Created: 2008-12-18 20:18
Updated: 2016-05-08 19:48
License: mit


In this installment of the Ruby Advent Calendar 2008, I'll show you how easy it is to use JRuby to mix and match both Ruby and Java tools to quickly create a simple memory monitoring program.

We'll be using the venerable Gruff graphing library, along with RMagick4J, a Java port of the ImageMagick/RMagick imaging library for JRuby, and the jmx gem as a nice Ruby wrapper around Java's JMX (Java Management Extensions) apis.

If you already have a JRuby installation you'd like to use to play around with the code for this article, you can install the gems we need as follows:

jruby -S gem install gruff rmagick4j jmx

(If you haven't played with JRuby before or don't have an install handy, a bundle is available of the code for this article, including JRuby and all dependencies.)

Gruff and RMagick4J

First, let's look at getting Gruff working with RMagick4J. Though RMagick4J is, as of this writing, an incomplete port of ImageMagick and RMagick, it runs well enough to support Gruff. And that's a pretty reasonable trade-off, considering how easy it is to install RMagick4J into JRuby (above) compared to the process of setting up RMagick in MatzRuby.

require 'rubygems'
gem 'rmagick4j'
gem 'gruff'
require 'gruff'

g ="400x300")
g.title = "My Graph""Apples", [1, 2, 3, 4, 4, 3])"Oranges", [4, 8, 7, 9, 8, 9])"Watermelon", [2, 3, 1, 5, 6, 8])"Peaches", [9, 9, 10, 8, 7, 9])
g.labels = {0 => '2003', 2 => '2004', 4 => '2005'}


However, since JRuby runs on Java, which sports the platform-independent Swing GUI toolkit, we can also create a frame to display the Gruff image rather than writing it to disk. Instead of g.write('my_fruity_graph.png') try this:

include Java
import javax.imageio.ImageIO
import javax.swing.JFrame

class ImagePanel < javax.swing.JPanel
  def initialize(image, x=0, y=0)
    @image, @x, @y = image, x, y

  def getPreferredSize, @image.height)

  def paintComponent(graphics)
    graphics.draw_image(@image, @x, @y, nil)

image =
frame ="My Graph")
frame.set_bounds 0, 0, image.width + 20, image.height + 40
frame.add(, 10, 10))
frame.visible = true

A few words about this code. First, include Java mixes in JRuby's Java integration features. The next three lines import Java classes as Ruby classes and stores them in constants named the same as the Java class name. (Aside: Be careful when using this technique with a Java class whose name clashes with a Ruby class of the same name. Instead, consider an alternate style such as JFile =

ImagePanel is a Ruby class that extends Swing's JPanel, a Java class. Note that the super() invocation in ImagePanel's initializer is necessary due to Ruby's default behavior of passing all the arguments to the superclass initializer, which, in this case, is a Java constructor that doesn't take any. getPreferredSize and paintComponent are expected overrides on JPanel, and we use them to size and paint an image in the panel.

In the last part of the code, we create a Java image object, pass it into a new ImagePanel, pack the panel into a JFrame, and show the frame. to_java_bytes is a JRuby-specific method on Ruby's String class that converts the string to a Java byte[].

If you're following along, and all goes well, you should be seeing this:



Java Management Extensions (JMX) is a management and monitoring API and framework for Java. It allows you to query and manipulate a large number of system metrics and components (called "MBeans"), as well as register your own. You can control these MBeans both from within a VM and via remote connections to other VMs. Some examples of things you can do with JMX are: check memory usage, list threads and their status, gather garbage collection, class loading and compiler statistics, or force a garbage collection cycle.

Typical of Java APIs, JMX is pretty verbose, and has its share of enterprisey Managers and Factorys. But thanks to Tom Enebo's JMX gem, it's easy to get connected to JMX and retrieve useful information. Here's the code to print out the current heap memory usage of the JVM:

def in_mb(value)
  format "%0.2f MB" % (value.to_f / (1024 * 1024))

server =
memory = server["java.lang:type=Memory"]
heap = memory.heap_memory_usage.used
puts "Size of Heap: #{in_mb(heap)}"

When you run this snippet with JRuby, you should see:

Size of Heap: 13.83 MB

It's especially nice to explore the JMX MBeans in an IRB session.

