kubernetes—Controller详解（一）-阿里云开发者社区

Pod控制器介绍

Pod是kubernetes的最小管理单元，在kubernetes中，按照pod的创建方式可以将其分为两类：

自主式pod：kubernetes直接创建出来的Pod，这种pod删除后就没有了，也不会重建
控制器创建的pod：kubernetes通过控制器创建的pod，这种pod删除了之后还会自动重建

什么是Pod控制器

Pod控制器是管理pod的中间层，使用Pod控制器之后，只需要告诉Pod控制器，想要多少个什么样的Pod就可以了，它会创建出满足条件的Pod并确保每一个Pod资源处于用户期望的目标状态。如果Pod资源在运行中出现故障，它会基于指定策略重新编排Pod。

在kubernetes中，有很多类型的pod控制器，每种都有自己的适合的场景，常见的有下面这些：

ReplicationController：比较原始的pod控制器，已经被废弃，由ReplicaSet替代
ReplicaSet：保证副本数量一直维持在期望值，并支持pod数量扩缩容，镜像版本升级
Deployment：通过控制ReplicaSet来控制Pod，并支持滚动升级、回退版本
Horizontal Pod Autoscaler：可以根据集群负载自动水平调整Pod的数量，实现削峰填谷
DaemonSet：在集群中的指定Node上运行且仅运行一个副本，一般用于守护进程类的任务
Job：它创建出来的pod只要完成任务就立即退出，不需要重启或重建，用于执行一次性任务
Cronjob：它创建的Pod负责周期性任务控制，不需要持续后台运行
StatefulSet：管理有状态应用

ReplicaSet(RS)

RS概述

ReplicaSet的主要作用是保证一定数量的pod正常运行，它会持续监听这些Pod的运行状态，一旦Pod发生故障，就会重启或重建。同时它还支持对pod数量的扩缩容和镜像版本的升降级。

ReplicaSet的资源清单文件：

apiVersion: apps/v1 # 版本号
kind: ReplicaSet # 类型
metadata: # 元数据
  name: # rs名称
  namespace: # 所属命名空间
  labels: #标签
    controller: rs
spec: # 详情描述
  replicas: 3 # 副本数量
  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: nginx-pod
    matchExpressions: # Expressions匹配规则
      - { key: app, operator: In, values: [ nginx-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: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx
          ports:
            - containerPort: 80

在这里面，需要新了解的配置项就是spec下面几个选项：

replicas：指定副本数量，其实就是当前rs创建出来的pod的数量，默认为1
selector：选择器，它的作用是建立pod控制器和pod之间的关联关系，采用的Label Selector机制在pod模板上定义label，在控制器上定义选择器，就可以表明当前控制器能管理哪些pod了

template：模板，就是当前控制器创建pod所使用的模板板，里面其实就是专栏pod详解学的定义

创建与删除RS

创建rs-controller.yaml文件，内容如下：

apiVersion: apps/v1
kind: ReplicaSet
metadata:
  name: rs-controller
  namespace: dev
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx-pod
  template:
    metadata:
      labels:
        app: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx

创建RS

# 创建rs
[root@master k8sYamlForCSDN]# kubectl apply -f rs-controller.yaml 
replicaset.apps/rs-controller created
# 查看rs
# DESIRED:期望副本数量  
# CURRENT:当前副本数量  
# READY:已经准备好提供服务的副本数量
[root@master k8sYamlForCSDN]# kubectl get rs rs-controller -n dev -o wide
NAME            DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES   SELECTOR
rs-controller   3         3         3       75s   nginx        nginx    app=nginx-pod
# 查看当前控制器创建出来的pod
# 这里发现控制器创建出来的pod的名称是在控制器名称后面拼接了-xxxxx随机码
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                  READY   STATUS    RESTARTS   AGE
rs-controller-rr598   1/1     Running   0          47s
rs-controller-tmfd7   1/1     Running   0          47s
rs-controller-xgrst   1/1     Running   0          47s

删除RS

# 只推荐第一种，第二三种做了解即可
# 1. 使用yaml直接删除(推荐)
[root@master k8sYamlForCSDN]# kubectl delete -f rs-controller.yaml 
replicaset.apps "rs-controller" deleted
# 2. 使用kubectl delete命令会删除此RS以及它管理的Pod
# 在kubernetes删除RS前，会将RS的replicasclear调整为0，等待所有的Pod被删除后，在执行RS对象的删除
[root@master k8sYamlForCSDN]# kubectl delete rs rs-controller -n dev
replicaset.apps "rs-controller" deleted
[root@master k8sYamlForCSDN]# kubectl get pod -n dev
No resources found in dev namespace.
# 3. 如果希望只删除RS对象，而留下正在运行的pod，需要在kubectl delete命令的时候增加--cascade=orphan选项
[root@master k8sYamlForCSDN]# kubectl get rs -n dev
NAME            DESIRED   CURRENT   READY   AGE
rs-controller   6         6         3       15s
[root@master k8sYamlForCSDN]# kubectl delete rs rs-controller -n dev --cascade=orphan
replicaset.apps "rs-controller" deleted
[root@master k8sYamlForCSDN]# kubectl get rs -n dev
No resources found in dev namespace.
[root@master k8sYamlForCSDN]# kubectl get pod -n dev
NAME                  READY   STATUS              RESTARTS   AGE
rs-controller-9h5zd   1/1     Running             0          38s
rs-controller-jk4h6   1/1     Running             0          38s
rs-controller-lmlwn   1/1     Running             0          38s

扩缩容

# 第一种方式
# 修改yaml文件，把replicas=3改成1，重新apply，此时rs会更新
[root@master k8sYamlForCSDN]# kubectl apply -f rs-controller.yaml 
replicaset.apps/rs-controller created
[root@master k8sYamlForCSDN]# vi rs-controller.yaml 
[root@master k8sYamlForCSDN]# kubectl apply -f rs-controller.yaml 
replicaset.apps/rs-controller configured
# 命令运行完毕，立即查看，发现已经有2个开始准备退出了
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                  READY   STATUS        RESTARTS   AGE
rs-controller-h654s   0/1     Terminating   0          18s
rs-controller-l67tb   0/1     Terminating   0          18s
rs-controller-s4g7b   1/1     Running       0          18s
# 稍等片刻，就只剩下1个了
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                  READY   STATUS    RESTARTS   AGE
rs-controller-s4g7b   1/1     Running   0          81s
# 第二种方式
# 编辑rs的副本数量，修改spec:replicas: 2即可，注意用这种方法，yaml文件中的数量是不改变的
[root@master k8sYamlForCSDN]# kubectl edit rs rs-controller -n dev
replicaset.apps/rs-controller edited
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                  READY   STATUS    RESTARTS   AGE
rs-controller-9mf77   1/1     Running   0          26s
rs-controller-s4g7b   1/1     Running   0          3m23s
# 第三种方式
# 使用scale命令实现，后面--replicas=n直接指定目标数量即可，注意用这种方法，yaml文件中的数量是不改变的
[root@master k8sYamlForCSDN]# kubectl scale rs rs-controller -n dev --replicas=3 
replicaset.apps/rs-controller scaled
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                  READY   STATUS              RESTARTS   AGE
rs-controller-7d9wf   0/1     ContainerCreating   0          6s
rs-controller-9mf77   1/1     Running             0          119s
rs-controller-s4g7b   1/1     Running             0          4m56s

镜像升级

# 第一种方式(推荐)
# 修改yaml文件，把image=nginx改成nginx:1.17.1，重新apply，此时rs会更新
[root@master k8sYamlForCSDN]# vi rs-controller.yaml 
[root@master k8sYamlForCSDN]# kubectl apply -f rs-controller.yaml 
replicaset.apps/rs-controller configured
[root@master k8sYamlForCSDN]# kubectl get rs -n dev -o wide
NAME            DESIRED   CURRENT   READY   AGE     CONTAINERS   IMAGES         
rs-controller   1         1         1       6m42s   nginx        nginx:1.17.1   
# 第二种方式
# 编辑rs的image，修改imagenginx:1.17.2即可，注意用这种方法，yaml文件中的配置是不改变的
[root@master k8sYamlForCSDN]# kubectl edit rs rs-controller -n dev
replicaset.apps/rs-controller edited
[root@master k8sYamlForCSDN]# kubectl get rs -n dev -o wide
NAME            DESIRED   CURRENT   READY   AGE     CONTAINERS   IMAGES         
rs-controller   1         1         1       8m41s   nginx        nginx:1.17.2  
# 第三种方式，也可以使用命令完成这个工作
# kubectl set image rs rs名称 容器=镜像版本 -n namespace
[root@master k8sYamlForCSDN]# kubectl set image rs rs-controller -n dev nginx=nginx:1.17.3
replicaset.apps/rs-controller image updated
[root@master k8sYamlForCSDN]# kubectl get rs -n dev -o wide
NAME            DESIRED   CURRENT   READY   AGE     CONTAINERS   IMAGES        
rs-controller   1         1         1       9m50s   nginx        nginx:1.17.3

Deployment(Deploy)

Deploy概述

为了更好的解决服务编排的问题，kubernetes在V1.2版本开始，引入了Deployment控制器。这种控制器并不直接管理pod，而是通过管理ReplicaSet来间接管理Pod，即：Deployment管理ReplicaSet，ReplicaSet管理Pod。所以Deployment比ReplicaSet功能更加强大。

Deployment主要功能有下面几个：

支持ReplicaSet的所有功能

支持发布的停止、继续
支持滚动升级和回滚版本

Deployment的资源清单文件：

apiVersion: apps/v1 # 版本号
kind: Deployment # 类型       
metadata: # 元数据
  name: # Deploy名称 
  namespace: # 所属命名空间 
  labels: #标签
    controller: deploy
spec: # 详情描述
  replicas: 3 # 副本数量
  revisionHistoryLimit: 3 # 保留历史版本（rs，用于版本回退）
  paused: false # 暂停部署，默认是false
  progressDeadlineSeconds: 600 # 部署超时时间（s），默认是600
  strategy: # 策略
    type: RollingUpdate # 滚动更新策略
    rollingUpdate: # 滚动更新
      maxSurge: 30% # 最大额外可以存在的副本数，可以为百分比，也可以为整数
      maxUnavailable: 30% # 最大不可用状态的 Pod 的最大值，可以为百分比，也可以为整数
  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: nginx-pod
    matchExpressions: # Expressions匹配规则
      - { key: app, operator: In, values: [ nginx-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: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx
          ports:
            - containerPort: 80

创建deployment

创建deploy-controller.yaml，内容如下：

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deploy-controller
  namespace: dev
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx-pod
  template:
    metadata:
      labels:
        app: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx

# 创建deployment
[root@master k8sYamlForCSDN]# kubectl apply -f deploy-controller.yaml 
deployment.apps/deploy-controller created
# 查看deployment
# UP-TO-DATE 最新版本的pod的数量
# AVAILABLE  当前可用的pod的数量
[root@master k8sYamlForCSDN]# kubectl get deploy deploy-controller -n dev
NAME                READY   UP-TO-DATE   AVAILABLE   AGE
deploy-controller   3/3     3            3           40s
[root@master k8sYamlForCSDN]# kubectl get deployments.apps -n dev
NAME                READY   UP-TO-DATE   AVAILABLE   AGE
deploy-controller   3/3     3            3           47s
# 查看rs
# 发现rs的名称是在原来deployment的名字后面添加了一个10位数的随机串
[root@master k8sYamlForCSDN]# kubectl get rs -n dev
NAME                           DESIRED   CURRENT   READY   AGE
deploy-controller-7f87645cd8   3         3         3       65s
# 查看pod
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                                 READY   STATUS    RESTARTS   AGE
deploy-controller-7f87645cd8-28vcs   1/1     Running   0          73s
deploy-controller-7f87645cd8-dgnsb   1/1     Running   0          73s
deploy-controller-7f87645cd8-qfczb   1/1     Running   0          73s

扩缩容

# 第一种方式
# 修改yaml文件，把replicas=3改成2，重新apply，此时rs会更新
[root@master k8sYamlForCSDN]# vi deploy-controller.yaml  
[root@master k8sYamlForCSDN]# kubectl apply -f deploy-controller.yaml 
deployment.apps/deploy-controller configured
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                                 READY   STATUS    RESTARTS   AGE
deploy-controller-7f87645cd8-dgnsb   1/1     Running   0          5m53s
deploy-controller-7f87645cd8-qfczb   1/1     Running   0          5m53s
# 第二种方式
# 编辑rs的副本数量，修改spec:replicas: 1即可，注意用这种方法，yaml文件中的数量是不改变的
[root@master k8sYamlForCSDN]# kubectl edit deploy deploy-controller -n dev
deployment.apps/deploy-controller edited
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                                 READY   STATUS    RESTARTS   AGE
deploy-controller-7f87645cd8-qfczb   1/1     Running   0          7m56s
# 第三种方式
# 使用scale命令实现，后面--replicas=n直接指定目标数量即可，注意用这种方法，yaml文件中的数量是不改变的
[root@master k8sYamlForCSDN]# kubectl scale deploy deploy-controller -n dev --replicas=3
deployment.apps/deploy-controller scaled
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                                 READY   STATUS              RESTARTS   AGE
deploy-controller-7f87645cd8-2dwpp   0/1     ContainerCreating   0          1s
deploy-controller-7f87645cd8-bb9xt   0/1     ContainerCreating   0          1s
deploy-controller-7f87645cd8-qfczb   1/1     Running             0          8m45s

镜像更新

deployment支持两种更新策略:重建更新和滚动更新,可以通过strategy指定策略类型,支持两个属性:

strategy：指定新的Pod替换旧的Pod的策略， 支持两个属性：
  type：指定策略类型，支持两种策略
    Recreate：在创建出新的Pod之前会先杀掉所有已存在的Pod
    RollingUpdate：滚动更新，就是杀死一部分，就启动一部分，在更新过程中，存在两个版本Pod
  rollingUpdate：当type为RollingUpdate时生效，用于为RollingUpdate设置参数，支持两个属性：
    maxUnavailable：用来指定在升级过程中不可用Pod的最大数量，默认为25%。
    maxSurge： 用来指定在升级过程中可以超过期望的Pod的最大数量，默认为25%。

重建更新:在创建出新的Pod之前会先杀掉所有已存在的Pod

编辑deploy-controller.yaml,在spec节点下添加更新策略

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deploy-controller
  namespace: dev
spec:
  strategy: # 策略
    type: Recreate # 重建更新
  replicas: 3
  selector:
    matchLabels:
      app: nginx-pod
  template:
    metadata:
      labels:
        app: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx

创建deploy进行验证

# 变更镜像
[root@master k8sYamlForCSDN]# kubectl get pods -n dev -o wide
NAME                                 READY   STATUS    RESTARTS   AGE   IP            NODE    NOMINATED NODE   READINESS GATES
deploy-controller-7f87645cd8-925wr   1/1     Running   0          35s   10.244.2.65   node2   <none>           <none>
deploy-controller-7f87645cd8-l4f82   1/1     Running   0          35s   10.244.2.64   node2   <none>           <none>
deploy-controller-7f87645cd8-ljwfx   1/1     Running   0          35s   10.244.1.44   node1   <none>           <none>
[root@master k8sYamlForCSDN]# kubectl set image deploy deploy-controller -n dev nginx=nginx:1.17.1
deployment.apps/deploy-controller image updated
# 观察升级过程
[root@master k8sYamlForCSDN]# kubectl get pods -n dev -o wide -w
NAME                                 READY   STATUS    RESTARTS   AGE   IP            NODE    NOMINATED NODE   READINESS GATES
deploy-controller-7f87645cd8-925wr   1/1     Running   0          39s   10.244.2.65   node2   <none>           <none>
deploy-controller-7f87645cd8-l4f82   1/1     Running   0          39s   10.244.2.64   node2   <none>           <none>
deploy-controller-7f87645cd8-ljwfx   1/1     Running   0          39s   10.244.1.44   node1   <none>           <none>
deploy-controller-7f87645cd8-925wr   1/1     Terminating   0          49s   10.244.2.65   node2   <none>           <none>
deploy-controller-7f87645cd8-l4f82   1/1     Terminating   0          49s   10.244.2.64   node2   <none>           <none>
deploy-controller-7f87645cd8-ljwfx   1/1     Terminating   0          49s   10.244.1.44   node1   <none>           <none>
deploy-controller-6f44b7b85d-whjxq   0/1     Pending       0          0s    <none>        <none>   <none>           <none>
deploy-controller-6f44b7b85d-4nvzz   0/1     Pending       0          0s    <none>        <none>   <none>           <none>
deploy-controller-6f44b7b85d-hjk7x   0/1     Pending       0          0s    <none>        <none>   <none>           <none>
deploy-controller-6f44b7b85d-whjxq   0/1     ContainerCreating   0          0s    <none>        node1    <none>           <none>
deploy-controller-6f44b7b85d-hjk7x   0/1     ContainerCreating   0          0s    <none>        node1    <none>           <none>
deploy-controller-6f44b7b85d-4nvzz   0/1     ContainerCreating   0          0s    <none>        node2    <none>           <none>
deploy-controller-6f44b7b85d-4nvzz   1/1     Running             0          2s    10.244.2.66   node2    <none>           <none>
deploy-controller-6f44b7b85d-hjk7x   1/1     Running             0          16s   10.244.1.46   node1    <none>           <none>
deploy-controller-6f44b7b85d-whjxq   1/1     Running             0          31s   10.244.1.45   node1    <none>           <none>

滚动更新:滚动更新，就是杀死一部分，就启动一部分，在更新过程中，存在两个版本Pod

编辑deploy-controller.yaml,在spec节点下添加更新策略

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deploy-controller
  namespace: dev
spec:
  strategy: # 策略
    type: RollingUpdate # 滚动更新
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
  replicas: 3
  selector:
    matchLabels:
      app: nginx-pod
  template:
    metadata:
      labels:
        app: nginx-pod
    spec:
      containers:
        - name: nginx
          image: nginx

创建deploy进行验证

# 创建deploy
[root@master k8sYamlForCSDN]# kubectl apply -f deploy-controller.yaml 
deployment.apps/deploy-controller configured
[root@master k8sYamlForCSDN]# kubectl get pods -n dev
NAME                                 READY   STATUS    RESTARTS   AGE
deploy-controller-7f87645cd8-4b2rf   1/1     Running   0          74s
deploy-controller-7f87645cd8-7tqnr   1/1     Running   0          57s
deploy-controller-7f87645cd8-cgmft   1/1     Running   0          55s
# 更新镜像
[root@master k8sYamlForCSDN]# kubectl set image deploy deploy-controller -n dev nginx=nginx:1.17.1
deployment.apps/deploy-controller image updated
# 滚动更新，不是先全杀掉在启动，而是一个杀掉，启动一个，滚动的形式
[root@master k8sYamlForCSDN]# kubectl get pods -n dev -o wide -w
NAME                                 READY   STATUS    RESTARTS   AGE   IP            NODE    NOMINATED NODE   READINESS GATES
deploy-controller-7f87645cd8-4b2rf   1/1     Running   0          86s   10.244.2.67   node2   <none>           <none>
deploy-controller-7f87645cd8-7tqnr   1/1     Running   0          69s   10.244.1.47   node1   <none>           <none>
deploy-controller-7f87645cd8-cgmft   1/1     Running   0          67s   10.244.1.48   node1   <none>           <none>
deploy-controller-6f44b7b85d-xckc9   0/1     Pending   0          0s    <none>        <none>   <none>           <none>
deploy-controller-6f44b7b85d-xckc9   0/1     ContainerCreating   0          0s    <none>        node2    <none>           <none>
deploy-controller-6f44b7b85d-xckc9   1/1     Running             0          16s   10.244.2.68   node2    <none>           <none>
deploy-controller-7f87645cd8-7tqnr   1/1     Terminating         0          92s   10.244.1.47   node1    <none>           <none>
deploy-controller-6f44b7b85d-mxm5q   0/1     Pending             0          0s    <none>        node1    <none>           <none>
deploy-controller-6f44b7b85d-mxm5q   0/1     ContainerCreating   0          0s    <none>        node1    <none>           <none>
deploy-controller-6f44b7b85d-mxm5q   1/1     Running             0          17s   10.244.1.49   node1    <none>           <none>
deploy-controller-7f87645cd8-7tqnr   0/1     Terminating         0          93s   10.244.1.47   node1    <none>           <none>
deploy-controller-7f87645cd8-cgmft   1/1     Terminating         0          107s   10.244.1.48   node1    <none>           <none>
deploy-controller-6f44b7b85d-ftztf   0/1     Pending             0          0s     <none>        <none>   <none>           <none>
deploy-controller-6f44b7b85d-ftztf   0/1     ContainerCreating   0          0s     <none>        node2    <none>           <none>
deploy-controller-7f87645cd8-cgmft   0/1     Terminating         0          108s   10.244.1.48   node1    <none>           <none>
deploy-controller-6f44b7b85d-ftztf   1/1     Running             0          17s    10.244.2.69   node2    <none>           <none>
deploy-controller-7f87645cd8-4b2rf   1/1     Terminating         0          2m23s   10.244.2.67   node2    <none>           <none>
# 至此，新版本的pod创建完毕，旧版本的pod销毁完毕
# 中间过程是滚动进行的，也就是边销毁边创建

滚动更新的过程：

版本回退

镜像更新中rs的变化:

# 创建deploy，并添加一个参数记录版本
[root@master k8sYamlForCSDN]# kubectl apply -f deploy-controller.yaml --record
Flag --record has been deprecated, --record will be removed in the future
deployment.apps/deploy-controller created
# 更新镜像
[root@master k8sYamlForCSDN]# kubectl set image deploy deploy-controller -n dev nginx=nginx:1.17.1
deployment.apps/deploy-controller image updated
# 查看RS
# 发现原来的RS中的pod被删除了，又重新创建了新的RS
[root@master k8sYamlForCSDN]# kubectl get rs -n dev -w
NAME                           DESIRED   CURRENT   READY   AGE
deploy-controller-7f87645cd8   3         3         1       11s
...
deploy-controller-6f44b7b85d   3         3         3       35s
...
deploy-controller-7f87645cd8   0         0         0       90s

eployment支持版本升级过程中的暂停、继续功能以及版本回退等诸多功能，下面具体来看

kubectl rollout：版本升级相关功能，支持下面的选项：

status 显示当前升级状态
history 显示升级历史记录
pause 暂停版本升级过程
resume 继续已经暂停的版本升级过程

restart 重启版本升级过程
undo 回滚到上一级版本（可以使用–to-revision回滚到指定版本）

# 查看当前升级版本的状态
[root@master k8sYamlForCSDN]# kubectl rollout status deployment deploy-controller -n dev
deployment "deploy-controller" successfully rolled out
# 查看升级历史记录
[root@master k8sYamlForCSDN]# kubectl rollout history deployment deploy-controller -n dev
deployment.apps/deploy-controller 
REVISION  CHANGE-CAUSE
1         kubectl apply --filename=deploy-controller.yaml --record=true
2         kubectl apply --filename=deploy-controller.yaml --record=true
# 升级镜像
[root@master k8sYamlForCSDN]# kubectl set image deploy deploy-controller -n dev nginx=nginx:1.17.2
deployment.apps/deploy-controller image updated
# 查看当前升级版本的状态
[root@master k8sYamlForCSDN]# kubectl rollout status deployment deploy-controller -n dev
Waiting for deployment "deploy-controller" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "deploy-controller" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "deploy-controller" rollout to finish: 1 old replicas are pending termination...
deployment "deploy-controller" successfully rolled out
# 查看升级历史记录
# 可以发现有三次版本记录，说明完成过两次升级
[root@master k8sYamlForCSDN]# kubectl rollout history deployment deploy-controller -n dev
deployment.apps/deploy-controller 
REVISION  CHANGE-CAUSE
1         kubectl apply --filename=deploy-controller.yaml --record=true
2         kubectl apply --filename=deploy-controller.yaml --record=true
3         kubectl apply --filename=deploy-controller.yaml --record=true
# 版本回滚
# 这里直接使用--to-revision=1回滚到了1版本(nginx)， 如果省略这个选项，就是回退到上个版本，就是2版本(nginx1.17.1),目前是3版本（1.17.2）
[root@master k8sYamlForCSDN]# kubectl rollout undo deployment deploy-controller -n dev --to-revision=1
deployment.apps/deploy-controller rolled back
# 查看发现，通过nginx镜像版本可以发现到了第一版
[root@master k8sYamlForCSDN]# kubectl get deploy -n dev -o wide
NAME                READY   UP-TO-DATE   AVAILABLE   AGE   CONTAINERS   IMAGES   SELECTOR
deploy-controller   3/3     2            3           10m   nginx        nginx    app=nginx-pod
# 查看rs，发现第三个rs中有3个pod运行
# 其实deployment之所以可是实现版本的回滚，就是通过记录下历史rs来实现的，
# 一旦想回滚到哪个版本，只需要将当前版本pod数量降为0，然后将回滚版本的pod提升为目标数量就可以了
[root@master k8sYamlForCSDN]# kubectl get rs -n dev 
NAME                           DESIRED   CURRENT   READY   AGE
deploy-controller-67db8df7f5   0         0         0       7m36s
deploy-controller-6f44b7b85d   0         0         0       12m
deploy-controller-7f87645cd8   3         3         3       13m

金丝雀发布

eployment控制器支持控制更新过程中的控制，如“暂停(pause)”或“继续(resume)”更新操作。

比如有一批新的Pod资源创建完成后立即暂停更新过程，此时，仅存在一部分新版本的应用，主体部分还是旧的版本。然后，再筛选一小部分的用户请求路由到新版本的Pod应用，继续观察能否稳定地按期望的方式运行。确定没问题之后再继续完成余下的Pod资源滚动更新，否则立即回滚更新操作。这就是所谓的金丝雀发布。

