A Buildkite plugin for running pipeline steps in Docker containers.

Also see the Docker Compose Buildkite Plugin which supports building images, docker-compose.yml, multiple containers, and overriding many of Docker’s defaults.

The following pipeline will build a binary in the dist directory using golang Docker image and then uploaded as an artifact.

steps:
  - command: "go build -o dist/my-app ."
    artifact_paths: "./dist/my-app"
    plugins:
      - docker#v5.12.0:
          image: "golang:1.11"

Windows images are also supported:

steps:
  - command: "dotnet publish -c Release -o published"
    plugins:
      - docker#v5.12.0:
          image: "microsoft/dotnet:latest"
          always-pull: true

⚠️ Warning: you should be careful when using an array or multi-line string as the command at the step level with this plugin. You will need to ensure that each line finishes with ;, execute a script in your repository or use the plugin's command option instead. You will also have to take into account the image's entrypoint and shell option.

If you want to control how your command is passed to the docker container, you can use the command parameter on the plugin directly:

steps:
  - plugins:
      - docker#v5.12.0:
          image: "mesosphere/aws-cli"
          always-pull: true
          command: ["s3", "sync", "s3://my-bucket/dist/", "/app/dist"]
    artifact_paths: "dist/**"

You can pass in additional environment variables and customize what is mounted into the container.

Note: If you are utilizing Buildkite's Elastic CI Stack S3 Secrets plugin, you must specify the environment variable key names as they appear in your S3 bucket's environment hook in order to access the secret from within your container.

steps:
  - command: "yarn install; yarn run test"
    plugins:
      - docker#v5.12.0:
          image: "node:7"
          always-pull: true
          environment:
            - "MY_SECRET_KEY"
            - "MY_SPECIAL_BUT_PUBLIC_VALUE=kittens"

Environment variables available in the step can also automatically be propagated to the container:

Note: this will not automatically propagate Elastic CI Stack S3 Secrets plugin environment variables. Refer above for explicitly importing values from that plugin.

steps:
  - command: "yarn install; yarn run test"
    env:
      MY_SPECIAL_BUT_PUBLIC_VALUE: kittens
    plugins:
      - docker#v5.12.0:
          image: "node:7"
          always-pull: true
          propagate-environment: true

AWS authentication tokens can be automatically propagated to the container, for example from an assume role plugin or ECS IAM role:

steps:
  - command: "yarn install; yarn run test"
    env:
      MY_SPECIAL_BUT_PUBLIC_VALUE: kittens
    plugins:
      - docker#v5.12.0:
          image: "node:7"
          always-pull: true
          propagate-aws-auth-tokens: true

You can pass in additional volumes to be mounted. This is useful for running Docker:

steps:
  - command: "docker build . -t image:tag; docker push image:tag"
    plugins:
      - docker#v5.12.0:
          image: "docker:latest"
          always-pull: true
          volumes:
            - "/var/run/docker.sock:/var/run/docker.sock"

You can disable the default behaviour of mounting in the checkout to workdir:

steps:
  - command: "npm start"
    plugins:
      - docker#v5.12.0:
          image: "node:7"
          always-pull: true
          mount-checkout: false

Variable interpolation can be tricky due to the 3 layers involved (Buildkite, agent VM, and docker). For example, if you want to use ECR Buildkite plugin, you will need to use the following syntax. Note the $$ prefix for variables that would otherwise resolve at pipeline upload time, not runtime:

steps:
  - command: "yarn install; yarn run test"
    plugins:
      - ecr#v2.7.0:
          login: true
          account_ids:
          - "d"
          - "p"
          region: us-west-2
          no-include-email: true
      - docker#v5.12.0:
          image: "d.dkr.ecr.us-west-2.amazonaws.com/imagename"
          command: ["./run-integration-tests.sh"]
          expand-volume-vars: true
          volumes:
            - "/var/run/docker.sock:/var/run/docker.sock"
            - "$$BUILDKITE_DOCKER_CONFIG_TEMP_DIRECTORY/config.json:/root/.docker/config.json"
          propagate-environment: true
          environment:
            - "BUILDKITE_DOCKER_CONFIG_TEMP_DIRECTORY"

You can read more about runtime variable interpolation from the docs.

