Category: Technical Page 11 of 15

The default configuration for running JRuby on Rails using GoldSpike only allows 4 instances of the JRuby runtime. This means that if there are more than 4 simultaneous requests from clients, the web server will respond with a server overloaded error. The easiest way to see what’s going on is to take a look at the ‘createObjectPool’ method of RailsServlet.

With a dedicated production server with 1gig of memory, it is possible to allow a lot more JRuby runtimes, and hence allow for more simultaneous users. A little load testing on a fairly small app suggested that 20 instances of JRuby runtime would fit comfortably under the 1gig ceiling and not overload the processor.

An easy way to change the max number of runtimes is to edit ‘web.xml’ in a JRuby application’s WEB-INF directory. Parameter is set as follows:

  <context-param>
    <param-name>jruby.pool.maxActive</param-name>
    <param-value>20</param-value>
  </context-param>

It is also worth checking with the web server configuration to ensure that the web server allows more threads than the number of JRuby runtimes specified in ‘web.xml’.

Say you wanted to store some information across different calls to your JRuby/Rails application on the server side. You could use a database, or memcached. However, what if you just want to do something very simple like a basic cache, and you don’t have a database for your application? Memcached seems like overkill, and would complicate your deployment considerably. So, what about just using statics?

Well, you can’t do it in Ruby, as the Rails integration servlet spawns multiple instances of the JRuby interpreter.

Ruby code in a page view:

<h1>Ruby global number</h1>
<% $i ||= 0 %>
<% $i += 1 %>
<%= $i %>

<h1>Ruby obj id</h1>
<% $o ||= Object.new %>
<%= $o.object_id %>

Results between multiple refreshes of the page:

Ruby global number
1 
Ruby obj id
244 

Ruby global number
1 
Ruby obj id
256 

Ruby global number
2 
Ruby obj id
244 

Ruby global number
2 
Ruby obj id
256 

Our requests seem to be switching between two instances of the Ruby interpreter.

However, we *can* do it in Java. My clever, Java-literate colleagues explained that Java application servers and servlet containers use a single instance of the JVM, but have a class loader for each application. The class loader stores class information and static values. This means that within one application, the static values will be maintained between requests, but a different application on the same server will have its own set of unrelated static values (similar to AppDomains in .NET).

If we add a call to a static method in Java that increments an integer and returns it, our view looks like this:

<h1>Ruby global number</h1>
<% $i ||= 0 %>
<% $i += 1 %>
<%= $i %>

<h1>Ruby obj id</h1>
<% $o ||= Object.new %>
<%= $o.object_id %>

<h1>Java incrementing static integer</h1>
<%= TestClass.incrementAndReturnNumber() %>

Our results now look a lot more useful. The Java static integer is getting incremented each call to the view:

Ruby global number
1 
Ruby obj id
244 
Java number
1 

Ruby global number
1 
Ruby obj id
256 
Java number
2 

Ruby global number
2 
Ruby obj id
244 
Java number
3 

Ruby global number
2 
Ruby obj id
256 
Java number
4 

From this little experiment, Java statics seem like a possible way to go for storing temporary data on the server side (eg, a cache implemented as a singleton) for JRuby / Rails. Another option may be to use the ServletContext from the JRuby Rails integration servlet – probably an area worth investigating.