Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time



Please see HACKING


Please see the LICENSE file

Further documentation
Also see the files in the doc/ folder, they contain further detailed information
about setting up the application for specific linux distributions such as 
CentOS and Ubuntu/Debian

Installation to a production environment -- a partial walkthrough.

Make ready

You may want to make separate directories, away from everything, to hold the
site code and the git repository respectively.  In production, you'll be setting
up a special user account too, but don't worry about that yet.

For this intro we're going to use, as examples,

* /www/gitorious           -- directory for the website code
* /gitorious/repositories  -- root directory for the git repositories
* a MySQL database on localhost at port 3306 with a _mysql_ user 'gitorious'
* eventually, a system account named 'gitslave'

All of these can be adjusted to suit: specifically, dirs within your home
directory are fine, and (though MySQL has the best development coverage), the
website code should be free of mysql-isms/quirks.


First of all, **we assume you have a working Ruby installation on your
machine**. While Gitorious may run fine on Ruby 1.9, you are likely to run
into encoding issues, and possibly other problems. has reverted
its Ruby 1.9 setup to Ruby Enterprise Edition, which is the Ruby version we
recommend. If you don't want to run REE, we recommend stock 1.8.7.


First, install each of these Libraries/applications:

* Git                   (
* Sphinx                (
* MySQL                 (or whatever)
* ImageMagick           (need version >= 6.3.0)
* aspell (optional)

An ActiveMessaging ( compatible queue
server. runs ActiveMQ and STOMP, stompserver might be a nice
flyweight alternative for your needs

Getting the code

Next, get the gitorious code itself:

  # mkdir /www/gitorious ;# or another location of your choice
  # cd    /www/gitorious
  # git clone git:// gitorious

Now you need edit a couple of configuration files.


This file contains settings for the databases used for Gitorious. The
Gitorious distribution provides a sample config file - database.sample.yml -
that you could copy to config/database.yml.

Gitorious should be database agnostic, so feel free to use your database of
choice. uses MySQL, but there shouldn't be any MySQL specific
code in Gitorious, so use whatever you're most comfortable with.


This file holds Gitorious specific configuration for each Rails environment.
There's a sample file in config/gitorious.sample.yml that describes what each
instruction means. Do make sure you have a section for each Rails environment,
especially if you're upgrading an existing Gitorious installation.

* Create a directory to hold project files
  # sudo mkdir  /gitorious/repositories
* Add the path to this directory under repository_base_path in gitorious.yml
* Make a long, complicated string. You can run "apg -m 64", or if you lack 'apg'
    `dd if=/dev/random count=1 | md5sum` ,
  and put that on the 'cookie_secret' line (replacing the 'ssssht').

Install the gems required by Gitorious

Gems are handled by Bundler. Install Bundler, then use it to install all of
Gitorious' requirements. Note that Gitorious does not work out of the box with
RubyGems 1.5, so make sure you're on an earlier version (e.g. 1.4.x).

  [sudo] gem install bundler
  bundle install

Create your database

It is now time for creating the Gitorious database. 

  bundle exec rake db:create:all
will create an empty database for you.

Migrate your database

Now that you have a database, it's time to add the database schema

  bundle exec rake db:setup
will take care of this for you

Run the tests

Running all the tests will ensure your Gitorious installation is correctly set
up. It takes less than a minute to run all the tests, and gives confidence
that the code is working as intended.

  bundle exec rake test
Once all the tests pass, you're almost ready.

Messaging server

Many Gitorious operations are performed asynchronously to ensure good
performance. Examples of such tasks includes updating the database when pushing
to Gitorious, creating bare git repositories when creating repositories in the
web UI and more. To process these asynchronous actions, Gitorious uses a
messaging system where it sends messages to a queue, and a worker (i.e. another
process, usually some kind of daemon) fetches messages back for processing.

Gitorious provides several messaging implementations ("adapters"). The
alternatives along with how to install and run them are presented below. You
only need one of these alternatives.

Sync adapter

Processes messages synchronously, which means that no extra process is
required. This is a very simple solution, but will yield poor performance. It's
intended use is for development, but may also fit small setups where performance
is not an issue (e.g. if resources are scarce). To use it, simply set
messaging_adapter in gitorious.yml to "sync":

  messaging_adapter: sync

Stomp adapter

The default (backwards compatible) adapter. This loads the implementation that's
been in Gitorious for a while. If you are upgrading an earlier install, using
this adapter will "just work". To use it, set messaging_adapter to "stomp" in

  messaging_adapter: stomp

The Stomp adapter uses stomp to talk to your Stomp aware message queue of
choice. A pretty reliable solution is to use Apache's ActiveMQ messaging queue
system, located at Download a version for your
operating system, unpack, and follow the instructions to get running. Instructions
for installing on Ubuntu can be found in this Gitorious installation guide:

An init script for ActiveMQ can be found here:
With it, starting ActiveMQ is as simple as:

sudo /etc/init.d/activemq start

The ActiveMessaging plugin needs to know how to contact your MQ server, so it
needs a config/broker.yml file. The Gitorious distribution provides an example
(the defaults should work in a development environment). Just copy the
config/broker.yml.example file to config/broker.yml and start your server.

