After a short 15 minute introduction, Nick left us with our Rails app to start it with JRuby on a mongrel. I’ve been toying around with JRuby a bit before, and this was rather easy. Some people, however, encounter some issues with their apps because of incompatible gems, like OpenSSL or RMagic which are not available for the Java platform. Luckily, there are some nice Java alternatives like JRuby OpenSSL and ImageVooDoo. My app didn’t suffer from these issues, but from what I’ve heard, these alternatives work great.
Next for the big work: creating a WAR file and deploying it. I’ve done this before, and it’s dead easy. In stall the Warble gem, warble your Rails app and deploy the resulting WAR file.
Nick also talked about the differences between using the JVM and Unix. The biggest differences are memory management and the creation of new threads/processes.
There are quite a few things that Rails and a Java app server have in common, like logging and session handling. In his sheets (I’ll link to those later), you can see how you can tune your Rails app to make use of these features and improve your apps performance (by not logging and keeping sessions in two different places).
Continuing on the performance of Glassfish/JRuby, Nick shows several graphs on how well JRuby performs. A lot of attention goes to Rails 2.2, which will be thread-safe and what it means for JRuby.
JRuby can run with 1 runtime and fill your CPU with threads to handle requests using native Java threads. To handle database connections, Nick Sieger has written a connection pooling component for ActiveRecord (to prevent every thread having it’s own connection to your database). An unreleased version of warbler was distributed that will allow you to set max_instances to 1. This will create 1 Rails instance and use threading in Glassfish. Of course, this is only useful with Rails 2.2.
This was altogether a great tutorial afternoon. Besides the basic warble-your-app-and-deploy-it, Nick provided a lot of interesting facts and tips about deploying to Glassfish. I already was enthusiastic about JRuby, but now I really have to get my hands on a decent server that will run Glassfish comfortably. This means at least at least 1Gb of RAM, so I think I have found at least one drawback of using JRuby.