building chroots and running Ruby code in them
Ruby Shell
Switch branches/tags
Nothing to show



brig, noun, (...) • informal a prison, esp. on a warship.

Brig is a tool for building chroots and running Ruby code in them.

It runs code in the chroot via .popen3 or via right_popen if EventMachine is present (and its reactor is running).

The same chroot can be used over and over (:chroot) or it can be used as an 'original' (:chroot_original) and each code run is done in a clean copy of the original chroot.


a) build a chroot b) run code in it

let's first look at running Ruby code.

running in a chroot

Running Ruby code in the chroot goes like :, :chroot => '/brigs/alpha') do |stdout, stderr|
  p stdout, stderr

Usually you want the return value of the ruby code (as a Ruby object).

  :chroot => '/brigs/alpha',
  :on_success => lambda { |result| p [ :result, result ] },
  :on_failure => lambda { |stderr, code| p [ :issue, stderr ] })

Please note that the return from the Ruby in the chroot is done thanks to JSON.

running in a fresh chroot copy

Let's say you have an 'original chroot' at /brigs/alpha you want to run some Ruby code in a fresh copy of it (and discard the copy immediately) :

  :chroot_original => '/brigs/alpha',
  :on_success => lambda { |result| p [ :result, result ] },
  :on_failure => lambda { |stderr, code| p [ :issue, stderr ] })

All good, but it too more than a second to run...

running in a pool of chroot copies

It's time to create a runner (and its pool).

runner =
  :chroot_original => '/brigs/alpha',
  :batch_size => 20)

  :on_success => lambda { |result| p [ :result, result ] },
  :on_failure => lambda { |stderr, code| p [ :issue, stderr ] })

The batch size is set to 20. The runner when initialized will immediately create 20 copies of the chroot at /brigs/alpha (in directories adjacent to the original, but with timestampy suffixes) and.

After 2/3 of the batch has been consumed for running code, another batch of chroot copies will be prepared.

In order to be thread-safe, the runner uses a Queue as a pivot between the consumption and the production of chroot copies.

By default, the :batch_size option is set to 20.

building a chroot

This is a two steps operation. First, you need to compile a Ruby (1.9.2) and install gems for it. Second, you have to create the chroot itself.

From what I've seen so far, a chroot on GNU/Linux should weigh 60MB (and take something like half a second to copy).

preparing a ruby for the chroot

Building a 1.9.2 Ruby for brig (-v for verbose) :

sudo ./bin/build_ruby.rb -v

Adding gems to the brig :

sudo /brig_ruby/bin/gem install gem_0 gem_1


This will build the chroot, stealing the /brig_ruby on its way :

sudo ./bin/build.rb -v /brigs/alpha

-v is for 'verbose', it will print a small summary of the generated chroot at the end.



IRC #ruote