The scripts/poller script uses processors found in app/processors to consume
messages from a Stomp queue. The script can be run in a blocking mode (useful in
development) by issuing:

  bundle exec script/poller run

To start the poller as a non-blocking daemon (as you would in a production
environment), issue:

  bundler exec script/poller start

Not that Gitorious uses ActiveMessaging to publish messages to the queue. Even
though ActiveMessaging supports several brokers, your only option for Gitorious
is Stomp. The reason is that we use Stomp directly to consume messages, and in
the future we are likely to remove ActiveMessaging alltogether in favor of a
direct Stomp implementation.

Resque adapter

Resque uses Redis as a backend for messaging. It comes with a nice
administration interface that allows for resending of messages, introspection
and general statistics about your queue. To use it, set messaging_adapter to
"resque" in gitorious.yml:

  messaging_adapter: resque

To use Resque, you need to install and run Redis. This is described in detail on
the official Resque page:

To process messages from the queue with Resque, you need to run rake:

  RAILS_ENV=production QUEUE=* bundle exec rake environment resque:work

You can also run a worker for a single, or a handful of queues too. This allows
you to assign different priority to different queues. The list of queues in use
can be found in lib/gitorious/messaging/resque_adapter.rb.

Note that Gitorious generally uses JMS style queue names, e.g.
/queue/GitoriousPostReceiveWebHook. Because the Resque web frontend does not
handle queue names with slashes in them, we strip queue names such that the
aforementioned queue will be named GitoriousPostReceiveWebHook under Resque.

Get Sphinx going

  bundle exec rake ultrasphinx:configure RAILS_ENV=production
  bundle exec rake ultrasphinx:index RAILS_ENV=production
  bundle exec rake ultrasphinx:daemon:start RAILS_ENV=production
This sequence of commands will configure, index the database and start the
sphinx daemon

Tweak environment

* If you haven't set up your mailer, production mode will fail on login. Set
    config.action_mailer.delivery_method = :test
  for immediate gratification.

Start the server

Now that you have all the components running, it's time to try Gitorious on
your local machine:

  bundle exec script/server (for Mongrel/Webrick)
  thin start                (for Thin)

Remember the values you put for gitorious_host and gitorious_port in
config/gitorious.yml? Go to http://<gitorious_host>:<gitorious_port> to see
the main page of your local Gitorious. You are now officially up and running,

Production setup

In a production environment, you'll want to automate as much as possible. We
use Capistrano for deploying to, and use custom Capistrano
recipes for starting and stopping the various tasks in Gitorious.

The doc/recipes directory contains instructions for setting up init scripts
for these tasks for various platforms. Please consult these, and feel free to
contribute your own!


The sphinx daemon needs to be running in order for full text search to work.


For cloning over the Git protocol, Gitorious includes a custom git daemon
found in script/git-daemon. 


The script/poller script needs to be kept running. 

Web/application server

While Mongrel/thin is great for trying out Gitorious on a local install, in a
production environment you'll probably want something a little more robust.
Most Rails folks use Passenger with Apache or Nginx these days, Gitorious will
play happily in such an environment. 

Support for pushing via SSH

In order for people to be able to push to your repositories, you need an SSH
daemon running. You'll also need a system user account that's used for git
connections through SSH. Consult the recipes in doc/recipes for instructions
on how to set this up. Until you've done this, the web frontend will work, but
users won't be able to push their changes.

Button up

* In production, you'll want to have a limited-privileges user to run the git
  processes, just as you do for your webserver

* Make the tree invisible to any other non-root user; make the tree read-only
  by that user; but grant write access to the /tmp and /public/cache

* Consider setting up the standard (lighttpd|nginx|apache) 
  frontend <=> mongrel backend if you see traffic (or configure 
  mod_passenger for apache2).

More Help

* Consult the mailinglist ( or drop 
  by #gitorious on if you have questions.


Gitorious will add a 'forced command' to your ~/.ssh/authorized_keys file for
the target host: if you start finding ssh oddities suspect this first.  Don't
log out until you've ensured you can still log in remotely.


Gitorious on JRuby running the web app, git daemon and poller in a single process.




No releases published


No packages published