If the image that you want to run is not in a registry the load property can be used to load an image from a tar file.

This can be useful if a previous step builds an image and uploads it as an artifact. The below example shows how a artifact can be downloaded and then passed to the load property to load the image which can then be referred to in the image property to start a container and run a command as explained above.

steps:
  - command: "npm start"
    plugins:
      - artifacts#v1.9.0:
          download: "node-7-image.tar.gz"
      - docker#v5.12.0:
          load: "node-7-image.tar.gz"
          image: "node:7"

You need to be careful when/if running the BuildKite agent itself in docker that, itself, runs pipelines that use this plugin. Make sure to read all the documentation on the matter, specially the caveats and warnings listed.

The name of the Docker image to use.

Example: node:7

Additional lines can be added to /etc/hosts in the container, in an array of mappings. See https://docs.docker.com/engine/reference/run/#managing-etchosts for more details.

Example: buildkite.fake:123.0.0.7

Additional groups to be added to the user in the container, in an array of group names (or ids). See https://docs.docker.com/engine/reference/run/#additional-groups for more details.

Example: docker

Whether to always pull the latest image before running the command. Useful if the image has a latest tag.

Default: false

Sets the command for the Docker image, and defaults the shell option to false. Useful if the Docker image has an entrypoint, or doesn't contain a shell.

This option can't be used if your step already has a top-level, non-plugin command option present.

Examples: [ "/bin/mycommand", "-c", "test" ], ["arg1", "arg2"]

Enables debug mode, which outputs the full Docker commands that will be run on the agent machine.

Default: false

Limit read rate from a device (format: <device-path>:<number>[<unit>]). Number is a positive integer. Unit can be one of kb, mb, or gb.

Example: ["/dev/sda1:200mb"]

Limit read rate (IO per second) from a device (format: <device-path>:<number>). Number is a positive integer.

Example: ["/dev/sda1:400"]

Limit write rate to a device (format: <device-path>:<number>[<unit>]). Number is a positive integer. Unit can be one of kb, mb, or gb.

Example: ["/dev/sda1:200mb"]

Limit write rate (IO per second) to a device (format: <device-path>:<number>). Number is a positive integer.

Example: ["/dev/sda1:400"]

Override the image’s default entrypoint, and defaults the shell option to false. See the docker run --entrypoint documentation for more details. Set it to "" (empty string) to disable the default entrypoint for the image, but note that you may need to use this plugin's command option instead of the top-level command option or set a shell instead (depending on the command you want/need to run - see Issue 138 for more information).

Example: /my/custom/entrypoint.sh, ""

An array of additional environment variables to pass into to the docker container. Items can be specified as either KEY or KEY=value. Each entry corresponds to a Docker CLI --env parameter. Values specified as variable names will be passed through from the outer environment.

Example: [ "BUILDKITE_MESSAGE", "MY_SECRET_KEY", "MY_SPECIAL_BUT_PUBLIC_VALUE=kittens" ]

An array of additional files to pass into to the docker container as environment variables. Each entry corresponds to a Docker CLI --env-file parameter.

If you set this to VALUE, and VALUE is an environment variable containing a space-separated list of environment variables such as A B C D, then A, B, C, and D will all be propagated to the container. This is helpful when you've set up an environment hook to export secrets as environment variables, and you'd also like to programmatically ensure that secrets get propagated to containers, instead of listing them all out.

When set to true, it will activate interpolation of variables in the elements of the image configuration variable. When turned off (the default), attempting to use variables will fail as the literal $VARIABLE_NAME string will be passed as the image name.

