» Docker Builder

Type: docker

The docker Packer builder builds Docker images using Docker. The builder starts a Docker container, runs provisioners within this container, then exports the container for reuse or commits the image.

Packer builds Docker containers without the use of Dockerfiles. By not using Dockerfiles, Packer is able to provision containers with portable scripts or configuration management systems that are not tied to Docker in any way. It also has a simple mental model: you provision containers much the same way you provision a normal virtualized or dedicated server. For more information, read the section on Dockerfiles.

The Docker builder must run on a machine that has Docker Engine installed. Therefore the builder only works on machines that support Docker and does not support running on a Docker remote host. You can learn about what platforms Docker supports and how to install onto them in the Docker documentation.

 Please note: Packer does not yet have support for Windows containers.

» Basic Example: Export

Below is a fully functioning example. It doesn't do anything useful, since no provisioners are defined, but it will effectively repackage an image.

{
  "type": "docker",
  "image": "ubuntu",
  "export_path": "image.tar"
}

» Basic Example: Commit

Below is another example, the same as above but instead of exporting the running container, this one commits the container to an image. The image can then be more easily tagged, pushed, etc.

{
  "type": "docker",
  "image": "ubuntu",
  "commit": true
}

» Basic Example: Changes to Metadata

Below is an example using the changes argument of the builder. This feature allows the source images metadata to be changed when committed back into the Docker environment. It is derived from the docker commit --change command line option to Docker.

Example uses of all of the options, assuming one is building an NGINX image from ubuntu as an simple example:

{
  "type": "docker",
  "image": "ubuntu",
  "commit": true,
  "changes": [
    "USER www-data",
    "WORKDIR /var/www",
    "ENV HOSTNAME www.example.com",
    "VOLUME /test1 /test2",
    "EXPOSE 80 443",
    "LABEL version=1.0",
    "ONBUILD RUN date",
    "CMD [\"nginx\", \"-g\", \"daemon off;\"]",
    "ENTRYPOINT /var/www/start.sh"
  ]
}

Allowed metadata fields that can be changed are:

  • CMD
    • String, supports both array (escaped) and string form
    • EX: "CMD [\"nginx\", \"-g\", \"daemon off;\"]" corresponds to Docker exec form
    • EX: "CMD nginx -g daemon off;" corresponds to Docker shell form, invokes a command shell first
  • ENTRYPOINT
    • String, supports both array (escaped) and string form
    • EX: "ENTRYPOINT [\"/bin/sh\", \"-c\", \"/var/www/start.sh\"]" corresponds to Docker exec form
    • EX: "ENTRYPOINT /var/www/start.sh" corresponds to Docker shell form, invokes a command shell first
  • ENV
    • String, note there is no equal sign:
    • EX: "ENV HOSTNAME www.example.com" not "ENV HOSTNAME=www.example.com"
  • EXPOSE
    • String, space separated ports
    • EX: "EXPOSE 80 443"
  • LABEL
    • String, space separated key=value pairs
    • EX: "LABEL version=1.0"
  • ONBUILD
    • String
    • EX: "ONBUILD RUN date"
  • MAINTAINER
    • String, deprecated in Docker version 1.13.0
    • EX: "MAINTAINER NAME"
  • USER
    • String
    • EX: "USER USERNAME"
  • VOLUME
    • String
    • EX: "VOLUME FROM TO"
  • WORKDIR
    • String
    • EX: "WORKDIR PATH"

» Configuration Reference

Configuration options are organized below into two categories: required and optional. Within each category, the available options are alphabetized and described.

The Docker builder uses a special Docker communicator and will not use the standard communicators.

» Required:

You must specify (only) one of commit, discard, or export_path.

  • commit (bool) - If true, the container will be committed to an image rather than exported.

  • discard (bool) - Throw away the container when the build is complete. This is useful for the artifice post-processor.

  • export_path (string) - The path where the final container will be exported as a tar file.

  • image (string) - The base image for the Docker container that will be started. This image will be pulled from the Docker registry if it doesn't already exist.

  • message (string) - Set a message for the commit.

» Optional:

  • aws_access_key (string) - The AWS access key used to communicate with AWS. Learn how to set this.

  • aws_secret_key (string) - The AWS secret key used to communicate with AWS. Learn how to set this.

  • aws_token (string) - The AWS access token to use. This is different from the access key and secret key. If you're not sure what this is, then you probably don't need it. This will also be read from the AWS_SESSION_TOKEN environmental variable.

  • aws_profile (string) - The AWS shared credentials profile used to communicate with AWS. Learn how to set this.

  • author (string) - Set the author (e-mail) of a commit.

  • changes ([]string) - Dockerfile instructions to add to the commit. Example of instructions are CMD, ENTRYPOINT, ENV, and EXPOSE. Example: [ "USER ubuntu", "WORKDIR /app", "EXPOSE 8080" ]

  • container_dir (string) - The directory inside container to mount temp directory from host server for work file provisioner. This defaults to c:/packer-files on windows and /packer-files on other systems.

  • exec_user (string) - Username (UID) to run remote commands with. You can also set the group name/ID if you want: (UID or UID:GID). You may need this if you get permission errors trying to run the shell or other provisioners.

  • privileged (bool) - If true, run the docker container with the --privileged flag. This defaults to false if not set.

  • pull (bool) - If true, the configured image will be pulled using docker pull prior to use. Otherwise, it is assumed the image already exists and can be used. This defaults to true if not set.

