注：本文中主要討論 .NET6.0專案在 k8s 中運行的 Dapr 的持續集成流程, 但實際上不是Dapr的專案部署到K8s也是相同流程，只是k8s的yaml組態檔有所不同

流程選擇

基于 Dapr 的專案持續集成包含以下流程

1. 編譯并打包專案
2. 構建 Dockerfile,并推送鏡像push image至私有倉庫
3. 準備 k8s 部署的組態檔
4. 通過 kubectl 部署鏡像至 k8s 中

這里面有多種方案

鑒于以上優缺點，最終我選擇了第二種折衷方案，這種方案既不影響持續集成的速度，也不會產生過多的鏡像，只是在部署時會產生多余的鏡像構建時間，

專案結構

• 每個要發布的API的 project 檔案夾中增加以下檔案
  • dapr.yaml
  • Dockerfile

dapr.yaml

kind: Deployment
apiVersion: apps/v1
metadata:
  name: demo
  namespace: dapr-api
  labels:
    app: .api
    service: demo
spec:
  replicas: 1
  selector:
    matchLabels:
      service: demo
  template:
    metadata:
      labels:
        app: .api
        service: demo
      annotations:
        dapr.io/enabled: "true"
        dapr.io/app-id: "demo-api"
        dapr.io/app-port: "80"
        dapr.io/log-as-json: "true"
    spec:
      containers:
        - name: demo-api
          image: 倉庫地址/鏡像名:220310.13
          ports:
            - name: http
              containerPort: 80
              protocol: TCP
          imagePullPolicy: IfNotPresent
---
kind: Service
apiVersion: v1
metadata:
  name: demo-api
  namespace: dapr-api
  labels:
    app: .api
    service: demo
spec:
  type: NodePort
  selector:
    service: demo
  ports:
  - protocol: TCP
    port: 80
    targetPort: 80
    nodePort: 30004

Dockerfile

FROM mcr.microsoft.com/dotnet/aspnet:6.0 AS final
WORKDIR /app
EXPOSE 80
COPY ["./projectfolder", "/app"]
ENTRYPOINT ["dotnet", "projectdll.dll"]

這兩個檔案需要每個專案不同，后面在編譯和部署流程中會用到，

Pipelines 持續集成的組態檔

trigger:
  batch: true

pool:
  name: Default
name: $(Date:yy)$(Date:MM)$(Date:dd)$(Rev:.r)

variables:
    BuildConfiguration: 'Release'
steps:
- task: UseDotNet@2
  displayName: 'Check and Install .NET SDK 6.0'
  inputs:
    version: '6.0.x'
    includePreviewVersions: false

- task: DotNetCoreCLI@2
  displayName: 'Publish to zip'
  inputs:
    command: publish
    publishWebProjects: false
    projects: './src/projectfolder/project.csproj'
    arguments: '--configuration $(BuildConfiguration) --output $(build.artifactstagingdirectory) -v n'
    zipAfterPublish: false
    workingDirectory: '$(Build.SourcesDirectory)/src'
## 復制上文中的兩個檔案到 Artifact
- task: CopyFiles@2
  displayName: 'Copy dapr.yaml to: $(build.artifactstagingdirectory)'
  inputs:
    SourceFolder: './src/${{ parameters.project }}/'
    Contents: |
      Dockerfile
      dapr.yaml
    TargetFolder: '$(build.artifactstagingdirectory)'

- task: PublishBuildArtifacts@1
  displayName: 'Publish Artifact'
  inputs:
    PathtoPublish: '$(build.artifactstagingdirectory)'

Release 發布流程組態檔

發布流程新建兩個作業

作業1 Build Image

variables:
  image: '自定義鏡像名'

steps:
- task: Docker@2
  displayName: buildAndPush
  inputs:
    containerRegistry: harbor
    repository: '$(image)'
    Dockerfile: '$(System.DefaultWorkingDirectory)/_dapr-demo/drop/Dockerfile'
    tags: '$(Build.BuildNumber)'

作業2 KubeDeploy

variables:
  image: '自定義鏡像名,與上文須一致'

steps:
- task: KubernetesManifest@0
  displayName: deploy
  inputs:
    kubernetesServiceConnection: online
    namespace: '$(ns)' ## k8s的部署目標命名空間
    strategy: canary ## 灰度部署策略
    percentage: 50
    manifests: '$(System.DefaultWorkingDirectory)/_dapr-demo/drop/dapr.yaml'
    containers: '$(harborUrl)/$(image):$(Build.BuildNumber)'

這樣，在首次部署時執行全部管道，

后期回滾版本只，手動執行第二個管理即KubeDeploy即可

其它流程

本流程全部依賴 Azure DevOps 自身的配置，并不依賴 Agent 環境配置，如果依賴 Agent 環境的話有更多做法，

標籤：.NET Core

上一篇：MASA Blazor入門這一篇就夠了

下一篇：一個高頻問題：異步操作會創建執行緒嗎？