Getting Started

Projects in Ansible Container are identified by their path in the filesystem. The project path contains all of the source and files for use inside of your project, as well as the Ansible Container build and orchestration instructions.

Conductor Container

The heavy lifting of Ansible Container happens within a special container, generated during the build process, called the Conductor.

It contains everything you need to build your target container images, including Ansible Core itself. Since Ansible requires a Python runtime on the hosts it is configuring, and so that you don’t need to install Python on those target container images you are building, the Python runtime and all library dependencies are mounted from the Conductor container into the target containers during builds.

Because of this, the Conductor container is built upon a base image of your choice, and it’s recommended that you select the same flavor of Linux distribution that your target containers will be built from. For example, if you’re building container images based on Alpine Linux, it’s a good idea to use Alpine Linux as your Conductor container base image as well, so that what the Conductor exports to your target containers contains apk and other binary dependencies you will likely need.

For more about how the Conductor container gets built, available pre-baked Conductor images, and how to build your own Conductor image, see The Conductor Container.

Dipping a Toe In - Starting from Scratch

Ansible Container provides a convenient way to start your app by simply running ansible-container init from within your project directory, which creates:


Other ansible-container subcommands enable the container development workflow:

  • ansible-container build initiates the build process. It builds and launches the Conductor Container. The Conductor then runs instances of your base container images as specified in container.yml. The Conductor container applies Ansible roles against them, committing each role as new image layers. Ansible communicates with the other containers through the container engine, not through SSH.
  • ansible-container run orchestrates containers from your built images together as described by the container.yml file. The container.yml can specify overrides to make development faster without having to rebuild images for every code change.
  • ansible-container deploy uploads your built images to a container registry of your choice and generates an Ansible playbook to orchestrate containers from your built images in production container platforms, like Kubernetes or Red Hat OpenShift.

So what goes into the files that make this work?


The container.yml file is a file in YAML-syntax that describes the services in your project, how to build and run them, the repositories to push them to, and more.

The container.yml file is similar to the other multi-container orchestration formats, like Docker Compose or OpenCompose. Much like these other formats, this file describes the orchestration of your app. Ansible Container uses this file to determine what images to build, what containers to run and connect, and what images to push to your repository. Additionally, when Ansible Container generates an Ansible playbook to ship and orchestrate your images in the cloud, this file describes the configuration target for your entire container infrastructure. It is automatically run, but it also saves the playbook for you to examine or reuse.

By way of an example, consider the below container.yml file:

version: "2"
    from: "centos:7"
      - common
      - apache
      - "80:80"
    command: ["/usr/bin/dumb-init", "/usr/sbin/apache2ctl", "-D", "FOREGROUND"]
        - "DEBUG=1"

Things to note:

  1. In this example the schema is set to version 2. (Version 2 support was add in version 0.3.0.)
  2. Each of the containers you wish to orchestrate should be under the services key.
  3. For supported service keys, see Container.yml Specification.
  4. The image you specify should be the base image that your containers will start from. Ansible Container will use your roles to build upon this base image. Each role you specify needs to be in a roles/ directory in your project, in your requirements.yml file, or in the –roles-path you specify at runtime in the command line.
  5. You may optionally specify a dev_overrides section. During build and in generating the Ansible roles to deploy your app to the cloud, this section will be ignored. However, when running your containers locally for your development environment, you may use this section to override settings from your production configuration. For instance, a Javascript developer may wish to use Gulp and BrowserSync to dynamically rebuild assets while development is taking place, versus rebuilding the entire container for every code change. Thus that developer may wish to include dev_overrides that run a BrowserSync server for those assets, whereas in production Gulp would build those assets and exit.


You can share your project on Ansible Galaxy for others to use as a template for building projects of their own. These templates are called “Container Apps”. Provide the requested information in meta.yml, and then log into Galaxy to import it into the Ansible Container project template registry.


Running Ansible inside of the Conductor container may have Python library dependencies that your modules require. Use the ansible-requirements.txt file to specify those dependencies. This file follows the standard pip format for Python dependencies. When your Conductor container image is created, these dependencies are installed.


If the roles in your container.yml file are in Ansible Galaxy or a remote SCM repository, and your project depends upon them, add them to requirements.yml. For more information about requirements.yml see Installing Roles From a File.


Set Ansible configuration settings within the build container. For more information see Configuration File. Do note that overriding some of the settings, like roles_path, might have unexpected results, due to Ansible using the Conductor container as its execution environment.


