I am aware of Capistrano, but it is a bit too heavyweight for me. Personally, I set up two Mercurial repositories, one on the production server and another on my local dev machine. Regularly, when a new feature is ready, I push changes from repository on my local machine to repository on the server, then update on the server. This is a pretty simple and quick way to keep files in sync on several computers, but does not help to update databases.
What is your solution to the problem?