# 更新deployment的版本，并配置暂停deployment
[root@master k8sYamlForCSDN]# kubectl set image deploy deploy-controller -n dev nginx=nginx:1.17.1 && kubectl rollout pause deploy deploy-controller -n dev
deployment.apps/deploy-controller image updated
deployment.apps/deploy-controller paused
#观察更新状态
# 监控更新的过程，可以看到已经新增了一个资源，但是并未按照预期的状态去删除一个旧的资源，就是因为使用了pause暂停命令
[root@master k8sYamlForCSDN]# kubectl get rs -n dev 
NAME                           DESIRED   CURRENT   READY   AGE
deploy-controller-67db8df7f5   0         0         0       19m
deploy-controller-6f44b7b85d   1         1         1       24m
deploy-controller-7f87645cd8   3         3         3       25m
[root@master k8sYamlForCSDN]# kubectl rollout status deploy deploy-controller -n dev
Waiting for deployment "deploy-controller" rollout to finish: 1 out of 3 new replicas have been updated...
# 确保更新的pod没问题了，继续更新
[root@master k8sYamlForCSDN]# kubectl rollout resume deploy deploy-controller -n dev
deployment.apps/deploy-controller resumed
# 查看最后的更新情况
[root@master k8sYamlForCSDN]# kubectl get rs -n dev -o wide
NAME                           DESIRED   CURRENT   READY   AGE   CONTAINERS   IMAGES         SELECTOR
deploy-controller-67db8df7f5   0         0         0       23m   nginx        nginx:1.17.2   
deploy-controller-6f44b7b85d   3         3         3       28m   nginx        nginx:1.17.1   
deploy-controller-7f87645cd8   0         0         0       29m   nginx        nginx        
[root@master k8sYamlForCSDN]# kubectl get pods -n dev 
NAME                                 READY   STATUS    RESTARTS   AGE
deploy-controller-6f44b7b85d-7gfp7   1/1     Running   0          2m13s
deploy-controller-6f44b7b85d-b6n2b   1/1     Running   0          5m30s
deploy-controller-6f44b7b85d-cg4vq   1/1     Running   0          116s

删除Deployment

# 删除deployment，其下的rs和pod也将被删除
[root@master k8sYamlForCSDN]# kubectl delete -f deploy-controller.yaml 
deployment.apps "deploy-controller" deleted

Horizontal Pod Autoscaler(HPA)

HPA概述

我们已经可以实现通过手工执行kubectl scale命令或者edit和改yaml实现Pod扩容或缩容，但是这显然不符合Kubernetes的定位目标：自动化、智能化。 Kubernetes期望可以实现通过监测Pod的使用情况，实现pod数量的自动调整，于是就产生了Horizontal Pod Autoscaler（HPA）这种控制器。

HPA可以获取每个Pod利用率，然后和HPA中定义的指标进行对比，同时计算出需要伸缩的具体值，最后实现Pod的数量的调整。其实HPA与之前的Deployment一样，也属于一种Kubernetes资源对象，它通过追踪分析RC控制的所有目标Pod的负载变化情况，来确定是否需要针对性地调整目标Pod的副本数，这是HPA的实现原理。

1.安装metrics-server

metrics-server可以用来收集集群中的资源使用情况

# 从一个能访问外网的地方下载一个yaml文件
https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/high-availability.yaml
# 修改yaml的参数
# 找到这个  -> image: k8s.gcr.io/metrics-server/metrics-server:v0.6.1
# 替换为    -> image: registry.cn-hangzhou.aliyuncs.com/google_containers/metrics-server:v0.6.1
# 再增加几个参数
hostNetwork: true
args:
- --kubelet-insecure-tls
- --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP
# 如图

# 将编辑好的yaml文件复制到linux中
[root@master k8sYamlForCSDN]# vi components.yaml
# 运行
[root@master k8sYamlForCSDN]# kubectl apply -f components.yaml 
serviceaccount/metrics-server created
clusterrole.rbac.authorization.k8s.io/system:aggregated-metrics-reader created
clusterrole.rbac.authorization.k8s.io/system:metrics-server created
rolebinding.rbac.authorization.k8s.io/metrics-server-auth-reader created
clusterrolebinding.rbac.authorization.k8s.io/metrics-server:system:auth-delegator created
clusterrolebinding.rbac.authorization.k8s.io/system:metrics-server created
service/metrics-server created
deployment.apps/metrics-server created
apiservice.apiregistration.k8s.io/v1beta1.metrics.k8s.io created
# 查看pod运行情况
[root@master k8sYamlForCSDN]# kubectl get pods -n kube-system 
NAME                              READY   STATUS    RESTARTS   AGE
......
metrics-server-77ccdd66cf-h7279   1/1     Running   0          7m53s
# 使用kubectl top nodes 查看资源使用情况
[root@master k8sYamlForCSDN]# kubectl top nodes
NAME     CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%   
master   84m          4%     1116Mi          64%       
node1    21m          1%     646Mi           37%       
node2    16m          0%     466Mi           27%  
[root@master k8sYamlForCSDN]# kubectl top pods -n kube-system 
NAME                              CPU(cores)   MEMORY(bytes)   
coredns-64897985d-7bx2c           2m           20Mi            
coredns-64897985d-w5c4q           1m           17Mi            
etcd-master                       11m          126Mi           
kube-apiserver-master             31m          306Mi           
kube-controller-manager-master    13m          84Mi            
kube-flannel-ds-78vrj             1m           20Mi            
kube-flannel-ds-lv5jf             2m           19Mi            
kube-flannel-ds-tk9c6             2m           20Mi            
kube-proxy-vxflz                  1m           30Mi            
kube-proxy-w886d                  1m           20Mi            
kube-proxy-xb7kp                  1m           19Mi            
kube-scheduler-master             2m           36Mi            
metrics-server-77ccdd66cf-h7279   4m           18Mi
# 至此,metrics-server安装完成

