Chris Hepner

Developer-Specific Vagrant Configuration: Part I

Jul 30, 2014 •

Introduction to Vagrant

In my day job, I maintain a Vagrant box. Vagrant, if you’re not familiar, makes it incredibly easy to setup, configure, and provision virtual machines for development and testing purposes.

Vagrant, and Ansible, the tool I use for provisioning, are awesome. I work with a team of six other developers and designers, most of whom have no experience with virtual machines or server configuration, and Vagrant allows me to provide each of them a local, sandboxed environment for them to develop in that closely resembles the production environment. When I release a new version of the box all they need to do is a

$ git pull
$ vagrant reload --provision

to update.

Vagrant makes it to forward folders between the “host” and “guest” boxes (the host being your “native” operating system and the “guest” OS being the virtual machine being run by Vagrant).

Vagrant’s configuration is defined in a file in the root of your project, named Vagrantfile. The setup of folder forwarding looks like this:

Vagrant.configure("2") do |config|
  # Configuration goes here
  # ...
  config.vm.synced_folder "/Users/chrishepner/dev/mysite", "/var/www/mysite"
  config.vm.synced_folder "/Users/chrishepner/dev/work", "/var/www/work"

Here we have ~chrishepner/dev/mysite on my local machine forwarding to /var/www/mysite on the Vagrant VM. This just means that any changes I made in ~/dev/mysite are automatically synced with /var/www/mysite in the VM, and vice-versa.

You can probably see the problem here: not every developer is going to have the code they want forwarded to the VM on the same path. Since the Vagrantfile is part of the repository, it’s problematic for developers to be setting their synced folders in that file directly. Sure, we could force everyone to place all their files in some /vagrant directory or something similar, but that idea makes me sad.

Locally-Customized Folder Forwarding

I initially tried to achieve this by setting up a single synced folder within the vagrant directory, like so:

config.vm.synced_folder "site/", "/var/www/"

I then intended to have everyone just add symlinks within this site folder to the directories they want synced. This wasn’t the best idea for several reasons, the most important being that Virtualbox, the hypervisor I was using, really didn’t want to cooperate with syncing symlinked directories, even when I set it’s configuration to do so. Time for a different approach.

Thankfully, the Vagrantfile is just Ruby, so we can add on to it as we please. My day job is primarily a PHP shop, so I didn’t want to require the team to write in a language they weren’t familiar with. I created a local.yml file, intended to be excluded from version control, where each person can add key-value pairs of host:guest directories they want to sync. I modified the default Vagrantfile as follows:

require 'yaml'

# ...

# Include local configurations for paths synced with the webroot
local_exists = false
if File.exists?(File.join(vagrant_dir,'local.yml')) then
    local_exists = true
    local_config = YAML.load_file(File.join(vagrant_dir, 'local.yml'))

# ...

config.vm.define "web" do |web|
    # ...
    if local_exists then
          local_config['synced'].each do | synced |
            web.vm.synced_folder synced['local_path'], synced['server_path'], :mount_options => [ "dmode=775", "fmode=774" ]

I also added a local.yml.example within the repo, which the developer just needs to copy to local.yml and customize:

# Configuration file for individual developer.

# "synced" defines folders to forward to the
#  virtual machine. local_path is the
#  path on your local machine, and server_path
#  is on the remove machine.
# After modifying this file, you will need to run 
# 'vagrant reload' for your changes to take effect.
  - local_path: /home/yourlocal/dev/ansible
      server_path: /var/www/html/first
  - local_path: /home/yourlocal/dev/otherpath
      server_path: /var/www/html/second

I’ve created a repository, “my-lamp” using this pattern. (This example provisions a simple LAMP stack, but the logic detailed above does not depend on PHP or Apache.) In the next article, I’ll discuss adding support for adding MySQL databases and their user accounts.


  • vagrant
  • ansible


comments powered by Disqus