What is the most common way to configure static files in debug and production for Django

Question

When developing a Django application in debug mode, I serve static files using the following code:

if settings.DEBUG:
    urlpatterns += patterns('',
        (r'^m/(?P<path>.*)$', serve, {
            'document_root' : os.path.join(os.path.dirname(__file__), "media")
        })
    )

I am using nginx as a front end to server my static files in production mode using the following nginx config:

  location m/
      {
           root /path/to/folder/media/;
      }

Which seems sub-optimal because I have to create a "m" folder in the media directory. I am wondering what everyone else's Django/nginx config files look like. Specifically, can you please cut-and-paste sections of nginx.confg and urls.py (settings.DEBUG == True)

Thanks.

Answer 1

Using Django 1.3 django.contrib.staticfiles will take care of serving everything for you during development. You don't need to do anything particular in the urls.py. I wrote a little guide for myself after the Django 1.3 update that covers the settings to use:

# idiom to get path of project
import os
PROJECT_PATH = os.path.dirname(os.path.abspath(__file__))
# url prefix for user uploaded files, stuff that django has to serve directly
MEDIA_URL = '/media/'
# url prefix for static files like css, js, images
STATIC_URL = '/static/'
# url prefix for *static* /admin media
ADMIN_MEDIA_PREFIX = STATIC_URL + 'admin/'
# path to django-served media
MEDIA_ROOT = os.path.join(PROJECT_PATH, 'media')
# path used for collectstatic, *where the webserver not django will expect to find files*
STATIC_ROOT = '/home/user/public_html/static'
# path to directories containing static files for django project, apps, etc, css/js
STATICFILES_DIRS = (
    os.path.join(PROJECT_PATH, 'static'),
)
# List of finder classes that know how to find static files in various locations.
STATICFILES_FINDERS = (
    'django.contrib.staticfiles.finders.FileSystemFinder',
    'django.contrib.staticfiles.finders.AppDirectoriesFinder',
) 
# Required for all the magic
INSTALLED_APPS = (
    'django.contrib.staticfiles',
)

Refer to the docs for details: http://docs.djangoproject.com/en/1.3/howto/static-files/.

I use nginx and uwsgi for serving django apps in production (I use runserver for development). I symlink my /static and /media folders (from my django project) into /var/www/vhosts/domain.com/html for nginx to find. You could also use the collectstatic command instead of symlinking. If it can't find a static file it falls back to uwsgi (which is running the django app).

Instead of uwsgi you could use fast-cgi, or proxy_pass or whatever you want. I prefer uwsgi because it has an incredible number of features and great performance. I run uwsgi as a daemon with: uwsgi --emperor '/srv/*/*.ini'. This is a fairly new option, it tells uwsgi to scan a given path for configuration files. When the emperor uwsgi daemon finds a configuration file it launches a new instance of uwsgi using the configuration found. If you change your configuration the emperor uwsgi daemon will notice and restart your app for you. You can also touch the config file to reload like with mod_wsgi, and it's really easy to setup new apps once you have everything configured initially.

The path conventions I follow are:

/srv/venv/ - virtualenv for project
/srv/venv/app.ini - configuration for uwsgi
/srv/venv/app.sock - uwsgi sock for django
/srv/venv/app.wsgi - wsgi file for uwsgi
/srv/venv/proj - django project
/srv/venv/proj/settings.py - project settings file
/srv/venv/proj/static - static files dir, linked into var/www/vhosts/domain.com/html
/srv/venv/proj/static/admin - admin static files, linked as well
/srv/venv/proj/media - media files dir
/var/www/vhosts/domain.com/html - base directory for nginx to serve static resources from

This is my nginx.conf:

location / {
    root /var/www/vhosts/domain.com/html;
    index index.html index.html;
    error_page 404 403 = @uwsgi;
    log_not_found  off;
}

location @uwsgi {
    internal;
    include /etc/nginx/uwsgi_params;
    uwsgi_pass unix:/srv/venv/app.sock;
}

My uwsgi ini file (you can also use xml/yaml/etc):

[uwsgi]
home = /srv/%n
pp = /srv/%n
wsgi-file = /srv/%n/%n.wsgi
socket = /srv/%n/%n.sock
single-intepreter = true
master = true
processes = 2
logto = /srv/%n/%n.log

You should also check out gunicorn, it has really nice django integration and good performance.

