use fastcgi, not mod_ruby
I have deployed the fastcgi version quite successfully.
These are the simple steps I took.
After fcgi, ruby, gems then rails was installed, I simply create a new subdomain in my case right now, run these commands:
cd /home/httpd/vhosts/example.com/subdomain/ruby/
rails rails-project
vi conf/vhost.include
add the line:
DocumentRoot /home/httpd/vhosts/example.com/subdomains/ruby/rails-project/public
Modify the .htaccess file to use the .fcgi extension instead of the default cgi extension. And presto, rails is installed.
The major problem with this case study is that when doing any modification to the Rails database.yml file, I need to restart the entire web server. This is a pain, to put it bluntly. All clients suffer for this lack of support.
What would be ideal is this. Each client gets there own sandbox instance of lighttpd with fcgi installed and configured. The apache2 vhost.conf uses the proxy tools it has natively with Plesk and ports through to whatever port is configured within the lighttpd configuration. Although each client would need its own local port for this to work well, this will still be a much more practical solution when it comes time for modifying things such as the database.yml file.
At this point, a start/restart/stop tool would need to be provided to each client to be able to manage their own lighttpd installation.
In addition, the rails-project directory needs to be owned by the ftpuser
saserv user/group, otherwise FTP will not work. In my case I need to install Rails by hand for each client.
Rails is completely geared towards lighttpd, so support within Plesk should have this, but it is not required.
I am positive cPanel will be coming out with this feature within the next month or so, and I would be interested at adopting it if Plesk shows no immediate interest in looking at such a challenge.