During a build, it is commonplace to synchronize the source from your project into one of your target containers. However many such files are not appropriate to be part of that build context, such as your .git directory. The Docker Engine would ignore files or patterns contained in the .dockerignore file from the build context and for conformity, Ansible Container will do the same.

Hello World

Every good walkthrough needs a “Hello, world,” right? So let’s start there, and make a Flask web service.

Writing the code

$ mkdir hello-world
$ cd hello-world
$ ansible-container init
Ansible Container initialized.

Now that we have the skeleton of an empty project, let’s make our Flask app. First, we need to list our Python requirements. So let’s create a file called requirements.txt that lists the Python packages we need:


And let’s write the Hello, world, Flask app code, in a file called

from flask import Flask
app = Flask(__name__)

def hello_world():
    return 'Hello, World!'

Next, let’s write the Ansible role that will be used to build our container. First, we have to create a roles/ directory and create the skeleton of a new Ansible role using the ansible-galaxy utility (make sure you have Ansible installed to get this utility):

$ mkdir roles
$ cd roles
$ ansible-galaxy init flask
- flask was created successfully

We can put the steps to build our container in tasks/main.yml.

First, we grab dumb-init from GitHub. In Linux systems, the process with process id (PID) 1 is special, in that it is expected to react differently to different system signals. The dumb-init utility very simply handles those signals properly and then spawns the real command you want your container to run.

Second, we need to install the Python package management utility pip, which in the CentOS world is in EPEL (Extra Packages for Enterprise Linux). So our next tasks set up our container to look into EPEL for its packages, from which it can find Pip.

Next, we need to create an unprivileged user to run our application. Running a container as root is a Bad Idea™, and enterprise-grade Kubernetes platforms like Red Hat OpenShift will refuse to execute a container that tries to run as root. However, we can keep our user in the root group without the same security impact.

Using the syncrhonize plugin, we can copy our project source into the container to use. Our Ansible Container project path is staged in the Conductor container at /src, so synchronize can find it there. This content will have already been filtered by the .dockerignore file in our project, and by default that .dockerignore file includes sensible things to ignore like .git/.

Finally, we want to use the pip utility we installed to grab the contents of the requirements.txt file to install Python dependencies.

- name: Install dumb init
    dest: /usr/bin/dumb-init
    mode: 0775
    validate_certs: no
- name: Install epel
    name: epel-release
    state: present
    disable_gpg_check: yes
- name: Install pip
    name: python2-pip
    state: present
    disable_gpg_check: yes
- name: Create flask user
    name: flask
    uid: 1000
    group: root
    state: present
    createhome: yes
    home: "/app"
- name: Copy source into container
    src: "/src/"
    dest: "/app"
  remote_user: flask
- name: Install Python dependencies
    requirements: /app/requirements.txt

Having written the role’s tasks, we’re almost ready to build. We still have to describe how our application will be built and run in the container.yml file.

As generated by ansible-container init, the boilerplate container.yml file contains a lot of commented out hints about what it can contain. Our example below contains just the significant parts of the container.yml, not the commented parts, so feel free either to read the comments in the boilerplate before copy-and-pasting our example into your text editor or to make line by line edits. Your choice.

The container.yml file encompasses all of your services and containers, all of your container registries, all of your persistent storage, and all of your overrides of Ansible Container runtime defaults.

For our Ansible Container settings, we need to identify the Conductor base image that we’re going to use; this should be the distribution that our target images are also ultimately based on. Since we’re going to build a CentOS 7 container image for Flask, we need the CentOS 7 base. And since we’re going to deploy our app into Kubernetes, we should configure the Kubernetes namespace (also the OpenShift project) in which our application will be deployed.

For our services, we have only the one: our Flask service. We must state its base image and the list of roles from which it will be built. And then we also specify its runtime configuration, much like one might do in a Docker Compose file.

Registries is required to be present, even if we don’t have a private registry we’re going to deploy to just yet.

Our container.yml thus looks like this:

version: "2"
    base: centos:7
  project_name: hello-world
     name: hello-world
     description: Simple Demo
     display_name: Hello, World
    from: centos:7
      - flask
    entrypoint: /usr/bin/dumb-init
    command: gunicorn -w 4 -b helloworld:app
      - 4000:4000
    working_dir: /app
    user: flask
registries: {}

That’s it - we’re all set to start building and running.

