Kubernetes(K8S) Deployment 升级和回滚

kubernetes,k8s,deployment,升级 · 浏览次数 : 1116

小编点评

# Pod 的升级和回滚 **CHANGE-CAUSE：**没有值 **创建 Deployment 时使用 --record 参数，就可以在 CHANGE-CAUSE 列看到每个版本使用的命令了** **Pod 的升级和回滚** **1. 创建 Deployment** ``` kubectl create deployment javademo1Name: javademo1Namespace: defaultCreationTimestamp: Tue, 18 Oct 2022 18:51:25 +0800Labels: app=javademo1Annotations: deployment.kubernetes.io/revision: 4 ``` **2. 修改 Deployment 的 spec** ``` spec: strategy: type: RollingUpdate maxUnavailable: 25% max surgePod Template: Labels: app=javademo1 Containers: vipsoft: Image: registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0 Port: <none> Host Port: <none> Environment: <none> Mounts: <none> Volumes: <none> conditions: Type: Type Status Reason ---- ------ ------ Available True MinimumReplicasAvailable Progressing True NewReplicaSetAvailableOldReplicaSets: <none>NewReplicaSet: javademo1-84dd5c9485 (3/3 replicas created)Events: Type Reason Age From Message ---- ------ ---- ---- ------- Normal ScalingReplicaSet 51m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 1 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 2 Normal ScalingReplicaSet 50m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 2 Normal ScalingReplicaSet 50m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 3 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 1 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 0 Normal ScalingReplicaSet 20s (x2 over 5h57m) deployment-controller Scaled up replica set javademo1-84dd5c9485 to 1 Normal ScalingReplicaSet 17s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 2 Normal ScalingReplicaSet 17s deployment-controller Scaled up replica set javademo1-84dd5c9485 to 2 Normal ScalingReplicaSet 15s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 1 Normal ScalingReplicaSet 14s (x2 over 5h51m) deployment-controller Scaled up replica set javademo1-84dd5c9485 to 3 Normal ScalingReplicaSet 9s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 0** **3. 创建 Deployment 的资源** ``` kubectl create deployment javademo1Name: javademo1Namespace: defaultCreationTimestamp: Tue, 18 Oct 2022 18:51:25 +0800Labels: app=javademo1Annotations: deployment.kubernetes.io/revision: 4 ``` **4. 修改 Deployment 的 spec** ``` spec: strategy: type: RollingUpdate maxUnavailable: 50% max surgePod Template: Labels: app=javademo1 Containers: vipsoft: Image: registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0 Port: <none> Host Port: <none> Environment: <none> Mounts: <none> Volumes: <none> conditions: Type: Type Status Reason ---- ------ ------ Available True MinimumReplicasAvailable Progressing True NewReplicaSetAvailableOldReplicaSets: <none>NewReplicaSet: javademo1-84dd5c9485 (3/3 replicas created)Events: Type Reason Age From Message ---- ------ ---- ---- ------- Normal ScalingReplicaSet 50m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 1 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 2 Normal ScalingReplicaSet 50m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 2 Normal ScalingReplicaSet 50m deployment-controller Scaled up replica set javademo1-5476dc8d7d to 3 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 1 Normal ScalingReplicaSet 50m deployment-controller Scaled down replica set javademo1-84dd5c9485 to 0 Normal ScalingReplicaSet 20s (x2 over 5h57m) deployment-controller Scaled up replica set javademo1-84dd5c9485 to 1 Normal ScalingReplicaSet 17s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 2 Normal ScalingReplicaSet 17s deployment-controller Scaled up replica set javademo1-84dd5c9485 to 2 Normal ScalingReplicaSet 15s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 1 Normal ScalingReplicaSet 14s (x2 over 5h51m) deployment-controller Scaled up replica set javademo1-84dd5c9485 to 3 Normal ScalingReplicaSet 9s deployment-controller Scaled down replica set javademo1-5476dc8d7d to 0** ``` **5. 创建 Deployment 的资源** ``` kubectl create deployment javademo1Name: javademo1Namespace: defaultCreationTimestamp: Tue, 18 Oct 2022 18:51:25 +0800Labels: app=javademo1Annotations: deployment.kubernetes.io/revision: 4 ```

正文

创建部署详见 Kubernetes(K8S) Deployment 部署 Pod
传统应用升级，一般是V1.0的jar包，有一个应对 1.0 的 shell 启动脚本。升级时，传 2.0 的 jar包，配置 2.0 的 shell 脚本。
执行顺序为，停1.0的服务，启2.0的服务，有问题时，把2.0停掉再执行 1.0的shell脚本，手动恢复到1.0 版本，服务多的情况下，就很崩溃，而且服务会中断不可用。

