Although this is certainly not the usual use case for a Django site, I am deploying a new site that will be used by several programs at the school where I work. Each program will have its own specific Django settings file with its own values (for example, the database connection info) but for the most part the sites share a base file. Here is a sample of this file:

from mysite.settings import *

DEBUG = False





        'default': {
                'ENGINE': 'django.db.backends.postgresql_psycopg2',
                'NAME': os.environ.get('MYSITE_DB_NAME'),
                'USER': os.environ.get('MYSITE_DB_USER'),
                'PASSWORD': os.environ.get('MYSITE_DB_PASSWORD'),
                'HOST': os.environ.get('MYSITE_DB_HOST'),
                'PORT': '5432',

As you can see, we get the database connection values from the environment.

I am using Apache (with mod_proxy_uwsgi) and uwsgi in emperor mode to run the site, and each site has a uwsgi ini file that actually sets all the relevant per-site settings via the env configuration option. This was quite easy to set up, and this is an example uwsgi ini file:

env = DJANGO_SETTINGS_MODULE=mysite.settings_production
env = MYSITE_DB_NAME=mysite_dbname
env = MYSITE_DB_USER=mysite_username
env =

So far so good -- we can run uwsgi in emperor mode, point it at our conf/ directory full of ini files, and everything works fine.

Now, what if we want to use shell? We have fully factored these settings out of the file and we can't rely on envdir because the uwsgi emperor won't know how to use it.

The answer (for me anyway) is fabric and python's ConfigParser module. With a little work we can actually write helper functions in our fabfile that use ConfigParser objects to read the uwsgi ini files and set the appropriate environment variables before calling

The important wrinkle here is that the env setting is used several times in the ini file, while the default behavior of ConfigParser is to use a simple dict to store each option found in the ini file, and if an option is specified multiple times, only the last one is saved. We can override this however by specifying our own dict type to hold the option values:

from fabric.api import task, shell_env, local
from collections import OrderedDict
import ConfigParser

def shell(site_name):
    filename = 'conf/uwsgi-%s.ini' % (site_name,)
    env_settings = parse_uwsgi_ini_env_settings(filename)
    with shell_env(**env_settings):
            local('python shell --settings=mysite.settings_production')

class MultiOrderedDict(OrderedDict):
    def __setitem__(self, key, value):
        if isinstance(value, list) and key in self:
            super(OrderedDict, self).__setitem__(key, value)

def parse_uwsgi_ini_env_settings(filename):
    env_settings = {}
    config = ConfigParser.RawConfigParser(dict_type=MultiOrderedDict)
    for env_setting in config.get('uwsgi', 'env'):
        key, val = env_setting.split('=', 1)
        env_settings[key] = val

    return env_settings

Now rather than run python shell, run fab shell:foo your environment variables will be read from conf/uwsgi-foo.ini and set properly.


comments powered by Disqus