  • run_command ([]string) - An array of arguments to pass to docker run in order to run the container. By default this is set to ["-d", "-i", "-t", "--entrypoint=/bin/sh", "--", "{{.Image}}"] if you are using a linux container, and ["-d", "-i", "-t", "--entrypoint=powershell", "--", "{{.Image}}"] if you are running a windows container. {{.Image}} is a template variable that corresponds to the image template option. Passing the entrypoint option this way will make it the default entrypoint of the resulting image, so running docker run -it --rm will start the docker image from the /bin/sh shell interpreter; you could run a script or another shell by running docker run -it --rm -c /bin/bash. If your docker image embeds a binary intended to be run often, you should consider changing the default entrypoint to point to it.

  • volumes (map[string]string) - A mapping of additional volumes to mount into this container. The key of the object is the host path, the value is the container path.

  • fix_upload_owner (bool) - If true, files uploaded to the container will be owned by the user the container is running as. If false, the owner will depend on the version of docker installed in the system. Defaults to true.

  • windows_container (bool) - If "true", tells Packer that you are building a Windows container running on a windows host. This is necessary for building Windows containers, because our normal docker bindings do not work for them.

  • login (bool) - This is used to login to dockerhub to pull a private base container. For pushing to dockerhub, see the docker post-processors

  • login_password (string) - The password to use to authenticate to login.

  • login_server (string) - The server address to login to.

  • login_username (string) - The username to use to authenticate to login.

  • ecr_login (bool) - Defaults to false. If true, the builder will login in order to pull the image from Amazon EC2 Container Registry (ECR). The builder only logs in for the duration of the pull. If true login_server is required and login, login_username, and login_password will be ignored. For more information see the section on ECR.

» Using the Artifact: Export

Once the tar artifact has been generated, you will likely want to import, tag, and push it to a container repository. Packer can do this for you automatically with the docker-import and docker-push post-processors.

Note: This section is covering how to use an artifact that has been exported. More specifically, if you set export_path in your configuration. If you set commit, see the next section.

The example below shows a full configuration that would import and push the created image. This is accomplished using a sequence definition (a collection of post-processors that are treated as as single pipeline, see Post-Processors for more information):

{
  "post-processors": [
    [
      {
        "type": "docker-import",
        "repository": "hashicorp/packer",
        "tag": "0.7"
      },
      "docker-push"
    ]
  ]
}

In the above example, the result of each builder is passed through the defined sequence of post-processors starting first with the docker-import post-processor which will import the artifact as a docker image. The resulting docker image is then passed on to the docker-push post-processor which handles pushing the image to a container repository.

If you want to do this manually, however, perhaps from a script, you can import the image using the process below:

$ docker import - registry.mydomain.com/mycontainer:latest < artifact.tar

You can then add additional tags and push the image as usual with docker tag and docker push, respectively.

» Using the Artifact: Committed

If you committed your container to an image, you probably want to tag, save, push, etc. Packer can do this automatically for you. An example is shown below which tags and pushes an image. This is accomplished using a sequence definition (a collection of post-processors that are treated as as single pipeline, see Post-Processors for more information):

{
  "post-processors": [
    [
      {
        "type": "docker-tag",
        "repository": "hashicorp/packer",
        "tag": "0.7"
      },
      "docker-push"
    ]
  ]
}

In the above example, the result of each builder is passed through the defined sequence of post-processors starting first with the docker-tag post-processor which tags the committed image with the supplied repository and tag information. Once tagged, the resulting artifact is then passed on to the docker-push post-processor which handles pushing the image to a container repository.

Going a step further, if you wanted to tag and push an image to multiple container repositories, this could be accomplished by defining two, nearly-identical sequence definitions, as demonstrated by the example below:

{
  "post-processors": [
    [
      {
        "type": "docker-tag",
        "repository": "hashicorp/packer1",
        "tag": "0.7"
      },
      "docker-push"
    ],
    [
      {
        "type": "docker-tag",
        "repository": "hashicorp/packer2",
        "tag": "0.7"
      },
      "docker-push"
    ]
  ]
}

» Docker For Windows

You should be able to run docker builds against both linux and Windows containers. Windows containers use a different communicator than linux containers, because Windows containers cannot use docker cp.

If you are building a Windows container, you must set the template option "windows_container": true. Please note that docker cannot export Windows containers, so you must either commit or discard them.

The following is a fully functional template for building a Windows container.

{
    "builders": [
    {
      "type": "docker",
      "image": "microsoft/windowsservercore:1709",
      "container_dir": "c:/app",
      "windows_container": true,
      "commit": true
    }
  ]
}

» Amazon EC2 Container Registry

Packer can tag and push images for use in Amazon EC2 Container Registry. The post processors work as described above and example configuration properties are shown below:

{
  "post-processors": [
    [
      {
        "type": "docker-tag",
        "repository": "12345.dkr.ecr.us-east-1.amazonaws.com/packer",
        "tag": "0.7"
      },
      {
        "type": "docker-push",
        "ecr_login": true,
        "aws_access_key": "YOUR KEY HERE",
        "aws_secret_key": "YOUR SECRET KEY HERE",
        "login_server": "https://12345.dkr.ecr.us-east-1.amazonaws.com/"
      }
    ]
  ]
}

Learn how to set Amazon AWS credentials.

» Dockerfiles

This builder allows you to build Docker images without Dockerfiles.

With this builder, you can repeatedly create Docker images without the use of a Dockerfile. You don't need to know the syntax or semantics of Dockerfiles. Instead, you can just provide shell scripts, Chef recipes, Puppet manifests, etc. to provision your Docker container just like you would a regular virtualized or dedicated machine.

While Docker has many features, Packer views Docker simply as an container runner. To that end, Packer is able to repeatedly build these containers using portable provisioning scripts.

» Overriding the host directory

By default, Packer creates a temporary folder under your home directory, and uses that to stage files for uploading into the container. If you would like to change the path to this temporary folder, you can set the PACKER_TMP_DIR. This can be useful, for example, if you have your home directory permissions set up to disallow access from the docker daemon.