An Isolated and Reproducible Ansible and Vagrant Setup on Mac

Ansible is a great tool for automating system maintenance. This is how I install it. My notes differ from the official ones in the following ways:

  • I install as much of this software as I can isolated from the system tools. This means I can update my projects and the system code independently and I can uninstall my tools cleanly.
  • I've added notes for integrating VM creation with Vagrant/VirtualBox and Ansible. This trio really works nicely when you want to spin up a VM or network of VMs, do horrible dangerous things to them, and finally kill them when you mess up too badly. All of this information is stored in relatively portable files, so it's easy to make a base configuration for VMs then customize it for different projects.

Install Ansible

Ansible does not support Windows. These notes are specifically for Mac, though they should be trivial to modify for Linux. The Ansible docs suggests that we should use pip to install it. This means that we need to install Python to use its pip. Macs have an older version of python, so I like to install a new one with Anaconda. This has two purposes- it gets me a modern and easily uninstallable version of Python, and it will let me isolate my Ansible install from the rest of the system, making it easier to install multiple versions of Ansible and also uninstall it.

Install Anaconda Python

I use the 3.6 64-bit Command-Line Installer

You want to install it to the default location and prepend the install location to the PATH

Follow any notes at the end of the install.

If this is your first install of dbus, automatically load on login with:
    mkdir -p ~/Library/LaunchAgents
    cp /Users/bkane/anaconda3/org.freedesktop.dbus-session.plist ~/Library/LaunchAgents/
    launchctl load -w ~/Library/LaunchAgents/org.freedesktop.dbus-session.plist

Anaconda Python includes a lot more than what you need for Ansible (for example, type jupyter qtconsole to get a nice Python CLI).

Create an environment for Ansible

We could install Ansible globally, but we'll want to update and test different versions against each other. To that end, we'll make a separate environment for the current version of ansible.

  • First make a folder:
mkdir -p ~/Code/Python/ansible-2.4
cd !$
conda create --name ansible-2.4 python=3

Enter the environment

source activate ansible-2.4

You should see your prompt change and if you type which python, you'll see this environment has its own copy of Python.

Install ansible

pip install ansible==2.4

Test it:

ansible --version

Save the environment

Let's save the environment to a file. This will let you give this file another Mac user and they'll be able to reproduce your Ansible install. Very helpful for debugging.

conda env export -f environment.yaml

Exit the environment

When you're done playing with Ansible, type source deactivate to exit the environment.

Ansible is now installed. To actually play with it though, we can use the vagrant and VirtualBox tools to launch Linux VMs for us to experiment with and then kill when we screw them up too badly.

Install Virtualbox

Install Vagrant

Launch a small VM network

At this point, we've installed all the programs we need to. The rest of this is creating the right files and running commands.

mkdir -p ~/Code/Vagrant/small_network
cd !$

We'll be staying in this folder for all further commands and file creation.

Copy the following into a text file called ~/Code/Vagrant/small_network/Vagrantfile. If you don't have a text editor installed to do this, I recommend Visual Studio Code.

# -*- mode: ruby -*-
# vi: set ft=ruby :

Vagrant.configure('2') do |config|
  # set settings common to all VMs = 'centos/7'
  # Disable the default synced folder because it's too much trouble to set up
  config.vm.synced_folder '.', '/vagrant', disabled: true
  config.vm.synced_folder '.', '/home/vagrant/sync', disabled: true

  # Don't check the host key
  config.ssh.verify_host_key = false

  # Add some VirtualBox specific settings
  config.vm.provider 'virtualbox' do |vb|
    vb.gui = true
    vb.customize ['modifyvm', :id, '--clipboard', 'bidirectional']

  # Create and set VM specific settings
  config.vm.define :node1 do |node1|
    node1.vm.hostname = 'node1'

  config.vm.define :node2 do |node2|
    node2.vm.hostname = 'node2'

Feel free to read the docs, but we're really just using Vagrant to give us easily buildable and destroyable VMs.

With the Vagrantfile in place, we've built a small network of two nodes. Let's turn it on:

cd ~/Code/Vagrant/small_network
vagrant up

Test by SSHing into the two nodes:

vagrant ssh node1
exit  # exit from inside the VM
vagrant ssh node2
exit  # exit from inside the VM

Use Ansible with our VMs

Create a hosts.ini for Ansible

We need to tell Ansible about our VMs and how to connect to them. Copy the following into ~/Code/Vagrant/small_network/hosts.ini:


Tell Ansible to use Vagrant's SSH settings

If we tried to connect to node1 and node2 as-is from Ansible, Ansible won't be able to find them. To fix this we need to export Vagrant's SSH settings to an SSH config file and tell Ansible to use that file.

vagrant ssh-config > ssh_config

Next we need to connect that ssh_config to Ansible with an ansible.cfg file in the small_network directory. In this file, I've also enabled some other useful options. See the docs for more info:

host_key_checking = False
retry_files_enabled = False
inventory = hosts.ini
# This presents a window for a logged-in attacker,
# but it's a small window and I need what it enables
# See
allow_world_readable_tmpfiles = True


# generate ssh_config with `vagrant ssh-config`
ssh_args = -F ./ssh_config

And test it (make sure you are in ~/Code/Vagrant/small_network/ and are using your ansible environment (source activate ansible-2.4)):

ansible all -m ping

You should get something like the following:

(ansible-2.4) ✔ ~/Code/Vagrant/small_network
15:09 $ ansible -i hosts.yaml all -m ping
node1 | SUCCESS => {
    "changed": false,
    "failed": false,
    "ping": "pong"
node2 | SUCCESS => {
    "changed": false,
    "failed": false,
    "ping": "pong"

Manipulate the VMs

To turn off the VMs, type vagrant halt. To turn them back on again, type vagrant up

Further Steps

At this point we've installed Anaconda Python, Ansible (in it's own isolated environment), Virtualbox, and vagrant. We've taught Ansible how to find and login to the VMs we've spun up, and we're ready to start really learning Ansible. Head to the Getting Started and follow it. We've already done about the first 3rd, but we integrated Vagrant instead of modifying the global /etc/ansible/hosts file. The (See also) links at the bottom of the page are your next steps. Enjoy!

Uninstalling all of this

Eventually, you might stop working with Ansible, or find a way you like more, and you might want to uninstall everything or parts of this stack. Here's how, starting from the VMs down

Erase a VM

cd ~/Code/Vagrant/small_network/
vagrant destroy

Optionally delete ~/Code/Vagrant/small_network if you're absolutely sure you won't want to recreate your work.

Uninstall Vagrant

Follow the Uninstallation Page

Uninstall Virtualbox

Follow this guide. It basically recommends that you quit the app and run the uninstallation script.

Remove the virtual environmant

Run conda remove --name ansible-2.4 --all

Optionally delete ~/Code/Python/ansible-2.4 if you'll never need the environment again.

Remove Anaconda Python

Follow this StackOverflow link and remove the PATH manipulation in your ~/.bashrc or ~/.bash_profile.