咨询：VSTS中能否通过指定Docker镜像运行各阶段构建任务？

Absolutely! VSTS (now part of Azure DevOps) fully supports running your build stages inside custom Docker containers—this feature is called Container Jobs, and it works just like the image field in GitLab CI that you’re familiar with. You won’t need to set up local VMs or install agents manually; instead, Azure DevOps will use its hosted Linux agents to pull your specified image, spin up a container, and run all your build steps inside it.

How to configure it in Azure DevOps YAML #

In GitLab CI you define the base image with the top-level image field or per stage. In Azure DevOps, you use the container keyword at either the pipeline level (for all jobs) or the job level (for individual stages/jobs). Here’s how you’d translate your GitLab example to Azure DevOps YAML:

trigger:
- main

stages:
- stage: Deploy
  jobs:
  - job: DeployJob
    # Specify the Docker image for this job/stage
    container: monachus/hugo:v0.30.2
    steps:
    - script: git submodule init
      displayName: 'Initialize Git submodules'
    - script: git submodule update --force
      displayName: 'Update Git submodules'
    # Add your deployment steps here

- stage: Cert
  jobs:
  - job: CertJob
    # You can use the same image or a different one for this stage
    container: monachus/hugo:v0.30.2
    steps:
    # Add your certificate management steps here

Key details to note: #

• Per-job or pipeline-wide containers: If all your stages use the same image, you can define the container at the top of the pipeline instead of per job.
• Hosted agents: Azure DevOps’ hosted Linux agents handle all the container setup—they pull the image, start the container, and execute your steps inside it. No local infrastructure required.
• Private registries: If your image is stored in a private Docker registry, you can add authentication by linking a service connection:

  container:
    image: your-private-registry/hugo:v0.30.2
    containerRegistry: YourPrivateRegistryServiceConnection
  

This setup matches exactly what you’re looking for: each build stage runs in an isolated Docker container, using your preferred image, without needing to manage local VMs or agents.

内容的提问来源于stack exchange，提问作者r1ckr