Ansible and Google Compute Engine

I’ve discovered Ansible only last week and I’m already a huge fan. Why? It’s much simpler that what I’ve seen with Chef or Puppet. If really fits with my way of working.

When I prepare a machine manually, I type a list of commands in a sequential manner. With Ansible, I simply translate this list of commands into a playbook. A playbook is a list of tasks. Something like that:

- hosts: gce
  sudo: yes
    - shell: creates=/usr/share/doc/python-apt apt-get update & apt-get install -y python-apt
    - apt: pkg=python-pip
    - apt: pkg=python-pycurl
    - apt: pkg=nginx
    - service: name=nginx state=started
    - apt_repository: repo='deb trusty main' state=present
    - apt_repository: repo='deb-src trusty main' state=present
    - apt_key: data="{{ lookup('file', 'files/java/webupd8.key.asc') }}" state=present
    - shell: echo debconf shared/accepted-oracle-license-v1-1 select true | debconf-set-selections
    - apt: pkg=oracle-java8-installer update_cache=yes state=installed
    - apt: pkg=maven
    - apt: pkg=psmisc
    - apt_key: url=
    - apt_repository: repo='deb dist 10gen' state=present
    - apt: pkg=mongodb-10gen=2.2.3 update_cache=yes
    - service: name=mongodb state=started

When these tasks work well, you can refactor the playbook a little bit by forming groups of tasks. Then you can extract theses tasks into named roles. One baby step at a time. Ansible enables this way of working step by step. It’s not too strict about the way you organise your project. It’s up to you to stop when you think it’s good enough.

If you want to prepare a Google Compute Engine instance with Ansible, it’s very easy. Create the instance with the Api or the Console. Grab the IP and put it into a hosts file and run a playbook. There’s no agent to install on the instance before running Ansible. That’s nice.

However, because GCE uses its own set of ssh keys to connect to an instance, here’s the kind of ansible.cfg configuration file you have to write before you get started:

remote_user = YOUR_GCE_USERNAME

ssh_args = -o UserKnownHostsFile=/dev/null -o CheckHostIP=no -o StrictHostKeyChecking=no -i PATH_TO/google_compute_engine

Have fun!

9 thoughts on “Ansible and Google Compute Engine”

  1. Hello,

    Instead of specifying apt: pkg … every time, you can group them using ‘with_items’ construct, such as, e.g.:

    – name: ensure apt cache is up to date
    action: apt update_cache=yes
    – name: ensure packages are installed
    action: apt name={{item}}
    – postgresql
    – libpq-dev
    – python-psycopg2
    – …

    The example above is taken from and you can find very well commented examples at

      1. By the way, I’ve recently discovered Ansible Galaxy at and I think Ansible is converging to the idea of having a CANN (Comprehensive Ansible Archive Network, that is similar to CPAN, CTAN, CRAN, etc.) This will let Ansible users to really streamline their DevOps development I believe.

  2. Hi,
    What do you think about using Ansible to install apps (like ear or war) instead of provisionning a whole system ? I’m looking for replace a bunch of Ops shell scripts by something more “structured” and reusable.
    Ansible is one of tool I want to evaluate and I would appreciate your opinion.

    1. From what I’ve experienced so far, Ansible can be as lightweight as you want, in other words, I think it’s perfectly suitable for installing Java applications (.war and/or .ear), too. Moreover, if you have some form of CI (Continuous Integration) infrastructure set up (e.g. via utilizing Jenkins), combining that with Ansible could prove to be good solution for your declarative automation needs.

  3. David, while technically your playbook is correct, please use `name:` parameter with your actions. That will make playbook execution much more visual and will make things much easier to troubleshoot.

Comments are closed.