Building the service

To kick off a build, all you do is run:

$ ansible-container build

But let’s talk about what happens when you do. First, Ansible Container builds its Conductor container. This may involve pulling the Conductor base image from the Docker Hub registry, but once it has that base image it makes a custom Conductor image for your project.

Then, Ansible Container starts up this Conductor container, which in turn tells the container engine to start up instances of each service’s base image to use in the build. From there, the Conductor runs Ansible, applying each of the roles to the target containers, committing new image layers for each role.

Ansible Container also maintains a build cache. If you were to run the build again, Ansible Container would recognize that your service configuration and the flask role hadn’t changed and would exit quickly. It will only rebuild the layers of your image that it has detected likely to have changed. You can always disable the build cache with --no-cache when you build.

But back to our build. When we run, ansible-container build, we see some output like this:

Building Docker Engine context...
Starting Docker build of Ansible Container Conductor image (please be patient)...
Parsing conductor CLI args.
Copying build context into Conductor container.
Docker™ daemon integration engine loaded. Build starting.       project=hello-world
Building service...     project=hello-world service=flask
Fingerprint for this layer: 60ed99196f031a470d2dfbe39e9af02fb934b7d328a0e3f494f11ba1072e878e    parent_fingerprint=ccfe8c87363124f4f46aa60b2d727e06366c81d2a3d552672faa027e1cee144d parent_image_id=<Image: 'centos:7'> role=flask service=flask
Cached layer for for role flask not found or invalid.   cur_image_id=<Image: 'centos:7'> fingerprint=60ed99196f031a470d2dfbe39e9af02fb934b7d328a0e3f494f11ba1072e878e service=flask
Could not locate intermediate build container to reapply role flask. Applying role on image <Image: 'centos:7'> as container hello-world_flask-ccfe8c87-flask.  cur_image_fingerprint=ccfe8c87363124f4f46aa60b2d727e06366c81d2a3d552672faa027e1cee144d service=flask

PLAY [flask] *******************************************************************

TASK [Gathering Facts] *********************************************************
ok: [flask]

TASK [flask : Install dumb init] ***********************************************
changed: [flask]

TASK [flask : Install epel] ****************************************************
changed: [flask]

TASK [flask : Install pip] *****************************************************
changed: [flask]

TASK [flask : Create flask user] ***********************************************
changed: [flask]

TASK [flask : Copy source into container] **************************************
changed: [flask]

TASK [flask : Install Python dependencies] *************************************
changed: [flask]

PLAY RECAP *********************************************************************
flask                      : ok=7    changed=6    unreachable=0    failed=0

Applied role to service role=flask service=flask
Committed layer as image        fingerprint=60ed99196f031a470d2dfbe39e9af02fb934b7d328a0e3f494f11ba1072e878e image=sha256:5367a04394b09a145dc9d3b6302ae89cb5f0aeab9fe61589632c4775cc8bac94 role=flask service=flask
Build complete. service=flask
Cleaning up stale build artifacts.      service=flask
All images successfully built.
Conductor terminated. Cleaning up.      command_rc=0 conductor_id=54f566e42f800de2c00842fbac32e25c13d863a86b351f38c695aba27ed0604c save_container=False

Looking through it, we can see the steps it took: building the Conductor, the build cache looking at known fingerprints, the Ansible playbook executing, the and container image layer being committed. If you run docker images you can see that you’ve got new images for your project’s Conductor and the service it built. Ansible Container will use the build timestamp for the version label.

And if we inspect the image we’ve build, we can see it is the centos:7 image plus a single additional layer (note that the specific hash ID’s will almost certainly be different for your installation):

$ docker inspect --format '{{.RootFS.Layers}}' centos:7
$ docker inspect --format '{{.RootFS.Layers}}' hello-world-flask
[sha256:e15afa4858b655f8a5da4c4a41e05b908229f6fab8543434db79207478511ff7 sha256:f40f4bccee05c1406fa7812d2aa33345796306bd31f562470582410fdcb0e488]

Now we can go on ahead and run the project.