>> {|b| b.to_s =~ /jruby/ }.each {|b| puts b.to_s }
=> [#<Java::JavaxManagement::ObjectName:...]
>> b = server['org.jruby:type=Runtime,name=4469286,service=Config']
=> #<JMX::MBeans::Org::Jruby::Management::Config:...>
>> b.version_string
=> "jruby 1.1.6RC1 (ruby 1.8.6 patchlevel 114) (2008-12-03 rev 8263) [i386-java]\n"
>> b = server['org.jruby:type=Runtime,name=4469286,service=ParserStats']
=> #<JMX::MBeans::Org::Jruby::Management::ParserStats:...>
>> b.total_parse_time
=> 0.843191

Here, we're exploring some of JRuby's own MBeans, and found out that the time JRuby spent parsing source code since startup is a little under a second.

Graphing the Memory in the VM

Now, let's put all these tools together. Let's start with a class that samples the memory every two seconds, and saves a rolling window of 50 samples.

class MemorySampler
  def initialize(&block)
    @samples = []
    @keep = 50
    @interval = 2
    @callback = block
    @server =
    @memory_bean ||= @server["java.lang:type=Memory"]

  def memory_usage

  def samples
    dat = @samples.dup.transpose
    { "Heap" => dat.first, "Non-Heap" => dat.last }

  def start
    @thread = do
      while true do
        @samples << {|m| m.to_f / (1024*1024) }
        @samples.shift if @samples.size > @keep if @callback
        sleep @interval

Here, we're encapsulating the JFrame and ImagePanel logic in an ImageApp class.

class ImageApp
  attr_accessor :title

  def initialize(image = nil)
    @title = "Preview"
    @image = image

  def image=(image)
    @image = image
    @panel.image = image if @panel

  def show
    @frame =
    @frame.set_bounds 0, 0, @image.width + 20, @image.height + 40
    @panel = @frame.add(, 10, 10))
    @frame.visible = true

We'll also need to add a method to ImagePanel to assign a new image to it and force a repaint.

class ImagePanel
  def image=(image)
    @image = image

Let's also add a convenience method to Gruff to create a Java image object.

class Gruff::Base
  def to_image

Now, let's start up a MemorySampler with a callback that repeatedly updates our ImageApp with a Gruff image generated from the most recent samples.

@app =
@app.title = "Memory" do |s|
  g ="400x300")
  g.title = @app.title
  g.y_axis_label = "MB"
  s.samples.each_pair {|key,val|, val) }
  g.labels = {}
  g.hide_dots = true
  @app.image = g.to_image unless @started
  @started = true

Now you should be seeing a window with a continuously updating graph of the VM's memory usage, all in right around 100 lines of code! Also, you can easily change the server to which you connect, and monitor the memory in another VM. For example, here's a profile of Glassfish starting up and serving requests to a Rails application:



In this article I've given you a taste of the kinds of mashups of Ruby and Java tech that JRuby so easily allows. JRuby makes installation and deployment of Gruff graphs easy, and allows you to access VM statistics and monitoring data in an idiomatically Ruby way with the JMX gem.

If you're interested in playing with these examples further, do check out the bundle I made available on Github. The example code is included (in a slightly different form from the examples above) in the lib/rubyadvent directory. If you want to use the JRuby that I include in the bundle, just add advent-jruby/bin to your path. Each example is accessible through a Rake task. Just run jruby -S rake to see what's available:

To run an example, re-run 'jruby -S rake' with one of the following tasks:
rake gruff:simple  # Simple Gruff example with JRuby
rake irb           # Run IRB with Gruff, RMagick4J and JMX loaded
rake jmx:memory    # Simple JMX memory monitoring example: print memory usage
rake jmx:tracker   # Gruff/JMX example plotting memory usage over time

Hope you enjoyed this installment of the Ruby Advent Calendar. Happy holidays, and happy Ruby hacking!

Cookies help us deliver our services. By using our services, you agree to our use of cookies Learn more