高并发、高可用系统普及的今天，服务的升级更新至少要做到“业务不中断”。而滚动更新(Rolling-update)恰是满足这一需求的一种系统更新升级方案。对于Kubernetes集群部署的Service来说，Rolling update就是指一次仅更新一个Pod，然后逐个进行更新，而不是在同一时刻将该Service下面的所有Pod shutdown，然后去更新，逐个更新可以避免将业务中断；

RollingUpdate命令：通过自动扩缩容实现自动滚动升级
指令格式：kubectl rolling-update 旧控制器新控制器 --image=新镜像

1. 发布新版到阿里云镜像仓库

Last login: Tue Oct 18 16:39:53 2022 from 172.16.0.66
[root@localhost ~]# cd /opt/demo/
[root@localhost demo]# ll
总用量 17160
-rw-r--r--. 1 root root 17566582 10月 19 11:50 demojenkins.jar
-rw-r--r--. 1 root root      126 10月 18 17:56 Dockerfile
[root@localhost demo]# docker login --username=hiXXXXXX@aliyun.com registry.cn-shanghai.aliyuncs.com
Password: 
WARNING! Your password will be stored unencrypted in /root/.docker/config.json.
Configure a credential helper to remove this warning. See
https://docs.docker.com/engine/reference/commandline/login/#credentials-store

Login Succeeded
[root@localhost demo]# docker build -t demo:2.0 .
Sending build context to Docker daemon  17.57MB
Step 1/4 : FROM openjdk:8-jdk-alpine
 ---> a3562aa0b991
Step 2/4 : VOLUME /tmp
 ---> Using cache
 ---> a98cf1fbeb9d
Step 3/4 : ADD ./demojenkins.jar demojenkins.jar
 ---> 12bfc15e5295
Step 4/4 : ENTRYPOINT ["java","-jar","/demojenkins.jar", "&"]
 ---> Running in 2908dd94dfa7
Removing intermediate container 2908dd94dfa7
 ---> 3041db93b6df
Successfully built 3041db93b6df
Successfully tagged demo:2.0
[root@localhost demo]# docker images
REPOSITORY                                          TAG            IMAGE ID       CREATED              SIZE
demo                                                2.0            3041db93b6df   About a minute ago   122MB
demo                                                1.0            1952e02daef9   19 hours ago         122MB 
openjdk                                             8-jdk-alpine   a3562aa0b991   3 years ago          105MB
java                                                8              d23bdf5b1b1b   5 years ago          643MB
[root@localhost demo]# docker tag 3041db93b6df registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:2.0
[root@localhost demo]# docker push registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:2.0
The push refers to repository [registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft]
c38fa2e9a247: Pushed 
ceaf9e1ebef5: Layer already exists 
9b9b7f3d56a0: Layer already exists 
f1b5933fe4b5: Layer already exists 
2.0: digest: sha256:785c4707e7e629d22b5609310d8f945c107b277b94925cc0dbebdb3621b size: 1159
[root@localhost demo]#

2. 查看现有Pod信息（升级后对比用）

# 查看创建的 Deployment 对象
[root@k8smaster ~]# kubectl get deployments
NAME        READY   UP-TO-DATE   AVAILABLE   AGE
javademo1   3/3     3            3           19h
nginx       1/1     1            1           5d21h

# 查看 Deployment 上线状态
[root@k8smaster ~]# kubectl rollout status deployment/javademo1
deployment "javademo1" successfully rolled out

# 查看 Deployment 对象创建的 ReplicaSet：
[root@k8smaster ~]# kubectl get rs
NAME                   DESIRED   CURRENT   READY   AGE 
javademo1-84dd5c9485   3         3         3       4h25m
nginx-f89759699        1         1         1       5d21h

# 查看 Deployment 对象操作 ReplicaSet 创建的 Pod，并显示生成的标签：
[root@k8smaster ~]# kubectl get pods --show-labels
NAME                         READY   STATUS    RESTARTS   AGE     LABELS
javademo1-84dd5c9485-7vgkr   1/1     Running   0          4h20m   app=javademo1,pod-template-hash=84dd5c9485
javademo1-84dd5c9485-8ckk6   1/1     Running   0          4h26m   app=javademo1,pod-template-hash=84dd5c9485
javademo1-84dd5c9485-8hfmd   1/1     Running   0          4h20m   app=javademo1,pod-template-hash=84dd5c9485
nginx-f89759699-5hkdw        1/1     Running   0          5d21h   app=nginx,pod-template-hash=f89759699

#查看当前运行的 Pod
[root@k8smaster ~]# kubectl get pods
NAME                         READY   STATUS    RESTARTS   AGE
javademo1-84dd5c9485-7vgkr   1/1     Running   0          4h31m
javademo1-84dd5c9485-8ckk6   1/1     Running   0          4h37m
javademo1-84dd5c9485-8hfmd   1/1     Running   0          4h31m
nginx-f89759699-5hkdw        1/1     Running   0          5d21h 

