TCP reverse proxy to copies of Twisted business logic processes

Question

I'm looking for a way implement a reverseProxy to copies of (Twisted) server processes. I'm think of a setup where the business logic is run in copies to allow for easy maintenance and upgrade, and stores shared data in a database and perhaps memcached.

I saw the reverseProxy class in twisted.web, but I don't think this is what I'm looking for for non-HTTP.

First off, is this a good design in general and/or is there a more "twisted" way to do it?

Answer 1

You might want to consider running an established load-balancing reverse proxy such as HAProxy or nginx, since it will be the weak point of your system. If you have a bug in your reverse proxy/load balancer, everything goes down. You could also handle SSL at the proxy so your twisted application servers don't need handle it.