SIO2 is a free platform for carrying out algorithmic contests and OIOIOI is its main component --- the web interface.

It should be easier to begin with a separate folder at first:

mkdir sio2
cd sio2

and to install OIOIOI inside a virtualenv:

virtualenv venv
. venv/bin/activate

Then OIOIOI and its dependencies can be installed using the following commands:

git clone git://github.com/sio2project/oioioi.git
cd oioioi
pip install -r requirements.txt

OIOIOI is a set of Django Applications, so you need to create a folder with Django settings and other deployment configuration:

cd ..
oioioi-create-config deployment
cd deployment

The created deployment directory looks like a new Django project, but already configured to serve the OIOIOI portal. You need to at least set the database configuration in settings.py.

In case of using PostgreSQL, install Psycopg2:

pip install psycopg2

Finally initialize the database:

./manage.py migrate

We use PostgreSQL.

Then you need to copy static files, like images and styles, to the deployment directory:

./manage.py collectstatic

You'll also need lessc (a LESS compiler, minimum version 1.3.3). First you need to install node.js with npm as described in their documentation, then install less; either globally as root:

sudo npm install less

or locally inside virtualenv:

cd ../venv/
npm --prefix . --global install less

In the simple configuration, OIOIOI will use the system-installed compilers, and will not use the safe execution environment. User's programs will be run with the normal user privileges. This is not a safe configuration and the judging will run quite slowly. It is to easily make OIOIOI up and running for testing purposes.

Ensure that required dependencies are installed:

• gcc/g++ (Ubuntu package: build-essential)
• fpc (Ubuntu package: fp-compiler)
• latex with support for Polish (Ubuntu packages: texlive-latex-base, texlive-lang-polish)

and in one terminal run the Django web server:

./manage.py runserver 0.0.0.0:8000

and in the other the evaluation daemons:

./manage.py supervisor

The supervisor process monitors all processes needed by OIOIOI, except the web server. It has many nice features.

You can create an administrator account by running:

./manage.py createsuperuser

If you see a locale error, you may want to circumvent it by providing another locale to the command:

LC_ALL=C ./manage.py createsuperuser

Now you're ready to access the site at http://localhost:8000.

1. Begin with the simple configuration described above.
2. Ensure that production-grade dependencies are installed:
   • lighttpd binary (Ubuntu package: lighttpd, shall not be run as service.)
   • uwsgi (pip install uwsgi)
3. Make sure you are in the deployment folder and the virtualenv is activated.
4. Install RabbitMQ. We tested version 2.8.6 from RabbitMQ Debian/Ubuntu Repos. Anything newer should work as well.
5. Uncomment and set BROKER_URL in settings.py to point to the configured RabbitMQ vhost. The default setting corresponds to the default RabbitMQ installation.

6. Download sandboxes:

   ./manage.py download_sandboxes

7. Disable system compilers and unsafe code execution by commenting out USE_UNSAFE_EXEC = True and USE_LOCAL_COMPILERS = True in settings.py.
8. (optionally) Disable starting the judging process on the server, especially if you want to configure judging machines (see below) for judging, what is strongly recommended. Comment out the RUN_LOCAL_WORKERS = True setting.
9. (optionally) Enable Filetracker server by uncommenting corresponding lines in settings.py and restart the daemons. This is required for dedicated judging machines.

10. Install and configure web server. We recommend using nginx with uwsgi plugin (included in nginx-full Ubuntu package). An example configuration is automatically created as nginx-site.conf. Have a look there. What you probably want to do is (as root):

    cp nginx-site.conf /etc/nginx/sites-available/oioioi
    ln -s ../sites-available/oioioi /etc/nginx/sites-enabled/
    service nginx reload

    Once this is done, you no more need to run manage.py runserver.

    If you prefer deploying with Apache, an example configuration is created as apache-site.conf. You would need to install apache2 and libapache2-mod-uwsgi packages.

11. Comment out DEBUG = True in settings.py. This is crucial for security and efficiency. Also set ALLOWED_HOSTS.
12. Set admin email in settings. Error reports and teacher account requests will be sent there.
13. Set SMTP server in settings. Otherwise new user registration (among others) will not work.

14. You probably want to run manage.py supervisor -d automatically when the system starts. One way is to add the following line to the OIOIOI user's crontab (crontab -e):

    @reboot <deployment_folder>/start_supervisor.sh

15. (optionally) If you have efficiency problems or expect heavy load, you may consider using gevent as uwsgi event loop. To do so, install gevent and set UWSGI_USE_GEVENT flag in settings.py.