我这里提供一份编写博文时用的并且已经修改好的yaml文件，2022年3月2日22点38分

apiVersion: v1
kind: ServiceAccount
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server
    rbac.authorization.k8s.io/aggregate-to-admin: "true"
    rbac.authorization.k8s.io/aggregate-to-edit: "true"
    rbac.authorization.k8s.io/aggregate-to-view: "true"
  name: system:aggregated-metrics-reader
rules:
  - apiGroups:
      - metrics.k8s.io
    resources:
      - pods
      - nodes
    verbs:
      - get
      - list
      - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server
  name: system:metrics-server
rules:
  - apiGroups:
      - ""
    resources:
      - nodes/metrics
    verbs:
      - get
  - apiGroups:
      - ""
    resources:
      - pods
      - nodes
    verbs:
      - get
      - list
      - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server-auth-reader
  namespace: kube-system
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: extension-apiserver-authentication-reader
subjects:
  - kind: ServiceAccount
    name: metrics-server
    namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server:system:auth-delegator
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:auth-delegator
subjects:
  - kind: ServiceAccount
    name: metrics-server
    namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server
  name: system:metrics-server
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:metrics-server
subjects:
  - kind: ServiceAccount
    name: metrics-server
    namespace: kube-system
---
apiVersion: v1
kind: Service
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
spec:
  ports:
    - name: https
      port: 443
      protocol: TCP
      targetPort: https
  selector:
    k8s-app: metrics-server
---
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    k8s-app: metrics-server
  name: metrics-server
  namespace: kube-system
spec:
  selector:
    matchLabels:
      k8s-app: metrics-server
  strategy:
    rollingUpdate:
      maxUnavailable: 0
  template:
    metadata:
      labels:
        k8s-app: metrics-server
    spec:
      hostNetwork: true
      containers:
        - args:
            - --cert-dir=/tmp
            - --secure-port=4443
            - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname
            - --kubelet-use-node-status-port
            - --metric-resolution=15s
            - --kubelet-insecure-tls
            - --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP
          # image: k8s.gcr.io/metrics-server/metrics-server:v0.6.1
          image: registry.cn-hangzhou.aliyuncs.com/google_containers/metrics-server:v0.6.1
          imagePullPolicy: IfNotPresent
          livenessProbe:
            failureThreshold: 3
            httpGet:
              path: /livez
              port: https
              scheme: HTTPS
            periodSeconds: 10
          name: metrics-server
          ports:
            - containerPort: 4443
              name: https
              protocol: TCP
          readinessProbe:
            failureThreshold: 3
            httpGet:
              path: /readyz
              port: https
              scheme: HTTPS
            initialDelaySeconds: 20
            periodSeconds: 10
          resources:
            requests:
              cpu: 100m
              memory: 200Mi
          securityContext:
            allowPrivilegeEscalation: false
            readOnlyRootFilesystem: true
            runAsNonRoot: true
            runAsUser: 1000
          volumeMounts:
            - mountPath: /tmp
              name: tmp-dir
      nodeSelector:
        kubernetes.io/os: linux
      priorityClassName: system-cluster-critical
      serviceAccountName: metrics-server
      volumes:
        - emptyDir: {}
          name: tmp-dir
---
apiVersion: apiregistration.k8s.io/v1
kind: APIService
metadata:
  labels:
    k8s-app: metrics-server
  name: v1beta1.metrics.k8s.io
spec:
  group: metrics.k8s.io
  groupPriorityMinimum: 100
  insecureSkipTLSVerify: true
  service:
    name: metrics-server
    namespace: kube-system
  version: v1beta1
  versionPriority: 100

2.准备deployment和servie

准备deployment和servie的yaml文件

apiVersion: apps/v1
kind: Deployment
metadata:
  name: deploy-nginx
  namespace: dev
spec:
  selector:
    matchLabels:
      run: "nginx"
  template:
    metadata:
      labels:
        run: "nginx"
    spec:
      containers:
        - name: nginx
          image: nginx
          ports:
            - containerPort: 80
              protocol: TCP
---
apiVersion: v1
kind: Service
metadata:
  name: svc-deploy-nginx
  namespace: dev
spec:
  type: NodePort
  clusterIP: 10.109.68.72
  selector:
    run: "nginx"
  ports:
    - port: 6872
      targetPort: 80
      protocol: TCP

[root@master k8sYamlForCSDN]# vi svc-deploy.yaml
# 创建
[root@master k8sYamlForCSDN]# kubectl apply -f svc-deploy.yaml 
deployment.apps/deploy-nginx created
# 查看
[root@master k8sYamlForCSDN]# kubectl get deploy,pods,svc -n dev
NAME                           READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/deploy-nginx   0/1     1            0           3s
NAME                               READY   STATUS              RESTARTS   AGE
pod/deploy-nginx-69ccc6c65-mbhds   0/1     ContainerCreating   0          3s
NAME                       TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE
service/svc-deploy-nginx   NodePort   10.109.68.72   <none>        6872:30709/TCP   3s
[root@master k8sYamlForCSDN]# kubectl get deploy,pods,svc -n dev
NAME                           READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/deploy-nginx   1/1     1            1           62s
NAME                               READY   STATUS    RESTARTS   AGE
pod/deploy-nginx-69ccc6c65-mbhds   1/1     Running   0          62s
NAME                       TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE
service/svc-deploy-nginx   NodePort   10.109.68.72   <none>        6872:30709/TCP   62s

