Capistrano to deploy rails application - how to handle long migrations?

Question

So I am using Capistrano to deploy a rails application to my production server (apache+passenger) and at the moment deployment usually goes along the lines:

$cap deploy
$cap deploy:migrations

It got me wondering, let's say my db:migrations took a long time to execute on the production server (a big refactor of the db schema) - in this case what is best practice with Capistrano? What happens if users are connected to my application at the time of deployment? Should I gracefully send users to a static placeholder page while the database is being updated? Does Capistrano handle this automagically? Do I need to code up a recipe to help with this? Or does the internal mechanisms of rails / passenger mean that I don't have to worry at all about this particular case?

Thanks.

Answer 1

You should put up a maintenance page if the application is not going to be available for a while. I use this Capistrano task:

namespace :deploy do
  namespace :web do
    desc <<-DESC
      Present a maintenance page to visitors. Disables your application's web \
      interface by writing a "maintenance.html" file to each web server. The \
      servers must be configured to detect the presence of this file, and if \
      it is present, always display it instead of performing the request.

      By default, the maintenance page will just say the site is down for \
      "maintenance", and will be back "shortly", but you can customize the \
      page by specifying the REASON and UNTIL environment variables:

        $ cap deploy:web:disable \\
              REASON="a hardware upgrade" \\
              UNTIL="12pm Central Time"

      Further customization will require that you write your own task.
    DESC
    task :disable, :roles => :web do
      require 'erb'
      on_rollback { run "rm #{shared_path}/system/maintenance.html" }

      reason = ENV['REASON']
      deadline = ENV['UNTIL']      
      template = File.read('app/views/admin/maintenance.html.erb')
      page = ERB.new(template).result(binding)

      put page, "#{shared_path}/system/maintenance.html", :mode => 0644
    end
  end
end

The app/views/admin/maintenance.html.erb file should contain:

<p>We’re currently offline for <%= reason ? reason : 'maintenance' %> as of <%= Time.now.utc.strftime('%H:%M %Z') %>.</p>
<p>Sorry for the inconvenience. We’ll be back <%= deadline ? "by #{deadline}" : 'shortly' %>.</p>

The final step is to configure the Apache virtual host with some directives to look for the maintenance.html file and redirect all requests to it if it's present:

<IfModule mod_rewrite.c>
  RewriteEngine On

  # Redirect all requests to the maintenance page if present
  RewriteCond %{REQUEST_URI} !\.(css|gif|jpg|png)$
  RewriteCond %{DOCUMENT_ROOT}/system/maintenance.html -f
  RewriteCond %{SCRIPT_FILENAME} !maintenance.html
  RewriteRule ^.*$ /system/maintenance.html [L]
</IfModule>

To put the application into maintenance mode, run cap deploy:web:disable and to make it live again do cap deploy:web:enable.

Answer 2

My production deploys generally follow this process:

1. cap production deploy:web:disable which directs all requests to a static maintenance page
2. cap production deploy
3. migrations etc, testing each of the servers individually to make sure things are OK
4. cap production deploy:web:enable to make the site work as it should

John Topley's response gives you some good in depth info here.