16. (optionally) You can also enable content caching. To do so, first you have to install dependencies:

    • memcached (Ubuntu package: memcached)
    • python-memcached (pip install python-memcached)

    Next, you have to uncomment the corresponding lines under "Cache" in settings.py and set the address of your memcached instance. Note that you can run memcached locally or on a remote server. For more information about memcached configuration see official documentation.

17. (optionally) You can ensure users are automatically notified of certain events in the system - or notify them on your own - just enable the Notifications Server. For more information, consult the notifications/README.rst file.

Before setting up judging machines, you need to configure the RabbitMQ server to accept remote connections. This can be done by creating a new user account or by allowing the default guest account to connect from a remote host, by creating the configuration file /etc/rabbitmq/rabbitmq.config with the following content:

[{rabbit, [{loopback_users, []}]}].

and restarting the RabbitMQ server. Then on every juding machine do the following:

1. Create a new user account for the judging processes and switch to it.

2. Set up virtualenv:

   virtualenv venv
   . venv/bin/activate

3. Install the sioworkers package:

   pip install sioworkers

4. Start the worker process:

   sio-celery-worker amqp://guest:guest@[server]:5672//

   The passed argument must point to the RabbitMQ server configured on the server machine.

5. That's all. You probably want to have the worker started automatically when system starts. We do not have a ready-made solution for this yet. Sorry!

The worker assumes that the Filetracker server is running on the same server as RabbitMQ, on the default port 9999. If this is not the case, you should pass the Filetracker server URL in the FILETRACKER_URL environment variable.

It is strongly recommended to install the librabbitmq Python module (on the server and the worker machines). We observed some not dispatched evaluation requests when running celery with its default AMQP binding library:

pip install librabbitmq

Celery will pick up the new library automatically, once you restart the daemons using:

./manage.py supervisor restart all

The sandboxes provided by the SIO2 Project contain 32-bit binaries. Therefore it is recommended that OIOIOI is installed on a 32-bit Linux system. Otherwise, required libraries may be missing. Here we list some of them, which we found needed when installing OIOIOI in a pristine Ubuntu Server 12.04 LTS (Precise Pangolin):

• libz (Ubuntu package: zlib1g:i386)

Make sure you are in the deployment folder and the virtualenv is activated. Then run:

pip install -e git://github.com/sio2project/oioioi.git#egg=oioioi
./manage.py migrate
./manage.py collectstatic
./manage.py supervisor restart all

and restart the judging machines.

If you're getting the "Upgrading from an old version" message when trying to sync the database, that means you had an old version of OIOIOI that was based on version 1.6 or 1.5 of the Django framework. Django 1.7 introduces a new migration system which requires a more complicated upgrade process.

IMPORTANT: BACKUP YOUR DATABASE BEFORE DOING THE NEXT STEP.

In the typical situation where you didn't create any custom migrations we've automated the process for you: make sure your database settings are valid and run:

./manage.py upgrade_to_17

That's all. If you have your own custom changes though and they are incompatible with our script or you want to understand what happens, the following needs to be done:

1. Install Django 1.6 and South and place all of the old migrations in proper directories. The easiest way is to 'git checkout' the last commit before the 1.7 commit and do 'pip install -r requirements.txt'. If you have custom changes in your OIOIOI directory and they conflict with our changes, you'll have to merge them yourself. For our automatic script we use a temporary virtualenv and a package containing all the necessary files to run the old migrations.
2. Now enable all aplications you have ever used (in the INSTALLED_APPS setting) and run ./manage.py migrate. If you don't know which applications you've used in the past, just enable them all and run ./manage.py syncdb and then ./manage.py migrate. Our script does that. If you have your own custom migrations they could be conflicting with ours. You'll have to solve these conflicts yourself.
3. Get the newest OIOIOI, install the needed packages and remove all of the old migrations. Again, the easiest way is to 'git checkout' the last commit and do 'pip install -r requirements.txt'.

4. Migrate all the new Django 1.7 migrations. The necessary changes are already in the database and in most cases Django will detect this by faking the migrations - marking them as applied without actually applying them. However some migrations need to be explicitly told to be faked. The commands that need to be run in the typical case are:

   ./manage.py migrate --fake balloons 0002
   ./manage.py migrate --fake complaints 0002
   ./manage.py migrate --fake contestexcl 0002
   ./manage.py migrate --fake contestlogo 0002
   ./manage.py migrate --fake contests 0002
   ./manage.py migrate

   assuming that these applications are in INSTALLED_APPS. If you've had your own custom migrations before and they introduced circular dependency loops on foreign keys in different applications than those mentioned above, you also have to run the ./manage.py migrate --fake command for them as well.

