October 29, 2011
Django is a very cool web development framework which comes with its own lightweight development server which is best for testing the application. But when it comes to deploying the application for the outer world, more efficient and secure web server like Apache is preferred. Now, web developers who have been developing applications using PHP might find it very easy to deploy their applications on Apache(as they have easy configurable PHP module, or they could just find a LAMP stack for if they were on Linux, so they would get everything cooked. But when it comes to combining python web frameworks like Django with Apache, things turn out to be less obvious. My experience weren’t smooth at all while trying to get my Django project work on Apache. And since I was trying to configure it with already installed LAMP stack on my Ubuntu system, things went really bad. Since, there are so many things integrated in a compilation like LAMP stack, there are as many things which can go wrong while configuring Django on an Apache server. So, its recommended that a clean and separate Apache installation should be used so that things are easier to diagnose one by one.
You should install a clean build of Apache in case you haven’t already. Its easy, just follow the docs given here http://httpd.apache.org/docs/2.2/instal l.html
Now that you’re done with Apache, you need to install mod_wsgi which is a WSGI module for Apache. Know more about it here. http://code.google.com/p/modwsgi/
Installation instructions http://code.google.com/p/modwsgi/wiki/InstallationO nLinux
This is where you edit the Apache configuration file and tell Apache to load mod_wsgi module using the DSO method. You need to edit /conf/httpd.conf file and add the following directive to it
LoadModule wsgi_module /usr/lib/apache2/modules/mod_wsgi.so
Also, at the end of the httpd.conf file, add the following directives
WSGIScriptAlias / <path-to-your-django-project>/test.wsgi <Directory path-to-your-django-project> Order deny,allow Allow from all </Directory>
The first line in the above directives loads mod_wsgi module into the Apache
instance. The second directive
Now that we’ve configured Apache to load the mod_wsgi module, its time to test
if it works. To do that, create a file
def application(environ, start_response): status = '200 OK' output = 'Hello World!' response_headers = [('Content-type', 'text/plain'), ('Content-Length', str(len(output)))] start_response(status, response_headers) return [output]
Now test if it works by trying http://localhost in your browser, you should get Hello World for obvious reasons.(and yes make sure Apache is running)
Now, that the simple WSGI application runs, we need to test something which generally has issues which need to be resolved before moving on to using Django as a WSGI application.
When we move to more complicated WSGI applications, and as the applications start using libraries whose versions conflict with the versions used by Apache, there are issues like Apache crashing and incorrect response. One such library is libexpat. Know more about it here http://expat.sourceforge.net/.
It happens in many situations that the expat module used by python(that is the default one on a system) has a different version than that used by Apache(it has its own). In such a situation, there is Segmentation Fault encountered by Apache and it crashes. To test it if you encounter that issue, just make the following entry in your test.wsgi file.
import pyexpat
Now, again try http://localhost. If you get proper response as earlier then you are free from the expat issue while if you get ‘no response’ error then you need to resolve the expat issue before turning on to Django.
There is this great post on handling expat issue. http://code.google.com/p/mod wsgi/wiki/IssuesWithExpatLibrary.
Since, python’s expat module is hard to replace, its recommended you change Apache’s expat module. Either update it or simply replace it with your system’s expat. I consider replacing it. After all done, restart Apache and test http://localhost, if all went good it should be working fine.
If you’ve successfully dealt with running WSGI with expat loaded, you should then be able to run Django after a little bit of work if not-so-common issues don’t fall in your way. For connecting to Django with mod_wsgi, you should create a new django.wsgi in the same directory as test.wsgi and set WSGIScriptAlias to that very file. Enter the following code to django.wsgi and you’re on the go.
import os import sys root = os.path.join(os.path.dirname(__file__),'..') sys.path.insert(0,root) os.environ['DJANGO_SETTINGS_MODULE']='<your-project>.settings' import django.core.handlers.wsgi application = django.core.handlers.wsgi.WSGIHandler()
Replace <your-project>
with your project’s name.
So you’ve worked out all the way to the point where you would need to test if all your setup works fine. Its recommended that you use a clean Django application so that you don’t encounter issues related to database and other stuff.