This project shows how to deploy an app to a Kubernetes cluster with no downtime.
Using this project, you will deploy a Spring Boot app exposing this REST endpoint:
@RestController
class ZDT {
@GetMapping("/")
String zeroDowntime() {
return "V1";
}
}There are two versions of this app:
alexandreroman/k8s-zdt-demo:v1returning"V1"alexandreroman/k8s-zdt-demo:v2returning"V2"
Use these Kubernetes descriptors to deploy app V1:
$ kubectl apply -f k8s
namespace/zdt created
deployment.apps/zdt created
service/app-lb createdMonitor app deployment using this command:
$ kubectl -n zdt rollout status deployment/zdt
Waiting for deployment "zdt" rollout to finish: 0 of 3 updated replicas are available...
Waiting for deployment "zdt" rollout to finish: 1 of 3 updated replicas are available...
Waiting for deployment "zdt" rollout to finish: 2 of 3 updated replicas are available...
deployment "zdt" successfully rolled outCheck that all pods/services are up and running:
NAME READY STATUS RESTARTS AGE
pod/zdt-594f4794f5-8s5cg 1/1 Running 0 90s
pod/zdt-594f4794f5-ns24b 1/1 Running 0 90s
pod/zdt-594f4794f5-q996p 1/1 Running 0 90s
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/app-lb LoadBalancer 10.98.101.201 localhost 80:32045/TCP 90s
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/zdt 3/3 3 3 90s
NAME DESIRED CURRENT READY AGE
replicaset.apps/zdt-594f4794f5 3 3 3 90sThis app is now available:
$ curl localhost
V1%Kubernetes is using healthcheck endpoints to monitor app readiness:
$ curl localhost/actuator/health
{"status":"UP"}%This healthcheck is defined in the deployment descriptor:
containers:
- name: app
image: alexandreroman/k8s-zdt-demo:v1
ports:
- containerPort: 8080
livenessProbe:
httpGet:
port: 8080
path: /actuator/health
initialDelaySeconds: 30
readinessProbe:
httpGet:
port: 8080
path: /actuator/health
initialDelaySeconds: 10When this endpoint is not available from an app instance, Kubernetes will not expose this pod to the load balancer. Kubernetes makes sure there are enough running pods at any time.
Run this command to edit app deployment descriptor:
$ kubectl -n zdt edit deployment zdt
deployment.extensions/zdt editedFind the Docker image reference alexandreroman/k8s-zdt-demo:v1, and
change it to alexandreroman/k8s-zdt-demo:v2:
image: alexandreroman/k8s-zdt-demo:v1
# change to V2
image: alexandreroman/k8s-zdt-demo:v2Monitor app V2 deployment:
$ kubectl -n zdt rollout status deployment/zdt
Waiting for deployment "zdt" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 1 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 2 out of 3 new replicas have been updated...
Waiting for deployment "zdt" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "zdt" rollout to finish: 1 old replicas are pending termination...
deployment "zdt" successfully rolled outWhile the new app version is being deployed, the old version is still
available. If you hit the app endpoint while the new version is being
deployed, you may see two versions running at the same time: V1 and
V2. Kubernetes is gradually stopping old pods as new pods are seen
as ready.
You may also edit the Kubernetes descriptor to point at the new app version, in zdt-02-deployment.yml:
containers:
- name: app
image: alexandreroman/k8s-zdt-demo:v2Then, you need to update your Kubernetes deployment:
$ kubectl apply -f k8s
namespace/zdt unchanged
deployment.apps/zdt configured
service/app-lb unchangedUsing this command, you can rollback to the previously deployed app version:
$ kubectl -n zdt rollout undo deployment/zdt
deployment.extensions/zdt rolled backContributions are always welcome!
Feel free to open issues & send PR.
Copyright © 2019 Pivotal Software, Inc.
This project is licensed under the Apache Software License version 2.0.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent