image_build

Table of Contents

1. Module description - What is the image_build module, and what does it do?
2. Setup - The basics of getting started with image_build
3. Usage - How to build Docker containers with Puppet
   • A hello world example
   • Involving hiera
   • Building multiple images from one manifest
   • Using a Puppet Master
   • Minimizing image size with Rocker
   • Building ACI images
4. Reference - Sample help output from the tool
5. Limitations - OS compatibility, etc.
6. Maintainers - who maintains this project

Module description

The basic purpose of image_build is to enable building various images, including Docker images, from Puppet code. There are two main cases where this can be useful:

1. You have an existing Puppet codebase and you're moving some of your services to using containers. By sharing the same code between container and non-container based infrastructure you can cut down on duplication of effort, and take advantage of work you've already done.
2. You're building a lot of images, but scaling Dockerfile means either a complex hierachy of images or copy-and-pasting snippets between many individual Dockerfiles. image_build allows for sharing common functionality as Puppet modules, and Puppet itself provides a rich domain-specific language for declarative composition of images.

Setup

puppetlabs/image_build is a Puppet Module and is available on the Forge.

The following should work in most cases:

puppet module install puppetlabs/image_build

You don't need any additional gems installed unless you are looking to work on developing the module. All you need is a working Docker environment or acbuild, for which I'd recommend Docker for Mac or Docker for Windows or just installing Docker if you're on Linux. For acbuild you can use the rkt module.

Usage

With the module installed you should have access to two new puppet commands; puppet docker and puppet aci. These have two subcommands, one will trigger a build of an image, the other can be used to output the intermediary dockerfile or shell script.

The examples directory contains a set of examples for experimenting with. Simply open up examples/nginx and run:

puppet docker build

The above is the simplest example of a build. Some settings are provided in the accompanying metadata.yaml file, while others are defaults specific to the tool. You can change values in the metadata file (useful for version control) or you can override those values on the command line.

puppet docker build --image-name puppet/sample --cmd nginx --expose 80

See the full help page for other arguments for specifying different base images, setting a maintainer, using Rocker instead of Docker for the build and much more.

puppet docker build --help

You can also output the intermediary dockerfile using another subcommand. This is useful for both debugging and if you want to do something not natively supported by the tool.

puppet docker dockerfile

A hello world example - Nginx

Lets see a simple hello world example. We'll create a Docker image running Nginx and serving a simple text file.

First lets use a few Puppet modules from the Forge. We'll use the existing nginx module and we'll specify it's dependencies. We're also using dummy_service to ignore service resources in the Nginx module.

$ cat Puppetfile
forge 'https://forgeapi.puppetlabs.com'

mod 'puppet/nginx'
mod 'puppetlabs/stdlib'
mod 'puppetlabs/concat'
mod 'puppetlabs/apt'
mod 'puppetlabs/dummy_service'

Then lets write a simple manifest. Disabling nginx daemon mode isn't supported by the module yet so we drop a file in place. Have a look at manifests/init.pp:

include 'dummy_service'

class { 'nginx': }

nginx::resource::vhost { 'default':
  www_root => '/var/www/html',
}

file { '/var/www/html/index.html':
  ensure  => present,
  content => 'Hello Puppet and Docker',
}

exec { 'Disable Nginx daemon mode':
  path    => '/bin',
  command => 'echo "daemon off;" >> /etc/nginx/nginx.conf',
  unless  => 'grep "daemon off" /etc/nginx/nginx.conf',
}

And finally lets store the metadata in a file rather than pass on the command line. Take a look at metadata.yaml:

cmd: nginx
expose: 80
image_name: puppet/nginx

Now lets build a Docker image. Note that you'll need docker available on your host to do so, along with the image_build module installed.

puppet docker build

And finally lets run our new image. We expose the webserver on port 8080 to the local host.

$ docker run -d -p 8080:80 puppet/nginx
83d5fbe370e84d424c71c1c038ad1f5892fec579d28b9905cd1e379f9b89e36d
$ curl http://0.0.0.0:8080
Hello Puppet and Docker%

Involving hiera - Elasticsearch

The Elasticsearch example is similar to the above, with a few additional features demonstrated. In particular the use of Hiera to provide additional context for the Puppet build. You can find this in the examples/elasticsearch directory.

puppet docker build manifests/init.pp --image-name puppet/es --expose 9200 --cmd /docker-entrypoint.sh

A note on options with multiple arguments

Several of the arguments to image_build can take a list of values. This is done by passing in comma separated values. For instance, to specify an ENTRYPOINT like so:

ENTRYPOINT ["nginx", "-g", "daemon off"]

You can pass the following on the commandline:

--entrypoint nginx,'-g','daemon off'

Building multiple images from one manifest

One advantage of using Puppet for building Docker images is you are removed from the need to have a single Dockerfile per image. Meaning a single repository of Puppet code can be used to describe multiple images. This makes ensuring all images use (for example) the same repositories or same hardening scripts much easier to enforce. Change code in one place and rebuild multiple images.

Describing multiple images in Puppet is done using the existing node resource in your manifest. For instance:

node 'node1' {
  webserver { 'hello node 1': }
}

node 'node2' {
  webserver { 'hello node 2': }
}

You can then select which image to build when running the build command, by explicitly passing the image-name.

puppet docker build --image-name puppet/node1

The match for the node resource in the Puppet code is done without the repository name, in this case the puppet/ before node1.

Note that you may want different metadata for different images. image_build will attempt to detect additional metadata in the metadata folder, and will merge items from metadata/metadata.yaml with node specific metadata, for instance from metadata/node1.yaml

You can see an example of this settup in the examples/multi directory.

