cd7f9754d   10        10        10      9m55s

更新镜像

# 语法kubectl set image deployment/[deploymentName] [containerName]=[iamge]
[root@k8s-master01 ~]# kubectl set image deployment/myapp myapp=chinda.com/library/myapp:v2
deployment.extensions/myapp image updated
# 触发rs创建
[root@k8s-master01 ~]# kubectl get rs
NAME               DESIRED   CURRENT   READY   AGE
myapp-69f8478dfc   10        10        5       6s
myapp-cd7f9754d    3         3         3       13m
# 也可以使用edit命令来编辑Deployment
[root@k8s-master01 ~]# kubectl edit deployment/myapp

回滚

[root@k8s-master01 ~]# kubectl rollout undo deployment/myapp
deployment.extensions/myapp rolled back

查看rollout状态、历史、回滚指定revision

# 状态
[root@k8s-master01 ~]# kubectl rollout status deployment/myapp
deployment "myapp" successfully rolled out
# 历史
[root@k8s-master01 ~]# kubectl rollout history deployment/myapp
deployment.extensions/myapp 
REVISION  CHANGE-CAUSE
5         kubectl apply --filename=deployment.yaml --record=true
6         kubectl apply --filename=deployment.yaml --record=true
7         kubectl apply --filename=deployment.yaml --record=true
# 回滚指定revision
[root@k8s-master01 ~]# kubectl rollout undo deployment/myapp --to-revision=5
deployment.extensions/myapp rolled back
# rollout成功完成，将会返回一个0值得Exit Code
[root@k8s-master01 ~]# kubectl rollout status deployment/myapp
deployment "myapp" successfully rolled out
[root@k8s-master01 ~]# echo $?
0
# 暂停
[root@k8s-master01 ~]# kubectl rollout pause deployment/myapp

查看RS历史

[root@k8s-master01 ~]# kubectl get rs
NAME               DESIRED   CURRENT   READY   AGE
myapp-69f8478dfc   10        10        10      18m
myapp-6bf96697b6   0         0         0       4m47s
myapp-cd7f9754d    0         0         0       21m

Deployment更新策略

Deployment可以保证在升级时只有一定数量的Pod是down的。期望副本数的25%

Deployment同时可以确保只创建出超过期望数量的一定数量的Pod。期望副本数的25%

Rellover(多个rollout并行)

假如创建了一个5个myapp:v1副本的Deployment，当只有3个myapp:v1副本创建出来时，就开始更新含有5个副本的myapp:v2的Deployment。在这种情况下，Deployment会立即杀掉已经创建的3个myapp:v1的Pod，并开始创建myapp:v2的Pod。它不会等到所有5个myapp:v1的Pod都创建完成后才开始改变航道。

DaemonSet控制器

DaemonSet确保全部(或者一些)Node上运行一个Pod的副本。当有Node假如集群时也会为他们新增一个Pod。当有Node从集群移除时，这些Pod也会被回收。删除DaemonSet将会删除它创建的所有Pod。

DaemonSet的典型用法

• 运行集群存储daemon，例如再每个Node上运行glusterd、ceph。
• 在每个Node上运行日志收集daemon，例如fluentd、logstash。
• 在每个Node上运行监控daemon，例如Prometheus Node Exporter、collectd、Datadog代理、New Relic代理、Ganglia gmond。

编写damon.yaml

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: daemonset
  labels:
    app: daemonset
spec:
  template:
    metadata:
      name: daemonset
      labels:
        app: daemonset
    spec:
      containers:
        - name: daemonset
          image: chinda.com/library/myapp:v1
          imagePullPolicy: IfNotPresent
      restartPolicy: Always
  selector:
    matchLabels:
      app: daemonset

创建Ds

[root@k8s-master01 ~]# kubectl create -f daemonset.yaml 
daemonset.apps/daemonset created
[root@k8s-master01 ~]# kubectl get ds
NAME        DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
daemonset   2         2         2       2            2           <none>          11s
[root@k8s-master01 ~]# kubectl get pod -o wide
NAME              READY   STATUS    RESTARTS   AGE   IP             NODE         NOMINATED NODE   READINESS GATES
daemonset-48jrg   1/1     Running   0          63s   10.244.1.101   k8s-node01   <none>           <none>
daemonset-k6mvj   1/1     Running   0          63s   10.244.2.92    k8s-node02   <none>           <none>
[root@k8s-master01 ~]# kubectl  delete pod daemonset-48jrg
pod "daemonset-48jrg" deleted
# 注意: 新创建的Pod和删除的是不同的
[root@k8s-master01 ~]# kubectl get pod
NAME              READY   STATUS    RESTARTS   AGE
daemonset-k6mvj   1/1     Running   0          3m22s
daemonset-q4chg   1/1     Running   0          7s

Job控制器

Job负载批处理任务，即仅执行一次的任务，它保证批处理任务的一个活多个Pod成功结束。

特殊说明

• spec.template格式同Pod。
• RestartPolicy仅支持Never或Onfailure。
• 单个Pod时，默认Pod成功运行后Job即结束。
• .spec.comletions标志Job结束需要成功运行Pod个数，默认为1。
• .spec.parallelism标志并行运行的Pod个数，默认为1。
• spec.activeDeadlineSeconds标志失败Pod的重试最大时间，超过这个时间不会继续重试。

