This project is not maintained at the moment. We have plans to work on a universal profiling protocol after which we intend to rewrite major parts of Rbkit.

rbkit is a Ruby gem that plugs into your ruby process, taps profiling data in realtime and sends it across the wire to the rbkit-client as packed messages.

Rbkit requires the following executables to be present :

• libtool
• autoconf
• automake

Install these with your operating system's package manager before proceeding to install Rbkit gem.

Add the following to the project's Gemfiles

gem 'rbkit', path: <RBKIT_PATH>

and run bundle install

You can also install rbkit without bundler/rubygems. This can be useful, if you are profiling a Ruby app where you want to measure overhead of Rubygems/Bundler.

Just clone the repository or download from a release tag and run:

# Run from root of rbkit directory.
~> ruby setup.rb

This should install rbkit in ruby's site_dir and then you don't need rbkit added to your Gemfile for requiring rbkit.

Wherever you want to start profiling, add the following :

require 'rbkit' # Not needed in Rails
Rbkit.start_server

If using Rails, and you want to measure everything from the boot process, a good place to put this would be at the end of config/boot.rb.

Without any arguments, it's same as :

Rbkit.start_server(pub_port: nil, request_port: nil)

Starts the Rbkit server and waits for a client to connect and issue commands to the request_port, until then there's zero performance overhead. Profiling data is sent asynchronously over pub_port. This method can be called early in a ruby application so that whenever profiling needs to be done, the client can attach itself to the inactive server, do the profiling and leave. Returns an instance of Rbkit::Server if server was started successfully, else returns false.

Without any arguments, it's same as :

Rbkit.start_profiling(
  pub_port: nil,
  request_port: nil,
  enable_object_trace: true,
  enable_gc_stats: true,
  enable_cpu_profiling: true,
  clock_type: :wall,
  cpu_profiling_mode: :sampling,
  cpu_sampling_interval_usec: 1000
)

Starts the server with all tracepoints enabled by default. User can optionally disable tracepoints using the optional arguments. This method can be used to profile the startup process of a ruby application where sending commands from the client to enable profiling is not feasible. Returns an instance of Rbkit::Server if server was started successfully, else returns false.

Arguments:

If zmq and msgpack are not installed, Rbkit automatically downloads and installs the two libraries from source during gem installation. But if you are developing Rbkit, it makes sense to have these preinstalled:

On OSX - Using homebrew following command should suffice:

~> brew install zeromq
~> brew install msgpack

On Linux - we recommend to download these libraries from their respective home pages and manually compiling and installing.

Recommended versions are:

Zeromq: 4.0.5_2 Msgpack: 0.5.9

git clone git@github.com:code-mancers/rbkit.git

We'll call this <RBKIT_PATH>.

Set the environment variable RBKIT_DEV to true. If using bash, put export RBKIT_DEV=true in your ~/.bashrc.

This compiles the C extension with debug flag and also sets a macro named RBKIT_DEV inside the C extension.

Two ways to do this :

cd <RBKIT_PATH>
bundle install
bundle exec rake compile

cd <RBKIT_PATH/ext>
ruby extconf.rb
make
# Create a symlink at `lib/rbkit_server.bundle` (or .so if on linux)
# that points to `ext/rbkit_server.bundle`
# (in order to use `rbkit` gem in Gemfiles using `path` option)

bundle exec rake

TODOs are tracked as github issues.