Environment variable interporation rules apply here. $VARIABLE_NAME is resolved at pipeline upload time, whereas $$VARIABLE_NAME is at run time. All things being equal, you likely want $$VARIABLE_NAME.

⚠️ Important: this is considered an unsafe option as the most compatible way to achieve this is to run the strings through eval which could lead to arbitrary code execution or information leaking if you don't have complete control of the pipeline

Whether or not to automatically propagate all* pipeline environment variables into the docker container. Avoiding the need to be specified with environment.

Note that only pipeline variables will automatically be propagated (what you see in the Buildkite UI). Variables set in proceeding hook scripts will not be propagated to the container.

* Caveat: only environment variables listed in $BUILDKITE_ENV_FILE will be propagated. This does not include e.g. variables that you exported in an environment hook. If you wish for those to be propagated, try env-propagation-list.

Whether or not to automatically propagate aws authentication environment variables into the docker container. Avoiding the need to be specified with environment. This is useful for example if you are using an assume role plugin or you want to pass the role of an agent running in ECS or EKS to the docker container.

Will propagate AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN, AWS_REGION, AWS_DEFAULT_REGION, AWS_STS_REGIONAL_ENDPOINTS, AWS_WEB_IDENTITY_TOKEN_FILE, AWS_ROLE_ARN, AWS_CONTAINER_CREDENTIALS_FULL_URI, AWS_CONTAINER_CREDENTIALS_RELATIVE_URI, and AWS_CONTAINER_AUTHORIZATION_TOKEN, only if they are set already.

When the AWS_WEB_IDENTITY_TOKEN_FILE is specified, it will also mount it automatically for you and make it usable within the container.

Whether or not to automatically propagate gcp auth credentials into the docker container. Avoiding the need to be specified with environment. This is useful if you are using a workload identity federation to impersonate a service account and you want to pass it to the docker container. This is compatible with the gcp-workload-identity-federation plugin.

Will propagate GOOGLE_APPLICATION_CREDENTIALS, CLOUDSDK_AUTH_CREDENTIAL_FILE_OVERRIDE and BUILDKITE_OIDC_TMPDIR and also mount the dir specified by BUILDKITE_OIDC_TMPDIR into the container.

Whether to match the user ID and group ID for the container user to the user ID and group ID for the host user. It is similar to specifying user: 1000:1000, except it avoids hardcoding a particular user/group ID.

Using this option ensures that any files created on shared mounts from within the container will be accessible to the host user. It is otherwise common to accidentally create root-owned files that the agent will be unable to remove, since a lot of images by default run as the root user. This can also be mitigated for the checkout directory by using the chown option of this plugin.

Whether or not to run the container in privileged mode

Whether or not to run an init process inside the container. This ensures that responsibilities like reaping zombie processes are performed inside the container.

See Docker's documentation for background and implementation details.

Default: true for Linux and macOS, false for Windows.

Whether or not to leave the container after the run, or immediately remove it with --rm.

Default: false

Specify a file to load a docker image from. If omitted no load will be done.

Whether to automatically mount the current working directory which contains your checked out codebase. Mounts onto /workdir, unless workdir is set, in which case that will be used.

If there's a git mirror path and mount-checkout is enabled, the (mirror path)[https://buildkite.com/docs/pipelines/environment-variables#BUILDKITE_REPO_MIRROR] is mounted into the docker container as an added volume. Otherwise, the git mirror path will have to be explicitly added as an extra volume to mount into the container.

Default: true

Whether to chown the directory mounted with mount-checkout to the Buildkite agent user before exiting, to ensure that the agent can clean up any additional files created there. This will happen even if the command fails or is cancelled. This option has no effect on Windows or if mount-checkout is false.

Prefer using propagate-uid-gid over this option, as the chown–which can take some time if your checkout is of considerable size—is likely not needed at all in that case.