Answer 2

I'm putting this here in case someone wants an example of how to do this for Apache and WSGI. The question title is worded such that it's not just covering nginx.

Apache/WSGI Daemon

In my deployment, I decided to keep the database connection info out of the settings.py file. Instead I have a path /etc/django which contains the files with the database configuration. This is covered in some detail in an answer to another question. However, as a side effect, I can check for the presence of these files and the project being in a certain path to determine if this is running as a deployment, and in settings.py I define the settings IS_DEV, IS_BETA, and IS_PROD as True or False. Finding the project's directory from settings.py is just:

# Find where we live.
import os
BASE_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), os.pardir))

Anything that needs a path into the project uses BASE_DIR. So in urls.py, I have at the end:

# Only serve static media if in development (runserver) mode.
if settings.IS_DEV:
    urlpatterns += patterns('',
        url(r'^static/(?P<path>.*)$', 'django.views.static.serve', 
            {'document_root': settings.MEDIA_ROOT, 
            'show_indexes': True}),
    )

(I also have another URL in there that I use for UI testing and don't want on beta or production.)

This covers the development server case. For production, I just have to set the Apache configuration up to serve the static stuff. (This is an intranet application with low to medium load, so I don't have a lightweight webserver like lighttpd to serve the static stuff, contrary to the Django docs' recommendation.) Since I'm using Fedora Core, I add a django.conf file in /etc/httpd/conf.d that reads similar to:

WSGIDaemonProcess djangoproject threads=15
WSGISocketPrefix /var/run/wsgi/wsgi

Alias /django/static/ /var/www/djangoproject/static/
Alias /django/admin/media/ /usr/lib/python2.6/site-packages/django/contrib/admin/media/
WSGIScriptAlias /django /var/www/djangoproject/django.wsgi
WSGIProcessGroup djangoproject

<Directory /var/www/djangoproject>
    Order deny,allow
    Allow from all
</Directory>

<Directory /usr/lib/python2.6/site-packages/django/contrib/admin/media>
    Order deny,allow
    Allow from all
</Directory>

IIRC, the key here is to put your Alias lines before your WSGIScriptAlias line. Also make sure that there's no way for the user to download your code; I did that by putting the static stuff in a static directory that is not in my Django project. That's why BASE_DIR gives the directory containing the Django project directory.

You can omit the WSGISocketPrefix line. I have it because the admin wants the sockets in a non-default location.

My WSGI file is at /var/www/djangoproject/django.wsgi (i.e. /django.wsgi in the Mercurial repository) and contains something like:

import os
import sys

os.environ['DJANGO_SETTINGS_MODULE'] = 'djangoproject.settings'
os.environ['DB_CONFIG'] = '/etc/django/db_regular.py'
thisDir = os.path.dirname(__file__)
sys.path.append(thisDir)
sys.path.append(os.path.join(thisDir, 'djangoproject'))
sys.path.append(os.path.join(thisDir, 'lib'))

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

The nice thing about WSGI daemons is that you just have to touch django.wsgi to restart your Django WSGI daemon; you don't need to reload or restart the Apache server. This makes the admin happy.

Finally, since my /var/www/djangoproject is just a Mercurial repo, I have the following in /var/www/djangoproject/.hg/hgrc:

[hooks]
changegroup.1=find . -name \*.py[co] -exec rm -f {} \;
changegroup.2=hg update
changegroup.3=chgrp -Rf djangoproject . || true
changegroup.4=chmod -Rf g+w,o-rwx . || true
changegroup.5=find . -type d -exec chmod -f g+xs {} \;
changegroup.6=touch django.wsgi # Reloads the app

This clears Python bytecode, updates the working copy, fixes up all the perms, and restarts the daemon whenever a developer pushes into deployment, so anyone in the djangoproject group can push into it and not just the last one who added a file. Needless to say, be careful who you put in the djangoproject group.

---

zeekay sets STATIC_URL, STATIC_ROOT, STATICFILES_DIRS, STATICFILES_FINDERS and uses "django.contrib.staticfiles" and "django.core.context_processors.static" in his settings. I don't have those since my code dates back to Django 1.1 and don't use {{ STATIC_ROOT }}.

Hope this helps.