3.部署HPA

创建hpa.yaml

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: hpa
  namespace: dev
spec:
  minReplicas: 1  #最小pod数量
  maxReplicas: 10 #最大pod数量
  targetCPUUtilizationPercentage: 3 # CPU使用率指标3%
  scaleTargetRef:   # 指定要控制的nginx信息
    apiVersion: apps/v1
    kind: Deployment
    name: deploy-nginx

[root@master k8sYamlForCSDN]# vi hpa.yaml
# 创建hpa
[root@master k8sYamlForCSDN]# kubectl apply -f hpa.yaml 
horizontalpodautoscaler.autoscaling/hpa created
# 查看hpa，发现是unknow，为什么呢？
[root@master k8sYamlForCSDN]# kubectl get hpa -n dev
NAME   REFERENCE                 TARGETS        MINPODS   MAXPODS   REPLICAS   AGE
hpa    Deployment/deploy-nginx   <unknown>/3%   1         10        1          21s

好，现在启动，压测，发现没有自动伸缩，怎么回事呢？我们查看hpa看看

[root@master k8sYamlForCSDN]# kubectl describe hpa -n dev 
Events:
  Type     Reason                        Age                     From                       Message
  ----     ------                        ----                    ----                       -------
  Warning  FailedComputeMetricsReplicas  7m5s (x12 over 9m50s)   horizontal-pod-autoscaler  invalid metrics (1 invalid out of 1), first error is: failed to get cpu utilization: missing request for cpu
  Warning  FailedGetResourceMetric       4m50s (x21 over 9m50s)  horizontal-pod-autoscaler  failed to get cpu utilization: missing request for cpu

这是因为我们上面创建的 Pod 对象没有添加 request 资源声明，这样导致 HPA 读取不到 CPU 指标信息，所以如果要想让 HPA 生效，对应的 Pod 资源必须添加 requests 资源声明，好，现在我们需要更新之前的deploy的yaml文件

# 其实这里与之前的就加了个resources.cpu资源配额的限制
apiVersion: apps/v1
kind: Deployment
metadata:
  name: deploy-nginx
  namespace: dev
spec:
  selector:
    matchLabels:
      run: "nginx"
  template:
    metadata:
      labels:
        run: "nginx"
    spec:
      containers:
        - name: nginx
          image: nginx
          ports:
            - containerPort: 80
              protocol: TCP
          resources:
            requests:
              cpu: 100m
---
apiVersion: v1
kind: Service
metadata:
  name: svc-deploy-nginx
  namespace: dev
spec:
  type: NodePort
  clusterIP: 10.109.68.72
  selector:
    run: "nginx"
  ports:
    - port: 6872
      targetPort: 80
      protocol: TCP

# 修改完后重新启动
[root@master k8sYamlForCSDN]# kubectl apply -f svc-deploy.yaml
deployment.apps/deploy-nginx created
service/svc-deploy-nginx created
[root@master k8sYamlForCSDN]# kubectl apply -f hpa.yaml 
horizontalpodautoscaler.autoscaling/hpa created
[root@master k8sYamlForCSDN]# kubectl get svc -n dev -o wide
NAME                       TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)          AGE     SELECTOR
service/svc-deploy-nginx   NodePort   10.109.68.72   <none>        6872:30627/TCP   3m16s   run=nginx
[root@master k8sYamlForCSDN]# kubectl get hpa -n dev -w
NAME   REFERENCE                 TARGETS        MINPODS   MAXPODS   REPLICAS   AGE
hpa    Deployment/deploy-nginx   0%/3%          1         10        1          75s

4.测试

使用压测工具对service地址192.168.109.100:30627进行压测，然后通过控制台查看hpa和pod的变化

# deployment变化
[root@master k8sYamlForCSDN]# kubectl get deploy -n dev -w
NAME           READY   UP-TO-DATE   AVAILABLE   AGE
deploy-nginx   8/8     8            8           5m9s
deploy-nginx   8/10    8            8           7m4s
deploy-nginx   8/10    8            8           7m4s
deploy-nginx   8/10    8            8           7m4s
deploy-nginx   8/10    10           8           7m4s
deploy-nginx   9/10    10           9           7m7s
deploy-nginx   10/10   10           10          7m21s
......
# hpa变化
[root@master k8sYamlForCSDN]# kubectl get hpa -n dev -w
NAME   REFERENCE                 TARGETS        MINPODS   MAXPODS   REPLICAS   AGE
hpa    Deployment/deploy-nginx   25%/3%         1         10        2          3m45s
hpa    Deployment/deploy-nginx   23%/3%         1         10        4          4m
hpa    Deployment/deploy-nginx   7%/3%          1         10        8          4m15s
hpa    Deployment/deploy-nginx   12%/3%         1         10        8          4m31s
......
# pod变化
[root@master k8sYamlForCSDN]# kubectl get pods -n dev -w
NAME                           READY   STATUS    RESTARTS   AGE
deploy-nginx-dbcdd8c84-gc5tk   1/1     Running   0          3m39s
deploy-nginx-dbcdd8c84-zm7gf   1/1     Running   0          6s
deploy-nginx-dbcdd8c84-2whpd   0/1     Pending   0          0s
deploy-nginx-dbcdd8c84-qf62m   0/1     Pending   0          0s
deploy-nginx-dbcdd8c84-2whpd   0/1     Pending   0          0s
deploy-nginx-dbcdd8c84-qf62m   0/1     Pending   0          0s
......

kubernetes—Controller详解（二）：https://developer.aliyun.com/article/1417727

kubernetes—Controller详解（一）