capistrano doesn't call deploy:start when executing deploy:cold

Question

I'm using capistrano 2 to deploy my application. According to the documentation, the deploy:cold task will call deploy:update followed by deploy:migrate and finally call deploy:start.

In my application I don't have a migration step, and when I run cap deploy:cold, it does the update but then shows this:

* 2014-04-09 14:57:45 executing `deploy:migrate'
`deploy:migrate' is only run for servers matching {:roles=>:db, :only=>{:primary=>true}}, but no servers matched

and deploy:start is not being called. If I call it manually, it works fine - but never through deploy:cold. What am I missing?

Answer 1

The problem is that Capistrano 2 defines a default task called deploy:migrate that is specified to run only for the role :db. If you don't have that role in your application (which most do not), then when deploy:migrate runs it will error out on the missing role, and that kind of error normally terminates the deployment scenario.

This can happen for other tasks, if you have a complex setup where some tasks are only run for specific servers. This behavior is not documented at all at Capistrano's site, which is a shame.

Starting with Capistrano 2.7 there is a way to circumvent the "is only run for servers matching role" error, by specifying a task property that tells Capistrano to skip tasks if no servers matches any of its specified roles:

For example:

namespace :deploy do
  task :stuff, :roles => :somerole, :on_no_matching_servers => :continue do
    # stuff
  end
end

Now this works well for custom tasks, but how to set this for the default task migrate ? you have to override the task with your own setting and add the :on_no_matching_servers option. I've added this to the global.rb file:

namespace :deploy do
  task :migrate, :on_no_matching_servers => :continue do end
end