Technobabble

The setup

You will need a short Python "bootstrap" script to create a WSGI-handler for your Django project. Here is an example (call it wsgi_handler.py and place it in the root directory of your Django project - the one with manage.py and settings.py):

import sys
import os

sys.path.append(os.path.dirname(os.path.abspath(__file__)) + '/..')
os.environ['DJANGO_SETTINGS_MODULE'] = 'projectname.settings'

import django.core.handlers.wsgi

application = django.core.handlers.wsgi.WSGIHandler()

Finally set up your Apache virtualhost to use mod_wsgi:

<VirtualHost *>

  ServerName www.projectname.org
  ServerAlias *projectname.org

  Alias /admin_media /usr/lib/python2.4/site-packages/django/contrib/admin/media

  <Location /admin_media>
    Order allow,deny
    Allow from all
  </Location>

  Alias /media /home/user/projectname/media

  <Location /media>
    Order allow,deny
    Allow from all
  </Location>

  WSGIScriptAlias / /home/user/projectname/wsgi_handler.py

  WSGIDaemonProcess projectname user=user group=user processes=1 threads=10
  WSGIProcessGroup projectname

</VirtualHost>

In the WSGIDaemonProcess line, you can easily manage the amount of system resources (measured in processes and threads) mod_wsgi should use. In my experience a single process with 10 threads will cover most small to medium loaded websites.

Why?

This is some of the reasons why you should ditch mod_python for mod_wsgi when hosting Django projects:

Faster

The load times of the websites now served with mod_wsgi really surprised me. Normally a page would be served within 150-300 ms. This was reduced to load times in the range of 40-80 ms.

I also discovered that running mod_wsgi in embedded mode (as opposed to daemon mode) was not worth the effort. I didn't really see any difference between load times when using Django.

Less memory usage

Everyone hosting more than a couple of Django projects on a single Apache instance knows that Django projects squanders a bit with memory usage, and every single Apache child process will easily end up using 50 MB RAM.

mod_wsgi dedicates a process (or multiple processes) to a single interpreter for a single Django project, and keeps the memory usage low in the "normal" Apache child processes. On a server with 8 small Django projects, I went from using ~1500 MB RAM on Apache child processes to using 150 MB.

Secure

When using mod_python your Python interpreter will be running as the user running the Apache webserver itself (on Debian systems, the user is called www-data). Typically this will allow you to peek around in places where you do not want your users peeking. This is due to the fact that www-data must have read access to every file you use in your application (including settings/configuration/media files).

mod_wsgi addresses this problem by changing to a user id specified in the configuration file, and run your Python interpreter as another user than www-data, allowing you to lock down every project on your server to seperate user accounts.

These points cover mod_wsgi running in daemon mode.

Conclusion

mod_wsgi rocks!

So if you are thinking about moving your systems to, or just curious about, mod_wsgi, you should really get to it. I, for one, welcome our new mod_wsgi overlords! (sorry)

Futher reading

• How to use django with mod_wsgi (from the Django wiki)
• Integration With Django (from the mod_wsgi wiki):