用戶但願應用程序始終可用,開發人員應該天天屢次部署新版本的應用程序。在Kubernetes中,這是經過滾動更新完成的。滾動更新容許經過使用新的實例逐步更新Pods實例來實現部署的更新,從而實現零停機。新的Pod將在具備可用資源的節點上進行調度。bash
在上一個模塊中,咱們將應用程序擴展爲運行多個實例。這是執行更新而不影響應用程序可用性的要求。默認狀況下,更新期間可用的最大Pod數和能夠建立的新Pod的最大數量是一個。兩個選項均可以配置爲數字或百分比(Pod)。在Kubernetes中,更新是版本化的,任何部署更新均可以恢復爲之前的(穩定)版本。mvc
版本記錄:每次對Deployment的操做,都能保存下來,給予後續可能的回滾使用。curl
[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 <none> <none> kubernetes-bootcamp-7f56b875dc-cp7cq 1/1 Running 0 19m 10.244.0.16 master <none> <none> kubernetes-bootcamp-7f56b875dc-n4wkn 1/1 Running 0 19m 10.244.0.17 master <none> <none> kubernetes-bootcamp-7f56b875dc-pmvck 1/1 Running 0 19m 10.244.0.18 master <none> <none>
[root@master ~]# kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=jocatalin/kubernetes-bootcamp:v2 deployment.extensions/kubernetes-bootcamp image updated
[root@master ~]# kubectl rollout status deployment/kubernetes-bootcamp deployment "kubernetes-bootcamp" successfully rolled out
能夠看見Pod的IP都變成新的了ide
[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 <none> <none> kubernetes-bootcamp-5bf4d5689b-c4zgm 1/1 Running 0 3m41s 10.244.0.22 master <none> <none> kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 3m45s 10.244.0.19 master <none> <none> kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 3m41s 10.244.0.21 master <none> <none>
第一種狀況:google
咱們這裏假設更新錯了一個不存在的鏡像url
kubectl set image deployments/kubernetes-bootcamp kubernetes-bootcamp=gcr.io/google-samples/kubernetes-bootcamp:v10 deployment.extensions/kubernetes-bootcamp image updated
查看Podsspa
[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 <none> <none> kubernetes-bootcamp-597cfc5b76-pslgq 0/1 ImagePullBackOff 0 51s 10.244.0.24 master <none> <none> kubernetes-bootcamp-5bf4d5689b-5f68t 1/1 Running 0 23m 10.244.0.20 master <none> <none> kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 23m 10.244.0.19 master <none> <none> kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 23m 10.244.0.21 master <none> <none>
查看部署blog
[root@master ~]# kubectl get deployments -l run=kubernetes-bootcamp NAME READY UP-TO-DATE AVAILABLE AGE kubernetes-bootcamp 3/4 2 3 5h58m
回滾資源
[root@master ~]# kubectl rollout undo deployments/kubernetes-bootcamp deployment.extensions/kubernetes-bootcamp rolled back
查看pods開發
[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 <none> <none> kubernetes-bootcamp-5bf4d5689b-c99hl 1/1 Running 0 29m 10.244.0.19 master <none> <none> kubernetes-bootcamp-5bf4d5689b-jrrh9 1/1 Running 0 70s 10.244.0.25 master <none> <none> kubernetes-bootcamp-5bf4d5689b-rpff2 1/1 Running 0 28m 10.244.0.21 master <none> <none>
能夠看出兩個錯誤的Pods已經不見了,原來的10.244.0.22的pods被新的Pods(10.244.0.25)取代。
第二種狀況,咱們所有更新完成,是否還能夠回滾
[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
證實也是能夠回滾前一個版本的。