Running the service

To run your project, Ansible Container spins up the Conductor once more. Behind the scenes, it uses your container.yml to generate an Ansible playbook to orchestrate your application in the container platform. Then it executes this playbook. So if we were to run our project, we might see:

$ ansible-container run
Parsing conductor CLI args.
Copying build context into Conductor container.
Engine integration loaded. Preparing run.   engine=Docker™ daemon
Verifying service image     service=flask

PLAY [Deploy hello-world] ******************************************************

TASK [docker_service] **********************************************************
changed: [localhost]

PLAY RECAP *********************************************************************
localhost                  : ok=1    changed=1    unreachable=0    failed=0

All services running.       playbook_rc=0
Conductor terminated. Cleaning up.  command_rc=0 conductor_id=d043ef69b3a71d919cfcdcac959e635dc18deb8020ba3c98706b7979f2f5678f save_container=False

Let’s verify that the service is actually running by asking Docker and hitting the Flask server:

$ docker ps
CONTAINER ID        IMAGE                              COMMAND                  CREATED             STATUS              PORTS                    NAMES
8a23d7571c05        hello-world-flask:20180127130321   "/usr/bin/dumb-init …"   14 seconds ago      Up 12 seconds>4000/tcp   helloworld_flask_1
$ docker logs helloworld_flask_1
[2018-01-27 13:04:18 +0000] [7] [INFO] Starting gunicorn 19.7.1
[2018-01-27 13:04:18 +0000] [7] [INFO] Listening at: (7)
[2018-01-27 13:04:18 +0000] [7] [INFO] Using worker: sync
[2018-01-27 13:04:18 +0000] [12] [INFO] Booting worker with pid: 12
[2018-01-27 13:04:18 +0000] [13] [INFO] Booting worker with pid: 13
[2018-01-27 13:04:18 +0000] [14] [INFO] Booting worker with pid: 14
[2018-01-27 13:04:18 +0000] [19] [INFO] Booting worker with pid: 19
$ curl
Hello, World!

Voila! Up and running.

Deploying to Kubernetes

If you recall, when we ran our project above, the Conductor container generated an Ansible playbook to orchestrate the application in the container engine. In Ansible Container, the deploy command pushes your build images to a registry and then writes the deployment playbook to disk for you to execute when you’re ready.

Ansible Container is designed with pluggable backends, called engines. So far, we have been using the engine for the Docker Engine. But Ansible Container supports other engines, and we’re actively working on new ones. To date, we support the Docker Engine, Kubernetes, and Red Hat OpenShift. Each of these requires additional dependencies be installed, as described in the install instructions.

For this walkthrough, the easiest thing to do is to install the open-source upstream for Red Hat OpenShift, an enterprise-grade distribution of Kubernetes, called OpenShift Origin. OpenShift Origin offers a conveniently packaged version for developer systems as minishift. We provide an easy Ansible role for getting Minishift set up here.

Once you’ve got Minishift going on your system, we use the deploy command to generate the deployment playbook:

$ ansible-container --engine openshift deploy
Parsing conductor CLI args.
Copying build context into Conductor container.
Engine integration loaded. Preparing deploy.        engine=OpenShift™
Verifying image for flask
  config file = /etc/ansible/ansible.cfg
  configured module search path = [u'/root/.ansible/plugins/modules', u'/usr/share/ansible/plugins/modules']
  ansible python module location = /usr/lib/python2.7/site-packages/ansible
  executable location = /usr/bin/ansible-galaxy
  python version = 2.7.5 (default, Aug  4 2017, 00:39:18) [GCC 4.8.5 20150623 (Red Hat 4.8.5-16)]
Using /etc/ansible/ansible.cfg as config file
Opened /root/.ansible_galaxy
Processing role ansible.kubernetes-modules
Opened /root/.ansible_galaxy
- downloading role 'kubernetes-modules', owned by ansible
- downloading role from
- extracting ansible.kubernetes-modules to /home/joeschmoe/devel/hello-world/ansible-deployment/roles/ansible.kubernetes-modules
- ansible.kubernetes-modules (v0.3.1-6) was installed successfully
Conductor terminated. Cleaning up.  command_rc=0 conductor_id=c078b981a08206aa09ce413aec4eb19c6b785c1cab73dd2a239d016d8cd263bb save_container=False

The generated output has been written out in our project path to a new directory ansible-deployment`, with the playbook ``ansible-deployment/hello-world.yml.

The playbook contains tagged tasks for each of the lifecycle management stages of your project: start, stop, restart, and destroy. To execute, change directory to the ansible-deployment directory and run:

$ ansible-playbook hello-world.yml --tags=start

You can then go to the console of your local Minishift instance and see your project at work. Run minishift console to bring up the web console, log in using the default credentials (username and password both developer), and look at the Hello World project.