k8s滚动升级_k8s(6)-滚动更新

用户希望应用程序始终可用，开发人员应该每天多次部署新版本的应用程序。在Kubernetes中，这是通过滚动更新完成的。滚动更新允许通过使用新的实例逐步更新Pods实例来实现部署的更新，从而实现零停机。新的Pod将在具有可用资源的节点上进行调度。

在上一个模块中，我们将应用程序扩展为运行多个实例。这是执行更新而不影响应用程序可用性的要求。默认情况下，更新期间可用的最大Pod数和可以创建的新Pod的最大数量是一个。两个选项都可以配置为数字或百分比(Pod)。在Kubernetes中，更新是版本化的，任何部署更新都可以恢复为以前的(稳定)版本。

版本记录：每次对Deployment的操作，都能保存下来，给予后续可能的回滚使用。

1. 查看正在运行的Pods

[root@master ~]# kubectl get pods -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

kubernetes-bootcamp-7f56b875dc-6zbjt 1/1 Running 0 5h31m 10.244.0.15 master

kubernetes-bootcamp-7f56b875dc-cp7cq 1/1 Running 0 19m 10.244.0.16 master

kubernetes-bootcamp-7f56b875dc-n4wkn 1/1 Running 0 19m 10.244.0.17 master

kubernetes-bootcamp-7f56b875dc-pmvck 1/1 Running 0 19m 10.244.0.18 master

2. 使用新的镜像更新程序

[root@master ~]# kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=jocatalin/kubernetes-bootcamp:v2

deployment.extensions/kubernetes-bootcamp image updated

3. 确认更新完成

[root@master ~]# kubectl rollout status deployment/kubernetes-bootcamp

deployment "kubernetes-bootcamp" successfully rolled out

4. 查看Pod的变化

可以看见Pod的IP都变成新的了

[root@master ~]# kubectl get pods -l run=kubernetes-bootcamp -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

kubernetes-bootcamp-5bf4d5689b-5f68t 1/1 Running 0 3m44s 10.244.0.20 master

kubernetes-bootcamp-5bf4d5689b-c4zgm 1/1 Running 0 3m41s 10.244.0.22 master

kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 3m45s 10.244.0.19 master

kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 3m41s 10.244.0.21 master

5. 回滚应用

第一种情况：

我们这里假设更新错了一个不存在的镜像

kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=gcr.io/google-samples/kubernetes-bootcamp:v10

deployment.extensions/kubernetes-bootcamp image updated

查看Pods

[root@master ~]# kubectl get pods -l run=kubernetes-bootcamp -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

kubernetes-bootcamp-597cfc5b76-jtmrr 0/1 ImagePullBackOff 0 51s 10.244.0.23 master

kubernetes-bootcamp-597cfc5b76-pslgq 0/1 ImagePullBackOff 0 51s 10.244.0.24 master

kubernetes-bootcamp-5bf4d5689b-5f68t 1/1 Running 0 23m 10.244.0.20 master

kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 23m 10.244.0.19 master

kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 23m 10.244.0.21 master

查看部署

[root@master ~]# kubectl get deployments -l run=kubernetes-bootcamp

NAME READY UP-TO-DATE AVAILABLE AGE

kubernetes-bootcamp 3/4 2 3 5h58m

回滚

[root@master ~]# kubectl rollout undo deployments/kubernetes-bootcamp

deployment.extensions/kubernetes-bootcamp rolled back

查看pods

[root@master ~]# kubectl get pods -l run=kubernetes-bootcamp -o wide

NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES

kubernetes-bootcamp-5bf4d5689b-5f68t 1/1 Running 0 29m 10.244.0.20 master

kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 29m 10.244.0.19 master

kubernetes-bootcamp-5bf4d5689b-jrrh9 1/1 Running 0 70s 10.244.0.25 master

kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 28m 10.244.0.21 master

可以看出两个错误的Pods已经不见了，原来的10.244.0.22的pods被新的Pods(10.244.0.25)取代。

第二种情况，我们全部更新完成，是否还可以回滚

[root@master ~]# curl http://192.168.0.10:31238

Hello Kubernetes bootcamp! | Running on: kubernetes-bootcamp-5bf4d5689b-h4wsw | v=2

[root@master ~]# kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=jocatalin/kubernetes-bootcamp:v1

deployment.extensions/kubernetes-bootcamp image updated

[root@master ~]# curl http://192.168.0.10:31238

Hello Kubernetes bootcamp! | Running on: kubernetes-bootcamp-7f56b875dc-kvxn5 | v=1

[root@master ~]# kubectl rollout undo deployments/kubernetes-bootcamp

deployment.extensions/kubernetes-bootcamp rolled back

[root@master ~]# curl http://192.168.0.10:31238

Hello Kubernetes bootcamp! | Running on: kubernetes-bootcamp-5bf4d5689b-rfhqx | v=2

证明也是可以回滚前一个版本的。