The answer in brief
The rails stack has your application and then the server that runs your application (aka the "application server"). This server could be webrick (not a good idea in production), thin, gunicorn, passenger etc etc.
You should be telling that server which port to run under. You'll (likely) need to specify this outside Rails - not in config/production.rb
because by the time Rails boots it's already running inside some application server.
A deeper dive with an example:
Let's use Heroku for an example, because port numbers there are essentially randomized (at least from the view of us looking in).
Heroku will pick a random port for us, then tell us through the PORT
environmental variable. With Heroku you need a Procfile to tell it what services to launch, and your Procfile may look something like this:
web: bundle exec unicorn -p $PORT -c ./config/unicorn.rb
See here, we use -p $PORT
to tell unicorn - our application server in this example - to run on some port Heroku gave us.
Back to your question:
However you kick off your application serving process in your production environment, you should tell it to specify the port number to your web server. There's a bunch of ways to kick off your application serving process at a production level: from upstart (built into Ubuntu), to supervisord to god... all these methods run commands and make sure the process stays up (an important part of production level deployment ;) )