Default: false

The Docker image in which to run the chown command.

Default: busybox

Whether to automatically mount the buildkite-agent binary from the host agent machine into the container.

Set to true if you want to enable and are sure that the binary running in the agent is compatible with the container's architecture and environment (for example, don't try to mount the OS X or Windows agent binary in a container running linux). If enabled in Windows agents your pipeline, step or agent must have the BUILDKITE_AGENT_BINARY_PATH environment variable defined with the executable to mount in the (Windows) agent.

Default: false

Important: enabling this option will share the BUILDKITE_AGENT_TOKEN and BUILDKITE_AGENT_JOB_API_TOKEN environment variables (and other related ones) with the container if present.

Whether to mount the ssh-agent socket (at /ssh-agent) from the host agent machine into the container or not. Instead of just true or false, you can specify absolute path in the container for the home directory of the user used to run on which the agent's .ssh/known_hosts will be mounted (by default, /root).

Default: false

Important: note that for this to work you will need the agent itself to have access to an ssh agent that is: up and running, listening on the appropriate socket, with the appropriate credentials loaded (or to be loaded). Please refer to the agent's documentation on using SSH agent for more information

Join the container to the docker network specified. The network will be created if it does not already exist. See https://docs.docker.com/engine/reference/run/#network-settings for more details.

Example: test-network

Platform defines the target platform containers for this service will run on, using the os[/arch[/variant]] syntax. The values of os, arch, and variant MUST conform to the convention used by the OCI Image Spec.

Example: linux/arm64

PID namespace provides separation of processes. The PID Namespace removes the view of the system processes, and allows process ids to be reused including pid 1. See https://docs.docker.com/engine/reference/run/#pid-settings---pid for more details. By default, all containers have the PID namespace enabled.

Example: host

GPUs selector. Dependencies: nvidia-container-runtime

Example: all

A number of times to retry failed docker pull. Defaults to 3. Only applies when always-pull is enabled.

Specify an explicit docker runtime. See the docker run options documentation for more details.

Example: nvidia

Specify the IPC mode to use. See the docker run options documentation for more details.

Example: host

If set to true, adds useful Docker labels to the container. See Container Labels for more info.

The default is true.

Set the shell to use for the command. Set it to false to pass the command directly to the docker run command. The default is ["/bin/sh", "-e", "-c"] unless you have provided an entrypoint or command.

Example: [ "powershell", "-Command" ]

Set the size of the /dev/shm shared memory filesystem mount inside the docker contianer. If unset, uses the default for the platform (typically 64mb). See docker run’s runtime constraints documentation for information on allowed formats.

Example: 2gb

Whether to skip the repository checkout phase.

Default: false

This allows setting the container rootfs size at the creation time. This is only available for the devicemapper, btrfs, overlay2, windowsfilter and zfs graph drivers. See docker documentation for more details.

Example: size=120G

If set to false, doesn't allocate a TTY. This is useful in some situations where TTY's aren't supported, for instance windows.

Default: true for Linux and macOS, and false for Windows.

If set to false, doesn't connect stdin to the process. Some scripts fall back to asking for user input in case of errors, if the process has stdin connected and this results in the process waiting for input indefinitely.

Allows a user to be set, and override the USER entry in the Dockerfile. See https://docs.docker.com/engine/reference/run/#user for more details.

Example: root

Allows to explicitly set the user namespace. This overrides the default docker daemon value. If you use the value host, you disable user namespaces for this run. See https://docs.docker.com/engine/security/userns-remap/ for more details. Due to limitations in this feature, a privileged container will override the user specified userns value to host

Example: mynamespace

Extra volume mounts to pass to the docker container, in an array. Items are specified as SOURCE:TARGET. Each entry corresponds to a Docker CLI --volume parameter. Relative local paths are converted to their full-path (e.g ./code:/app).

