تحرير

مشاركة عبر

Container jobs in YAML pipelines

  • مقالة

Azure DevOps Services | Azure DevOps Server 2022 - Azure DevOps Server 2019

This article explains container jobs in Azure Pipelines.

By default, Azure Pipelines jobs run directly on the host machines where the agent is installed. Hosted agent jobs are convenient, require little initial setup and infrastructure to maintain, and are well-suited for basic projects.

If you want more control over task context, you can define and run jobs in containers. Containers are a lightweight abstraction over the host operating system that provides isolation from the host. When you run jobs in containers, you can select the exact versions of operating systems, tools, and dependencies that your build requires.

Linux and Windows agents can run pipeline jobs directly on the host or in containers. Container jobs aren't available on macOS.

For a container job, the agent first fetches and starts the container. Then each step of the job runs inside the container.

If you need fine-grained control at the individual build step level, step targets let you choose a container or host for each step.

Requirements and limitations

The following requirements and limitations apply to Azure Pipelines agent hosts or containers.

Agent hosts

  • Only windows-* and ubuntu-* images support running containers. The macos-* images don't support running containers.
  • To run containers, Windows and Linux agent hosts must have Docker installed, and must have permission to access the Docker daemon.
  • Containers aren't supported when the agent is already running inside a container. You can't have nested containers.

Containers

Linux-based containers have the following requirements. For workarounds, see Nonglibc-based containers.

  • Bash installed
  • GNU C Library (glibc)-based
  • No ENTRYPOINT
  • Provide USER with access to groupadd and other privileged commands without using sudo
  • Can run Node.js, which the agent provides

    Note

    Node.js must be pre-installed for Linux containers on Windows hosts.

Some stripped-down containers available on Docker Hub, especially containers based on Alpine Linux, don't satisfy these requirements. Containers with an ENTRYPOINT might not work because Azure Pipelines docker create and docker exec expect that the container is always up and running.

Single job examples

The following examples define a Windows or Linux container for a single job.

The following simple example defines a Linux container:

pool:
  vmImage: 'ubuntu-latest'

container: ubuntu:18.04

steps:
- script: printenv

The preceding example tells the system to fetch the ubuntu image tagged 18.04 from Docker Hub and then start the container. The printenv command runs inside the ubuntu:18.04 container.

Multiple jobs

You can use containers to run the same step in multiple jobs. The following example runs the same step in multiple versions of Ubuntu Linux. You don't have to mention the jobs keyword because only a single job is defined.

pool:
  vmImage: 'ubuntu-latest'

strategy:
  matrix:
    ubuntu16:
      containerImage: ubuntu:16.04
    ubuntu18:
      containerImage: ubuntu:18.04
    ubuntu20:
      containerImage: ubuntu:20.04

container: $[ variables['containerImage'] ]

steps:
- script: printenv

Multiple jobs with agent pools on a single agent host

A container job uses the underlying host agent's Docker configuration file for image registry authorization. This file signs out at the end of the Docker registry container initialization. Registry image pulls for subsequent container jobs might be denied for unauthorized authentication because another job running in parallel already signed out the Docker configuration file.

The solution is to set a Docker environment variable DOCKER_CONFIG that's specific to each agent pool running on the hosted agent. Export the DOCKER_CONFIG in each agent pool's runsvc.sh script as follows:

export DOCKER_CONFIG=./.docker

Startup options

You can specify options to control container startup, as in the following example:

container:
  image: ubuntu:18.04
  options: --hostname container-test --ip 192.168.0.1

steps:
- script: echo hello

Running docker create --help gives you the list of options that you can pass to Docker invocation. Not all of these options are guaranteed to work with Azure DevOps. Check first to see if you can use a container property to accomplish the same goal.

For more information, see the docker create command reference and the resources.containers.container definition in the Azure DevOps YAML schema reference.

Reusable container definition

The following example defines the containers in the resources section, and then references them by their assigned aliases. The jobs keyword is explicitly listed for clarity.

resources:
  containers:
  - container: u16
    image: ubuntu:16.04

  - container: u18
    image: ubuntu:18.04

  - container: u20
    image: ubuntu:20.04

jobs:
- job: RunInContainer
  pool:
    vmImage: 'ubuntu-latest'

  strategy:
    matrix:
      ubuntu16:
        containerResource: u16
      ubuntu18:
        containerResource: u18
      ubuntu20:
        containerResource: u20

  container: $[ variables['containerResource'] ]

  steps:
  - script: printenv

Service endpoints

You can host containers on other registries than public Docker Hub. To host an image on Azure Container Registry or another private container registry, including a private Docker Hub registry, add a service connection to access the registry. Then you can reference the endpoint in the container definition.

Private Docker Hub connection:

container:
  image: registry:ubuntu1804
  endpoint: private_dockerhub_connection

Azure Container Registry connection:

container:
  image: myprivate.azurecr.io/windowsservercore:1803
  endpoint: my_acr_connection

Note

Azure Pipelines can't set up a service connection for Amazon Elastic Container Registry (ECR), because Amazon ECR requires other client tools to convert AWS credentials into something Docker can use to authenticate.

Nonglibc-based containers

The Azure Pipelines agent supplies a copy of Node.js, which is required to run tasks and scripts. To find out the version of Node.js for a hosted agent, see Microsoft-hosted agents.

The version of Node.js compiles against the C runtime used in the hosted cloud, typically glibc. Some Linux variants use other C runtimes. For instance, Alpine Linux uses musl.

If you want to use a nonglibc-based container, you need to:

  • Supply your own copy of Node.js.
  • Add a label to your image telling the agent where to find the Node.js binary.
  • Provide other dependencies that Azure Pipelines depends on: bash, sudo, which, and groupadd.

Supply your own Node.js

If you use a nonglibc-based container, you're responsible for adding a Node binary to your container. Node.js 18 is a safe choice. Start from the node:18-alpine image.

Tell the agent about Node.js

The agent reads the container label "com.azure.dev.pipelines.handler.node.path". If this label exists, it must be the path to the Node.js binary.

For example, in an image based on node:18-alpine, add the following line to your Dockerfile:

LABEL "com.azure.dev.pipelines.agent.handler.node.path"="/usr/local/bin/node"

Add required packages

Azure Pipelines assumes a Bash-based system with common administrative packages installed. Alpine Linux in particular doesn't come with several of the packages needed. Install bash, sudo, and shadow to cover the basic needs.

RUN apk add bash sudo shadow

If you depend on any in-box or Marketplace tasks, also supply the binaries they require.

Full Dockerfile example

FROM node:18-alpine

RUN apk add --no-cache --virtual .pipeline-deps readline linux-pam \
  && apk add bash sudo shadow \
  && apk del .pipeline-deps

LABEL "com.azure.dev.pipelines.agent.handler.node.path"="/usr/local/bin/node"

CMD [ "node" ]

Related content