编写job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: pi
spec:
  template:
    metadata:
      name: pi
    spec:
      containers:
        - name: pi
          image: chinda.com/library/perl:v1
          command: ['perl', '-Mbignum=bpi', '-wle', 'print bpi(2000)']
      restartPolicy: Never

创建Job

[root@k8s-master01 ~]# kubectl create -f job.yaml 
job.batch/pi created
[root@k8s-master01 ~]# kubectl get pod
NAME       READY   STATUS    RESTARTS   AGE
pi-smqkb   1/1     Running   0          6s
# 查看日志
[root@k8s-master01 ~]# kubectl log pi-smqkb
log is DEPRECATED and will be removed in a future version. Use logs instead.
3.1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679821480865132823066470938446095505822317253594081284811174502841027019385211055596446229489549303819644288109756659334461284756482337867831652712019091456485669234603486104543266482133936072602491412737245870066063155881748815209209628292540917153643678925903600113305305488204665213841469519415116094330572703657595919530921861173819326117931051185480744623799627495673518857527248912279381830119491298336733624406566430860213949463952247371907021798609437027705392171762931767523846748184676694051320005681271452635608277857713427577896091736371787214684409012249534301465495853710507922796892589235420199561121290219608640344181598136297747713099605187072113499999983729780499510597317328160963185950244594553469083026425223082533446850352619311881710100031378387528865875332083814206171776691473035982534904287554687311595628638823537875937519577818577805321712268066130019278766111959092164201989380952572010654858632788659361533818279682303019520353018529689957736225994138912497217752834791315155748572424541506959508295331168617278558890750983817546374649393192550604009277016711390098488240128583616035637076601047101819429555961989467678374494482553797747268471040475346462080466842590694912933136770289891521047521620569660240580381501935112533824300355876402474964732639141992726042699227967823547816360093417216412199245863150302861829745557067498385054945885869269956909272107975093029553211653449872027559602364806654991198818347977535663698074265425278625518184175746728909777727938000816470600161452491921732172147723501414419735685481613611573525521334757418494684385233239073941433345477624168625189835694855620992192221842725502542568876717904946016534668049886272327917860857843838279679766814541009538837863609506800642251252051173929848960841284886269456042419652850222106611863067442786220391949450471237137869609563643719172874677646575739624138908658326459958133904780275901

CronJob控制器

CronJob是基于时间的Job，即1. 在给定的时间点只运行一次；2. 周期性的在给定时间点运行。

典型用法

• 在给定的时间点调度Job运行。
• 创建周期性运行的Job，例如：数据库备份、发送邮件等。

字段解释

• spec.shuedule： 调度，必须字段，指定任务运行周期，格式同Cron。

• spec.jobTemplate： Job模板，必须字段，指定需要运行的任务，格式同Job。

• spec.startingDeadlineSeconds： 启动Job的期限(秒级别), 超过此时间Job认为是失败的，该字段是可选的。如果因为任何原因而错过了被调度的时间，那么错过执行时间的Job将被认为是失败的。如果没有指定，则没有期限。

• spec.concurrencyPolicy： 并发策略，该字段可选。它指定了如何处理被CronJob创建Job的并发执行。只允许指定一种策略。

  • Allow(默认)：允许并发运行Job。
  • Forbid：禁止并发运行，如果前一个还没有完成，则直接跳过下一个。
  • Replace：取消当前正在运行的Job，用一个新的来替换。

  注意：当前策略只能应用于同一个CronJob创建的Job。如果存在多个CronJob，他们创建的Job之间总是允许并发运行的。

• spec.suspend：挂起，该字段可选。如果设置为true，后续所有执行都会被挂起。它对已经开始执行的Job不起作用。默认值为false。

• spec.successfulJobsHistoryLimit和spec.failedJobsHistoryLimit：历史限制，该字段可选。它们指定了可以保留多少完成和失败的Job。默认情况下，它们分别设置为3和1。设置限制的值为0，相关类型的Job完成后将不会保留。

编写cronjob.yaml

apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: hello
spec:
  jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: hello
              image: busybox
              args:
                - /bin/sh
                - -c
                - date; echo Hello from the Kubernetes cluster
          restartPolicy: OnFailure
  schedule: '*/1 * * * *'

创建CronJob

[root@k8s-master01 ~]# kubectl apply -f cronjob.yaml 
cronjob.batch/hello created
[root@k8s-master01 ~]# kubectl get job -w
NAME               COMPLETIONS   DURATION   AGE
hello-1588114140   1/1           6s         2m33s
hello-1588114200   1/1           7s         93s
hello-1588114260   1/1           4s         33s
[root@k8s-master01 ~]# kubectl get pod -w
NAME                     READY   STATUS      RESTARTS   AGE
hello-1588114140-6l2g2   0/1     Completed   0          2m47s
hello-1588114200-snrtc   0/1     Completed   0          107s
hello-1588114260-zv8dq   0/1     Completed   0          47s
[root@k8s-master01 ~]# kubectl get cronjob
NAME    SCHEDULE      SUSPEND   ACTIVE   LAST SCHEDULE   AGE
hello   */1 * * * *   False     0        69s             4m9s
[root@k8s-master01 ~]# kubectl log hello-1588114140-6l2g2
log is DEPRECATED and will be removed in a future version. Use logs instead.
Tue Apr 28 22:53:12 UTC 2020
Hello from the Kubernetes cluster