Example: [ "/var/run/docker.sock:/var/run/docker.sock" ]

When set to true, it will activate interpolation of variables in the elements of the volumes configuration array as well as workdir. When turned off (the default), attempting to use variables will fail as the literal $VARIABLE_NAME string will be passed to the -v option.

Environment variable interporation rules apply here. $VARIABLE_NAME is resolved at pipeline upload time, whereas $$VARIABLE_NAME is at run time. All things being equal, you likely want $$VARIABLE_NAME.

⚠️ Important: this is considered an unsafe option as the most compatible way to achieve this is to run the strings through eval which could lead to arbitrary code execution or information leaking if you don't have complete control of the pipeline

Tmpfs mounts to pass to the docker container, in an array. Each entry corresponds to a Docker CLI --tmpfs parameter. See Docker's tmpfs mounts documentation for more information on this feature.

Example: [ "/tmp", "/root/.cache" ]

The working directory to run the command in, inside the container. The default is /workdir. This path is also used by mount-checkout to determine where to mount the checkout in the container.

Example: /app

Set namespaced kernel parameters in the container. More information can be found in https://docs.docker.com/engine/reference/commandline/run/.

Example: --sysctl net.ipv4.ip_forward=1

Add Linux capabilities to the container. Each entry corresponds to a Docker CLI --cap-add parameter.

Remove Linux capabilities from the container. Each entry corresponds to a Docker CLI --cap-drop parameter.

Add security options to the container. Each entry corresponds to a Docker CLI --security-opt parameter.

Add ulimit options to the container. Each entry corresponds to a Docker CLI --ulimit parameter.

You can give builds limited access to a specific device or devices by passing devices to the docker container, in an array. Items are specific as SOURCE:TARGET or just TARGET. Each entry corresponds to a Docker CLI --device parameter.

Example: [ "/dev/bus/usb/001/001" ]

You can allow the docker container to publish ports. More information can be found in https://docs.docker.com/config/containers/container-networking/. Each entry corresponds to a Docker CLI --publish or -p parameter.

Example: [ "8080:80" ] (Map TCP port 80 in the container to port 8080 on the Docker host.)

Set the CPU limit to apply when running the container. More information can be found in https://docs.docker.com/config/containers/resource_constraints/#cpu.

Example: 0.5

Set the memory limit to apply when running the container. More information can be found in https://docs.docker.com/config/containers/resource_constraints/#limit-a-containers-access-to-memory.

Example: 2g

Set the memory swap limit to apply when running the container. More information can be found in https://docs.docker.com/config/containers/resource_constraints/#limit-a-containers-access-to-memory.

Example: 2g

Set the swappiness level to apply when running the container. More information can be found in https://docs.docker.com/config/containers/resource_constraints/#--memory-swappiness-details.

Example: 0

When running a command, the plugin will automatically add the following Docker labels to the container:

• com.buildkite.pipeline_name=${BUILDKITE_PIPELINE_NAME}
• com.buildkite.pipeline_slug=${BUILDKITE_PIPELINE_SLUG}
• com.buildkite.build_number=${BUILDKITE_BUILD_NUMBER}
• com.buildkite.job_id=${BUILDKITE_JOB_ID}
• com.buildkite.job_label=${BUILDKITE_LABEL}
• com.buildkite.step_key=${BUILDKITE_STEP_KEY}
• com.buildkite.agent_name=${BUILDKITE_AGENT_NAME}
• com.buildkite.agent_id=${BUILDKITE_AGENT_ID}

These labels can make it easier to query containers on hosts using docker ps for example:

docker ps --filter "label=com.buildkite.job_label=Run tests"

This behaviour can be disabled with the run-labels: false option.

To run testing, shellchecks and plugin linting use use bk run with the Buildkite CLI.

bk run

Or if you want to run just the tests, you can use the docker Plugin Tester:

docker run --rm -ti -v "${PWD}":/plugin buildkite/plugin-tester:latest

MIT (see LICENSE)