实战演示k8s部署go服务，实现滚动更新、重新创建、蓝绿部署、金丝雀发布(转)

1 前言

本文主要实战演示k8s部署go服务，实现滚动更新、重新创建、蓝绿部署、金丝雀发布

2 go服务镜像准备

2.1 初始化项目

 cd /Users/flying/Dev/Go/go-lesson/src/
 mkdir goPublish
 cd goPublish
 go mod init goPublish

2.2 编写main.go

 package main
 ​
 import (
     "flag"
     "github.com/gin-gonic/gin"
     "net/http"
     "os"
 )
 ​
 var version = flag.String("v", "v1", "v1")
 ​
 func main() {
     router := gin.Default()
     router.GET("", func(c *gin.Context) {
         flag.Parse()
         hostname, _ := os.Hostname()
         c.String(http.StatusOK, "This is version:%s running in pod %s", *version, hostname)
     })
     router.Run(":8080")
 }

2.3 本地测试运行

 

 

 

 

2.4 编写Dockerfile

v1.3版本

 FROM golang:latest AS build
 ​
 WORKDIR /go/src/test
 COPY . /go/src/test
 RUN go env -w GOPROXY=https://goproxy.cn,direct
 RUN CGO_ENABLED=0 go build -v -o main .
 ​
 FROM alpine AS api
 RUN mkdir /app
 COPY --from=build /go/src/test/main /app
 WORKDIR /app
 ENTRYPOINT ["./main", "-v" ,"1.3 "]

2.5 构建镜像

 docker build -t go-publish:v1.3 .

2.6 发布镜像到docker hub

 docker tag go-publish:v1.3 joycode123/go-publish:v1.3
 docker push joycode123/go-publish:v1.3 # 这里需要先注册docker hub账号，并在命令行登陆

2.7 重复上述步骤，构建v2.0版本镜像并发布到docker hub

修改Dockerfile

 FROM golang:latest AS build
 ​
 WORKDIR /go/src/test
 COPY . /go/src/test
 RUN go env -w GOPROXY=https://goproxy.cn,direct
 RUN CGO_ENABLED=0 go build -v -o main .
 ​
 FROM alpine AS api
 RUN mkdir /app
 COPY --from=build /go/src/test/main /app
 WORKDIR /app
 ENTRYPOINT ["./main", "-v" ,"2.0 "]

构建并发布

 docker build -t go-publish:v2.0 .
 docker tag go-publish:v2.0 joycode123/go-publish:v2.0
 docker push joycode123/go-publish:v2.0

3 滚动更新

接下来的部分，需要k8s运行环境，这里是一主两从的k8s环境来演示

3.1 编写k8s的rolling-update.yaml文件

 apiVersion: apps/v1
 kind: Deployment
 metadata:
   name: rolling-update
   namespace: test
 spec:
   strategy:
     rollingUpdate:
       maxSurge: 25%
       maxUnavailable: 25%
     type: RollingUpdate
   selector:
     matchLabels:
       app: rolling-update
   replicas: 4
   template:
     metadata:
       labels:
         app: rolling-update
     spec:
       containers:
         - name: rolling-update
           command: ["./main","-v","v1.3"]
           image: joycode123/go-publish:v1.3
           ports:
             - containerPort: 8080
 ---
 apiVersion: v1
 kind: Service
 metadata:
   name: rolling-update
   namespace: test
 spec:
   ports:
     - port: 8080
       protocol: TCP
       targetPort: 8080
   selector:
     app: rolling-update
   type: ClusterIP

3.2 部署v1.3到k8s中

kubectl create namespace test # 创建test命名空间
kubectl apply -f rollingUpdate.yaml
kubectl get svc,deploy,pods -n test

 

3.3 测试

为了便于测试出效果，新开一个窗口，执行如下命令

while sleep 1;do curl 10.110.255.77:8080;echo "";done

 

 

3.4 部署v2.0到k8s中

修改rollingUpdate.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: rolling-update
  namespace: test
spec:
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  selector:
    matchLabels:
      app: rolling-update
  replicas: 4
  template:
    metadata:
      labels:
        app: rolling-update
    spec:
      containers:
        - name: rolling-update
          command: ["./main","-v","v2.0"]
          image: joycode123/go-publish:v2.0
          ports:
            - containerPort: 8080
---
apiVersion: v1
kind: Service
metadata:
  name: rolling-update
  namespace: test
spec:
  ports:
    - port: 8080
      protocol: TCP
      targetPort: 8080
  selector:
    app: rolling-update
  type: ClusterIP
kubectl apply -f rollingUpdate.yaml
kubectl get svc,deploy,pods -n test

 

 

注意观察测试窗体的变化，可以看到，2.0版本完全部署之前，v1.3和v2.0同时提供服务，当v2.0部署完成后，就只剩下v2.0提供服务了

 

