To use, call your cgi-fcgi -start -connect $host:$port script with
the -supervise option, like:

cgi-fcgi -start -supervise -connect 127.0.0.1:1791 /path/to/dispatch.fcgi

Full Supervise run script becomes

    #!/bin/sh
    RAIL_NUMBER=$(basename $PWD|awk -F'-' '{print $2}')
    RAILS_HOST=$(<env/RAILS_HOST)
    RAILS_ROOT=$(<env/RAILS_ROOT)
    RAILS_PORT=179$RAIL_NUMBER
    exec envdir ./env \
      cgi-fcgi -supervise -start -connect \
      $RAILS_HOST:$RAILS_PORT \
      $RAILS_ROOT/public/dispatch.fcgi

This would be in a script called ‘run’ in your ~/service/someapp-$RAIL_NUMBER
directory, where $RAIL_NUMBER is 1-99. The references to ./env require a
directory named ‘env’ to be set up in the same directory as the run script.
This should have at least 3 files in it:

RAILS_ROOT => contains one line that is the full path to your rails root directory.
RAILS_ENV => contains one word, either ‘production’ or ‘development’
RAILS_HOST => contains one IP address or FQDN

You can set any other environment variables in this way by simply creating a
file with the variable name and its contents will become the value of that
environment variable. Because of the envdir ./env call before the cgi-fcgi
call, your rails application has access to any variables set in this way.