# 查看 Deployeement 详情
[root@k8smaster ~]# kubectl describe deployment/javademo1
Name:                   javademo1
Namespace:              default
CreationTimestamp:      Tue, 18 Oct 2022 18:51:25 +0800
Labels:                 app=javademo1
Annotations:            deployment.kubernetes.io/revision: 2
Selector:               app=javademo1
Replicas:               3 desired | 3 updated | 3 total | 3 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=javademo1
  Containers:
   vipsoft:
    Image:        registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0
    Port:         <none>
    Host Port:    <none>
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Progressing    True    NewReplicaSetAvailable
  Available      True    MinimumReplicasAvailable
OldReplicaSets:  <none>
NewReplicaSet:   javademo1-84dd5c9485 (3/3 replicas created)
Events:          <none>

Deployment, kubectl get deployments 所显示的字段有：

NAME：列出了名字空间中 Deployment 的名称。
READY：显示应用程序的可用的“副本”数。显示的模式是“就绪个数/期望个数”。
UP-TO-DATE：显示为了达到期望状态已经更新的副本数。
AVAILABLE：显示应用可供用户使用的副本数。
AGE：显示应用程序运行的时间。

ReplicaSet, kubectl get rs 输出中包含以下字段：

NAME：列出名字空间中 ReplicaSet 的名称；[Deployment名称]-[随机字符串]
DESIRED：显示应用的期望副本个数，即在创建 Deployment 时所定义的值。此为期望状态；
CURRENT：显示当前运行状态中的副本个数；
READY：显示应用中有多少副本可以为用户提供服务；
AGE：显示应用已经运行的时间长度。

3. Deployment 升级

方法1

# 查看 Deployeement 详情 里可以看到 Containers 内容 
#  Containers:
#   vipsoft:
#    Image:        registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0
# 设置vipsoft容器中的镜像，镜像版本变了，就会触发重新拉取动作
[root@k8smaster ~]# kubectl set image deployment/javademo1 vipsoft=registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:2.0
deployment.apps/javademo1 image updated
[root@k8smaster ~]# kubectl rollout status deployment/javademo1
deployment "javademo1" successfully rolled out
[root@k8smaster ~]#

方法2：Jenkins 的时候比较常用，早期使用 Kubernetes Continuous Deploy 插件，现已重淘汰，可使用 Config File Provider 代替

# Containers:
#   vipsoft:
#    Image:        registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0  #将 ：1.0 改成 2.0
# 一旦镜像名（或 Pod 定义）发生了修改，则触发 k8s 系统完成 Deployment 所有运行 Pod 的滚动升级操作
[root@k8smaster ~]# kubectl edit deployment/javademo1
deployment "javademo1" edited

查看已更新的 Deployment 的信息

[root@k8smaster ~]# kubectl get deployments
NAME        READY   UP-TO-DATE   AVAILABLE   AGE
javademo1   3/3     3            3           20h
nginx       1/1     1            1           5d22h

# 扩容了3个副本，将旧的 ReplicaSet 缩容到了 0 个副本，完成了Pod的升级
[root@k8smaster ~]# kubectl get rs
NAME                   DESIRED   CURRENT   READY   AGE
javademo1-5476dc8d7d   3         3         3       11m
javademo1-84dd5c9485   0         0         0       5h17m
nginx-f89759699        1         1         1       5d22h

#只有最新的  5476dc8d7d 信息
[root@k8smaster ~]# kubectl get pods
NAME                         READY   STATUS    RESTARTS   AGE
javademo1-5476dc8d7d-bf6m5   1/1     Running   0          11m
javademo1-5476dc8d7d-ttt5b   1/1     Running   0          11m
javademo1-5476dc8d7d-zd4c7   1/1     Running   0          11m
nginx-f89759699-5hkdw        1/1     Running   0          5d22h
[root@k8smaster ~]# kubectl describe deployment/javademo1
Name:                   javademo1
Namespace:              default
CreationTimestamp:      Tue, 18 Oct 2022 18:51:25 +0800
Labels:                 app=javademo1
Annotations:            deployment.kubernetes.io/revision: 3
Selector:               app=javademo1
Replicas:               3 desired | 3 updated | 3 total | 3 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=javademo1
  Containers:
   vipsoft:
    Image:        registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:2.0
    Port:         <none>
    Host Port:    <none>
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    NewReplicaSetAvailable
OldReplicaSets:  <none>
NewReplicaSet:   javademo1-5476dc8d7d (3/3 replicas created)
Events:
  Type    Reason             Age   From                   Message
  ----    ------             ----  ----                   -------
  Normal  ScalingReplicaSet  12m   deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 1
  Normal  ScalingReplicaSet  11m   deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 2
  Normal  ScalingReplicaSet  11m   deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 2
  Normal  ScalingReplicaSet  11m   deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 1
  Normal  ScalingReplicaSet  11m   deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 3
  Normal  ScalingReplicaSet  11m   deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 0
[root@k8smaster ~]#