 

 

 

4 重建部署

4.1 删除前面部署的滚动更新deployment、service、pods

kubectl delete deploy rolling-update -n test
kubectl delete svc rolling-update -n test

 

 

4.2 编写recreate.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: recreate
  namespace: test
spec:
  strategy:
    type: Recreate
  selector:
    matchLabels:
      app: recreate
  replicas: 4
  template:
    metadata:
      labels:
        app: recreate
    spec:
      containers:
        - name: recreate
          image: joycode123/go-publish:v1.3
          ports:
            - containerPort: 8080
          livenessProbe:
            tcpSocket:
              port: 8080
---
apiVersion: v1
kind: Service
metadata:
  name: recreate
  namespace: test
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8080
  selector:
    app: recreate
  type: ClusterIP

4.3 部署v1.3到k8s中

kubectl apply -f recreate.yaml
kubectl get svc,deploy,pods -n test

 

 

4.4 测试

为了便于测试出效果，新开一个窗口，执行如下命令

while sleep 1;do curl 10.111.233.53;echo "";done

 

 

4.5 部署v2.0

修改recreate.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: recreate
  namespace: test
spec:
  strategy:
    type: Recreate
  selector:
    matchLabels:
      app: recreate
  replicas: 4
  template:
    metadata:
      labels:
        app: recreate
    spec:
      containers:
        - name: recreate
          image: joycode123/go-publish:v2.0
          ports:
            - containerPort: 8080
          livenessProbe:
            tcpSocket:
              port: 8080
---
apiVersion: v1
kind: Service
metadata:
  name: recreate
  namespace: test
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8080
  selector:
    app: recreate
  type: ClusterIP
kubectl apply -f recreate.yaml
kubectl get svc,deploy,pods -n test

注意观察，在重建过程中，服务会有短暂的中断

 

 

5 蓝绿部署

5.1 删除之前的部署

kubectl delete deploy recreate -n test
kubectl delete svc recreate -n test

5.2 编写blueGreen.yaml文件（green版本）

apiVersion: apps/v1
kind: Deployment
metadata:
  name: green
  namespace: test
spec:
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  selector:
    matchLabels:
      app: bluegreen
  replicas: 4
  template:
    metadata:
      labels:
        app: bluegreen
        version: v1.3
    spec:
      containers:
        - name: bluegreen
          image: joycode123/go-publish:v1.3
          ports:
            - containerPort: 8080

5.3 部署v1.3版本(green版本)

kubectl apply -f blueGreen.yaml
kubectl get deploy,pods -n test

5.4 修改blueGreen.yaml（blue版本）

apiVersion: apps/v1
kind: Deployment
metadata:
  name: blue
  namespace: test
spec:
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  selector:
    matchLabels:
      app: bluegreen
  replicas: 4
  template:
    metadata:
      labels:
        app: bluegreen
        version: v2.0
    spec:
      containers:
        - name: bluegreen
          image: joycode123/go-publish:v2.0
          ports:
            - containerPort: 8080

5.5 部署v2.0版本（blue版本）

kubectl apply -f blueGreen.yaml
kubectl get deploy,pods -n test

 

 

这时，在k8s中同时存在blue版本和green版本在运行着

5.6 编写blueGreenService.yaml

apiVersion: v1
kind: Service
metadata:
  name: bluegreen
  namespace: test
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8080
  selector:
    app: bluegreen #注意，这里的app要和blueGreen.yaml的label的app相一致
    version: v1.3
  type: ClusterIP
kubectl apply -f blueGreenService.yaml
kubectl get svc,deploy,pods -n test

5.7 测试访问

while sleep 1;do curl 10.98.73.127;echo "";done

 

 

5.8 切换到blue版本（v2.0）

修改blueGreenService.yaml文件

apiVersion: v1
kind: Service
metadata:
  name: bluegreen
  namespace: test
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8080
  selector:
    app: bluegreen
    version: v2.0
  type: ClusterIP
kubectl apply -f blueGreenService.yaml
kubectl get svc,deploy,pods -n test

 

 

可以看出，如果要在蓝绿版本之间切换，只要修改blueGreenService.yaml文件的版本号即可。

6 金丝雀发布

在蓝绿版本发布的基础上，修改blueGreenService.yaml

apiVersion: v1
kind: Service
metadata:
  name: bluegreen
  namespace: test
spec:
  ports:
    - port: 80
      protocol: TCP
      targetPort: 8080
  selector:
    app: bluegreen
  type: ClusterIP

相比之前的yaml主要是去掉selector选择器下的version整个KV键值对

kubectl  apply -f blueGreenService.yaml 

接着测试：

 

 

这个时候，blue版本和green版本都可以提供服务，如果要设置权重，则控制好blue版本和green版本的副本数量即可实现。

转自：https://zhuanlan.zhihu.com/p/617025075