Using a Puppet Master

The above examples all use local manifests copied to the image during build, but image_build also supports using a Puppet Master. You can provide metadata via a local metadata file or directory, or by passing command line arguments to the build command as shown in the examples above. The only change is passing --master like so.

puppet docker dockerfile --master puppet.example.com --image-name puppet/node1 --expose 80 --cmd nginx

The hostname passed to the Puppet Master will take the form node1.{datetime}.dockerbuilder. This means you can match on that pattern in your manifests, for instance like so:

node /^node1/ {
  webserver { 'hello node 1': }
}

A worked example is provided in the examples/master folder. You can either upload this to an existing Puppet Master or Puppet Enterprise install, or run a new local master using Docker.

First install the dependent modules into the local environment:

r10k puppetfile install --moduledir code/environments/production/modules

Create an autosign.conf file with the following:

*.dockerbuilder.*

Then, from the examples/master folder, use Docker to run an instance of Puppet Server:

docker run --name puppet -P --hostname puppet -v $(pwd)/code:/etc/puppetlabs/code -v $(pwd)/autosign.conf:/etc/puppetlabs/puppet/autosign.conf puppet/puppetserver-standalone

Determine the port on which the Puppet Server is exposed locally:

docker port puppet

You'll also need the IP address of your local machine. Replace the {ip} and {port} in the following with your own values.

puppet docker dockerfile --master {ip}:{port} --image-name puppet/node1 --expose 80 --cmd nginx

This should use the code on the Puppet Master to build the image.

Minimizing image size with Rocker

image_build supports using the Rocker build tool in place of the standard Docker build command. The Rocker output provides a little more detail about the build process, but also allows for mounting of folders at build time which minimizes the size of the resulting image.

puppet docker build --rocker

Note that when using Rocker the Puppet tools are not left in the final image, reducing it's file size.

Building ACI images

As well as Docker support, image_build also experimentally supports building ACI compatible images for use with Rkt or other supported runtimes. This works in the same manner as above. The following command should generate a shell script which, when run, generates an ACI:

puppet aci script

And if you simply want to build the ACI directly you can just run:

puppet aci build

Reference

$ puppet docker --help
USAGE: puppet docker <action> [--from STRING]
[--maintainer STRING]
[--os STRING]
[--os-version STRING]
[--puppet-agent-version STRING]
[--r10k-version STRING]
[--module-path PATH]
[--expose STRING]
[--cmd STRING]
[--entrypoint STRING]
[--labels KEY=VALUE]
[--rocker]
[--[no-]inventory]
[--hiera-config STRING]
[--hiera-data STRING]
[--image-user STRING]
[--puppetfile STRING]
[--image-name STRING]
[--config-file STRING]
[--config-directory STRING]
[--master STRING]

Build Docker images and Dockerfiles using Puppet code

OPTIONS:
  --render-as FORMAT             - The rendering format to use.
  --verbose                      - Whether to log verbosely.
  --debug                        - Whether to log debug information.
  --cmd STRING                   - The default command to be executed by the
                                   resulting image
  --config-directory STRING      - A folder where metadata can be loaded from
  --config-file STRING           - A configuration file with all the metadata
  --entrypoint STRING            - The default entrypoint for the resulting
                                   image
  --expose STRING                - A list of ports to be exposed by the
                                   resulting image
  --from STRING                  - The base docker image to use for the
                                   resulting image
  --hiera-config STRING          - Hiera config file to use
  --hiera-data STRING            - Hieradata directory to use
  --image-name STRING            - The name of the resulting image
  --image-user STRING            - Specify a user to be used to run the
                                   container process
  --[no-]inventory               - Enable or disable the generation of an
                                   inventory file at /inventory.json
  --labels KEY=VALUE             - A set of labels to be applied to the
                                   resulting image
  --maintainer STRING            - Name and email address for the maintainer of
                                   the resulting image
  --master STRING                - A Puppet Master to use for building images
  --module-path PATH             - A path to a directory containing a set of
                                   modules to be copied into the image
  --network STRING               - The Docker network to pass along to the
                                   docker build command
  --os STRING                    - The operating system used by the image if not
                                   autodetected
  --os-version STRING            - The version of the operating system used by
                                   the image if not autodetected
  --puppet-agent-version STRING  - Version of the Puppet Agent package to
                                   install
  --puppet-debug                 - Pass the debug flag to the Puppet process
                                   used to build the container image
  --puppetfile STRING            - Enable use of Puppetfile to install
                                   dependencies during build
  --r10k-version STRING          - Version of R10k to use for installing modules
                                   from Puppetfile
  --rocker                       - Use Rocker as the build tool
  --[no-]show-diff               - Enable or disable showing the diff when
                                   running Puppet to build the image
  --skip-puppet-install          - If the base image already contains Puppet we
                                   can skip installing it
  --volume STRING                - A list of volumes to be added to the
                                   resulting image

ACTIONS:
  build         Build a Docker image from Puppet code
  dockerfile    Generate a Dockerfile which will run the specified Puppet code

See 'puppet man docker' or 'man puppet-docker' for full help.

Limitations

The module currently does not support building Windows containers, or building containers from a Windows machine. We'll be adding support for these in the future.

The inventory functionality does not work correctly on Centos 6 based images, so if you're using Centos 6 then you need to pass the --no-inventory flag.

Maintainers

This repository is maintained by: Gareth Rushgrove gareth@puppet.com.