5. Run ./manage.py collectstatic and start the supervisor, your judging machines and the server.

When new features are added, the configuration files in your custom deployment directory may need an update. An example valid configuration can always be found in the oioioi sources (oioioi/deployment directory, *.template files). One of the simplest ways to learn about the changes is:

diff -u path_to_deployment/changed_file path_to_oioioi/oioioi/deployment/changed_file.template

Once you have made sure that your deployment directory is up-to-date, change CONFIG_VERSION in your custom deployment/settings.py so that it equals INSTALLATION_CONFIG_VERSION in oioioi/default_settings.py.

List of changes since the CONFIG_VERSION numbering was introduced:

1. • Added unpackmgr queue entry to deployment/supervisord.conf.:

     [program:unpackmgr]
     command={{ PYTHON }} {{ PROJECT_DIR }}/manage.py celeryd -E -l info -Q unpackmgr -c {{ settings.UNPACKMGR_CONCURRENCY }}
     startretries=0
     stopwaitsecs=15
     redirect_stderr=true
     stdout_logfile={{ PROJECT_DIR }}/logs/unpackmgr.log

   • Added USE_SINOLPACK_MAKEFILES and UNPACKMGR_CONCURRENCY options to deployment/settings.py.:

     USE_SINOLPACK_MAKEFILES = False
     #UNPACKMGR_CONCURRENCY = 1

2. • Added Notifications Server entries to deployment/supervisord.conf.:

     [program:notifications-server]
     command={{ PYTHON }} {{ PROJECT_DIR }}/manage.py notifications_server
     redirect_stderr=true
     {% if not settings.NOTIFICATIONS_SERVER_ENABLED %}exclude=true{% endif %}

   • Added NOTIFICATIONS_ options to deployment/settings.py.:

     # Notifications configuration (client)
     # This one is for JavaScript socket.io client.
     # It should contain actual URL available from remote machines.
     NOTIFICATIONS_SERVER_URL = 'http://localhost:7887/'
     
     # Notifications configuration (server)
     NOTIFICATIONS_SERVER_ENABLED = False
     
     # URL connection string to a Notifications Server instance
     NOTIFICATIONS_OIOIOI_URL = 'http://localhost:8000/'
     
     # URL connection string for RabbitMQ instance used by Notifications Server
     NOTIFICATIONS_RABBITMQ_URL = 'amqp://localhost'
     
     # Port that the Notifications Server listens on
     NOTIFICATIONS_SERVER_PORT = 7887

3. • Added prizesmgr queue entry to deployment/supervisord.conf:

     [program:prizesmgr]
     command={{ PYTHON }} {{ PROJECT_DIR }}/manage.py celeryd -E -l info -Q prizesmgr -c 1
     startretries=0
     stopwaitsecs=15
     redirect_stderr=true
     stdout_logfile={{ PROJECT_DIR }}/logs/prizesmgr.log

4. • Added ATOMIC_REQUESTS database option to deployment/settings.py:

     DATABASES = {
      'default': {
       'ENGINE': 'django.db.backends.', # Add 'postgresql_psycopg2', 'mysql', 'sqlite3' or 'oracle'.
       'NAME': '',                      # Or path to database file if using sqlite3.
       'USER': '',                      # Not used with sqlite3.
       'PASSWORD': '',                  # Not used with sqlite3.
       'HOST': '',                      # Set to empty string for localhost. Not used with sqlite3.
       'PORT': '',                      # Set to empty string for default. Not used with sqlite3.
       'ATOMIC_REQUESTS': True,         # Don't touch unless you know what you're doing.
      }
     }

Well, we don't have a full-fledged User's Guide, but feel free to propose what should be added here.

To run a contest, you obviously need some tasks. To add a task to a contest in OIOIOI, you need to create an archive, called task package. Here are some pointers, how it should look like:

• example task packages used by our tests,
• a rudimentary task package format specification.

Should you have any further questions regarding installation, configuration or usage of OIOIOI, feel free to contact us by an e-mail, via IRC (#tagtag at freenode) or through github issues system. Please use English on github and English or Polish elsewhere. You may find some additional information on our official website and in the official project documentation. You can also look at what we are currently working on by browsing current tickets on our issue tracker.