Hѱdra

Spread your tests over multiple machines to test your code faster.

Description

Hydra is a distributed testing framework. It allows you to distribute your tests locally across multiple cores and processors, as well as run your tests remotely over SSH.

Hydra’s goals are to make distributed testing easy. So as long as you can ssh into a computer and run the tests, you can automate the distribution with Hydra.

Usage

In your rakefile:

require 'hydra'
require 'hydra/tasks'

Hydra::TestTask.new('hydra') do |t|
  t.add_files 'test/unit/**/*_test.rb'
  t.add_files 'test/functional/**/*_test.rb'
  t.add_files 'test/integration/**/*_test.rb'
end

Run:

$ rake hydra

Hydra defaults to Single Core mode, so you may want to configure it to use two (or more) of your cores if you have a multi-processing machine.

Hydra + Cucumber

Hydra can run cucumber features, but because of cucumber’s specialized environment, cucumber features have to be run after your other tests.

Hydra::TestTask.new('hydra') do |t|
  t.add_files 'test/unit/**/*_test.rb'
  t.add_files 'test/functional/**/*_test.rb'
  t.add_files 'test/integration/**/*_test.rb'
  # cucumber
  t.autosort = false
  t.add_files 'features/**/*.feature'
end

Hydra’s autosort feature sorts files by their runtime, so we have to disable it in order to run cucumber features at the end

Supported frameworks

Right now hydra only supports a few frameworks:

Test::Unit
Cucumber

We’re working on adding more frameworks, and if you’d like to help, please send me a message and I’ll show you where to code!

Running Remote Tasks

You can run tasks across all of your remote workers easily with Hydra. In your rake file, add:

Hydra::RemoteTask.new('db:reset')

Then you can run:

rake hydra:remote:db:reset

Running Global Tasks

A Global task is a task run locally and remotely. It’s used in the same way as RemoteTask:

Hydra::GlobalTask.new('db:reset')

But it is invoked in a higher namespace:

rake hydra:db:reset

Configuration

Place the config file in the main project directory as ‘hydra.yml’ or ‘config/hydra.yml’.

Examples

Dual Core

workers:
  - type: local
    runners: 2

Dual Core, with a remote Quad Core server

The -p3022 tells it to connect on a different port

workers:
  - type: local
    runners: 2
  - type: ssh
    connect: [email protected]
    ssh_opts: -p3022
    directory: /absolute/path/to/project
    runners: 4

Two Remote Quad Cores with Synchronization

You can use the ‘sync’ configuration to allow rsync to synchronize the local and remote directories every time you run hydra.

workers:
  - type: ssh
    connect: [email protected]
    directory: /path/to/project/on/alpha/
    runners: 4
  - type: ssh
    connect: [email protected]
    directory: /path/to/project/on/beta/
    runners: 4

sync:
    directory: /my/local/project/directory
    exclude:
      - tmp
      - log
      - doc

Workers Options

type

Either “local” or “ssh”.

runners

The runners option is how many processes will be running on the machine. It’s best to pick the same number as the number of cores on that machine (as well as your own).

SSH Options

connect

The connect option is passed to SSH. So if you’ve setup an ssh config alias to a server, you can use that. It is also used in rsync, so you cannot use options.

ssh_opts

The ssh_opts option is passed to SSH and to Rsync’s RSH so that you can use the same ssh options for connecting and rsync. Use ssh_opts to set the port or compression options.

Using Hydra::Listeners

Hydra comes with a couple of listeners for the events it fires. By default, Hydra::Listener::MinimalOutput is used to display the files being tests and the ./F/E for each file and F/E output.

It also uses Hydra::Listener::ReportGenerator to generate reports of the test files to that it can order them by their run times.

To use another listener, just add a listeners node to the config file. For example, if you are on Ubuntu Linux (or have access to the notify-send command) you can add a notifier listener like this:

listeners:
  - Hydra::Listener::Notifier.new

Note that if you make a listener node, the default listeners will be overridden, so you will no longer have the standard minimal output unless you do:

listeners:
  - Hydra::Listener::Notifier.new
  - Hydra::Listener::MinimalOutput.new

Listeners take one argument to their contstructor: an IO object. So, you can easily output Hydra to a variety of log files. For example:

listeners:
  - Hydra::Listener::ReportGenerator.new(File.new('/home/ngauthier/Desktop/hydra_log.yml', 'w'))

More Information

For more information on Hydra, check out the rdocs:

rdoc.info/projects/ngauthier/hydra