Container Wrapper

The Container Wrapper extension allows running a build step inside the specified Docker/LXC image. Images are pulled via docker pull or podman pull commands, depending on which container manager is installed on the agent that runs the build.

TeamCity can pull containers anonymously (if images are publicly available) or after logging into a registry (for private registries or to avoid DockerHub penalties for anonymous downloads). If you need TeamCity to authorize to a registry before pulling an image, configure the Docker Support build feature, as follows:

In your project settings, select Connections from the sidebar and follow the instructions in Configuring Connections to Docker to add new Docker or Podman connections to your project.
In your build configuration settings, follow the instructions in Docker Support to configure the Docker Support build feature, adding the connections created in the previous step.

The extension is available for the following build runners:

Container Settings

In the Container Settings section of the build step settings, you can specify an image which will be used to run the build step. Once the image is specified, the following options become available.

Setting	Description
Run step within container	The image name as stated in Docker Hub or other registry. TeamCity will start a container from the specified image and will try to run this build step within this container. For example, `ruby:2.4` will run the step within the Ruby container, version 2.4. If an agent that runs the build has Podman installed instead of Docker, use either full image names (for example, `docker.io/library/alpine:latest` instead of `alpine:latest`), or ensure the registry domain is specified in the registries.conf file on your build agent machine. See also: How to manage Linux container registries.
Image platform	Select <Any> (default), Linux, or Windows. Note that Windows images are not supported by Podman.
Pull image explicitly	If enabled, the image will be pulled from the Hub repository via `docker pull <imageName>`/`podman pull <imageName` before the `docker run`/`podman run` command is launched.
Additional run arguments	Allows specifying additional options for the `docker run` and `podman run` commands. The default argument is `--rm`, but you can provide more, for instance, add an additional volume mapping. If you intend to utilize environment variables in this field (for example, `%env.FOO%`), note that TeamCity passes to containers only those variables that are declared in build configurations and projects. Agent-specific environment variables declared in the buildAgent.properties file are not passed to containers. If you need a parameter declared in this file, use this approach: Define a system property (`system.FOO=BAR`) in buildAgent.properties. Add a build configuration parameter with the name `system.FOO` and value `%system.FOO%`. Reference the system property as `%system.FOO%` in the "Additional run arguments" field. If you need to use the same value as an environment variable in your build script, you can also add a build configuration parameter with the name `env.FOO` and value `%system.FOO%`.

How Container Wrapper Works

Technically, the command of the build runner is wrapped in a shell script, and this script is executed inside a container with the docker run or podman run command. To view the details about the started process, text of the script, and so on, check the build log in Verbose mode.

The Container Wrapper maps paths to the build checkout directory and other agent directories like buildAgent/work, so that all these directories have the same location on a build agent and inside a Container Wrapper.

If the process environment contains the TEAMCITY_DOCKER_NETWORK environment variable set by the previous Docker Compose build step, this network is passed to the started docker run command with the --network switch.

Restoring File Ownership on Linux

Build agents that use Docker execute the chown command at the end of each step running inside a container to restore the buildAgent user's permissions to access the checkout directory. This action prevents potential issues related to build agents being unable to remove no longer needed container files that were created with the root ownership. Agents using Podman do not perform this step.

By default, a TeamCity agent uses the busybox image from Docker Hub to run the chown command. You can specify an alternative image name with the teamcity.internal.docker.busybox parameter, either in the buildAgent.properties file or in the build configuration parameters.

Environment Variables Handling

TeamCity passes environment variables from the build configuration into the Docker or Podman process, but it does not pass environment variables from the build agent, as they may not be relevant to the container environment. The list of the passed environment variables can be seen in the Verbose mode in the build log.

Setting Image Entrypoint

If a Docker image does not define an ENTRYPOINT, you can use still run a container with an ENTRYPOINT from the command line:

Add a Command Line build step.
Set the Run mode to Executable with parameters.
In the Command executable field, specify the full path to the ENTRYPOINT in the target container.
In Docker Settings, specify the name of the container.

TeamCity will start the specified Docker image with the defined ENTRYPOINT.

Setting Container User

If your step creates or accesses files or folders on local storage, ensure these actions are performed under the correct user with sufficient permissions. To do this, add --user=<value> to Additional run arguments of the runner.

steps {
    script {
        // ...
        dockerImage = "python:windowsservercore-ltsc2022"
        dockerRunParameters = "--user=1001"
    }
}

The host UID can be retrieved via the env.UID parameter (--user=%env.UID%).

Last modified: 18 September 2024