Deploying a Drupal Site with Capistrano 2

A little over a year ago I wrote up some instructions for deploying drupal sites using capistrano. It’s proved a popular entry, still getting a good bit of traffic, but in the time since I wrote it Capistrano 2 has joined us and my techniques have moved on, so it seemed high time I updated the instructions with some new ones.

As before, I’m going to presume that anyone reading this already has capistrano installed and has shell access to their server. If you need help with the former, I’d recommend stopping by the Capistrano website, and for the latter you should probably talk to your hosting company.

My approach is to keep each site’s assets (modules, themes, etc) within that site’s folder, and then store each site in version control (git or subversion). In common with capistrano-based rails deployments, I then have the site’s files and images stored in a shared folder and symlinked into the site, so that they can be preserved between deployments.

To get started go into the folder for the site you want to be able to deploy (eg. /path/to/drupal/sites/mysite.com) and type at the command line:

mkdir config
capify .

That will create the basic files for deployment, which you will then need to edit with your configuration details. I start by opening the file config/deploy.rb and deleting everything in it, so as to start with a blank slate. I then put in some overrides to capistrano’s default deployment methods so that they work better with drupal:

set :asset_folders, %W(images files)

namespace :deploy do
  desc "Link the asset folders from the shared folder into our site"
  task :finalize_update, :except => { :no_release => true } do
    logger.info 'finalizing update with custom method'
    run "chmod -R g+w #{latest_release}" if fetch(:group_writable, true)

    asset_folders.each do |asset|
      run "rm -rf #{release_path}/#{asset}"
      run "ln -nfs #{shared_path}/#{asset} #{release_path}/#{asset}"
    end
  end

  desc "Set up the expected application directory structure on all boxes"
  task :setup, :except => { :no_release => true } do
    run <<-CMD
      umask 02 &&
      mkdir -p #{deploy_to} #{releases_path} #{shared_path}
    CMD
    asset_folders.each do |asset|
      run "mkdir #{shared_path}/#{asset}"
    end
  end

  task :set_permissions, :except => { :no_release => true } do
    # do nothing
  end

  task :restart do
   # do nothing
  end
end

You then need to set some basic configuration, so also in deploy.rb place:

set :deploy_to, '/path/to/your/drupal/sites'

# Make sure this is the domain name of your app as it is
# what your sites folder will be named
set :site_name, 'staging.scodigo.com'
role :web, "your.server.com"
set :current_path, "#{deploy_to}/#{site_name}"
set :shared_path, "#{deploy_to}/shared/#{site_name}"
set :repository, "svn://your.svn.com/path/to/repos/trunk"

With that in place you’re all ready to go. But there was an extra step I decided to add this time around. Since I am frequently setting up new staging/test sites and often need to upgrade them to new versions of drupal, I wanted to be able to install and update the drupal files using capistrano.

To do that, I wrote a recipe that will grab the tarball for a given release, unpack it and install it, without wiping out an existing sites folder. To add that you will need to put the following code into your deploy.rb file:

namespace :drupal do
  desc <<-DESC
  Grab the specified version of drupal and install it

  This presumes that the variables drupal_version and drupal_path have been set
  DESC

  task :install do
    run "cd #{shared_path} && curl -O http://ftp.drupal.org/files/projects/drupal-#{drupal_version}.tar.gz"
    run "cd #{shared_path} && tar xzvf drupal-#{drupal_version}.tar.gz"
    run "rm -rf #{shared_path}/drupal-#{drupal_version}/sites"
    run "cd #{drupal_path} && rm -rf !(sites)"
    run "mv #{shared_path}/drupal-#{drupal_version}/* #{drupal_path}"
    run "mv #{shared_path}/drupal-#{drupal_version}/.htaccess #{drupal_path}/.htaccess"
    run "rm -rf #{shared_path}/drupal-#{drupal_version}"
  end
end

after "deploy:setup", "drupal:install"
set :drupal_path, File.join(deploy_to, '../')
set :drupal_version, "5.8" # Or whatever version you want to install

(You can download a copy of the combined deploy.rb file here or view it as a pastie here)

Now, to set up your drupal site you just need to change directory to your site and type and command line:

cap deploy:setup

and to deploy a new version:

cap deploy

In practice, I tend to use these techniques in combination with the Capistrano Multistage extension, so that I can deploy a site to a staging server first, and then a production server. I’d highly recommend that approach, but for now will leave it as an exercise for the reader.

Please note that this technique is also designed to support a multi-site drupal set up, but hasn’t been extensively tested in that context. It should work, but no guarantees!

Tags: , , , ,

4 comments

  1. Thanks for the post. This general idea can be applied for any non-rails setup.

  2. Thanks Nathan, and absolutely. I’m using capistrano for a variety of project types and find it works well across all of them.

    But having written an early and well-linked post about drupal and capistrano it seemed only right to post an update, plus I hadn’t yet seen anyone post a recipe for downloading and installing drupal from a tarball.

  3. Very informative post, thanks. Does this script also deploy the database?
    Thanks,
    Ramesh

  4. @Ramesh – No, you could add tasks to create/update the drupal database but since there are quite a few ways that people approach that I’ve not included any code to do that here.

    If you want to add tasks to manage your database I’d recommend exploring the capistrano documentation at:

    http://capify.org/

    and looking at the different ways to add tasks and harness hooks.