上面 Events: 可以看出，先 up javademo1-5476dc8d7d to 1,再停 javademo1-84dd5c9485 to 2，Pod 逐个滚动式更新，如下图:

Deployment 更新策略
在 Deployment 的定义中，可以通过 spec.strategy 指定 Pod 的更新策略，目前支持两种更新策略：
Recreate（重建）：设置 spec.strategy.type=Recreate，表示 Deployment 在更新 Pod 时，会先杀掉所有正在运行的 Pod，然后创建新的 Pod。
RollingUpdate（滚动更新）：设置 spec.strategy.type=RollingUpdate，表示 Deployment 会以滚动更新的方式来逐个更新 Pod。同时，可以通过设置 spec.strategy.rollingUpdate 下的两个参数（maxUnavailable 和 maxSurge）来控制滚动更新的过程。

4. Deployment 回滚

# 查看 Deployment 部署历史， CHANGE-CAUSE 里没有值，需要在创建 Deployment 时，使用 --record 参数，就可以在 CHANGE-CAUSE 列看到每个版本使用的命令了
[root@k8smaster ~]# kubectl rollout history deployment/javademo1
deployment.apps/javademo1 
REVISION  CHANGE-CAUSE
1         <none>
2         <none> 
# 查看版本 2 的信息： 
[root@k8smaster ~]# kubectl rollout history deployment/javademo1 --revision=2
deployment.apps/javademo1 with revision #2
Pod Template:
  Labels:       app=javademo1
        pod-template-hash=84dd5c9485
  Containers:
   vipsoft:
    Image:      registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0
    Port:       <none>
    Host Port:  <none>
    Environment:        <none>
    Mounts:     <none>
  Volumes:      <none>


# kubeetl rollout undo deployment/javademo1 --to-revision=2  加上参数，回滚到指定版本
# 这边没加参加，回滚到上一个版本
[root@k8smaster ~]# kubectl rollout undo deployment/javademo1
deployment.apps/javademo1 rolled back
[root@k8smaster ~]# kubectl describe deployment/javademo1
Name:                   javademo1
Namespace:              default
CreationTimestamp:      Tue, 18 Oct 2022 18:51:25 +0800
Labels:                 app=javademo1
Annotations:            deployment.kubernetes.io/revision: 4
Selector:               app=javademo1
Replicas:               3 desired | 3 updated | 3 total | 3 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=javademo1
  Containers:
   vipsoft:
    Image:        registry.cn-shanghai.aliyuncs.com/vipsoft/vipsoft:1.0
    Port:         <none>
    Host Port:    <none>
    Environment:  <none>
    Mounts:       <none>
  Volumes:        <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    NewReplicaSetAvailable
OldReplicaSets:  <none>
NewReplicaSet:   javademo1-84dd5c9485 (3/3 replicas created)
Events:
  Type    Reason             Age                  From                   Message
  ----    ------             ----                 ----                   -------
  Normal  ScalingReplicaSet  51m                  deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 1
  Normal  ScalingReplicaSet  50m                  deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 2
  Normal  ScalingReplicaSet  50m                  deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 2
  Normal  ScalingReplicaSet  50m                  deployment-controller  Scaled up replica set javademo1-5476dc8d7d to 3
  Normal  ScalingReplicaSet  50m                  deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 1
  Normal  ScalingReplicaSet  50m                  deployment-controller  Scaled down replica set javademo1-84dd5c9485 to 0
  Normal  ScalingReplicaSet  20s (x2 over 5h57m)  deployment-controller  Scaled up replica set javademo1-84dd5c9485 to 1
  Normal  ScalingReplicaSet  17s                  deployment-controller  Scaled down replica set javademo1-5476dc8d7d to 2
  Normal  ScalingReplicaSet  17s                  deployment-controller  Scaled up replica set javademo1-84dd5c9485 to 2
  Normal  ScalingReplicaSet  15s                  deployment-controller  Scaled down replica set javademo1-5476dc8d7d to 1
  Normal  ScalingReplicaSet  14s (x2 over 5h51m)  deployment-controller  Scaled up replica set javademo1-84dd5c9485 to 3
  Normal  ScalingReplicaSet  9s                   deployment-controller  Scaled down replica set javademo1-5476dc8d7d to 0
[root@k8smaster ~]#

history 中 CHANGE-CAUSE 里没有值，需要在创建 Deployment 时，使用 --record 参数，就可以在 CHANGE-CAUSE 列看到每个版本使用的命令了

看了很多贴子，这篇不错：Pod 的升级和回滚