[Tags]

In the beginning there was Cherokee...

Previously, Cherokee powered this website. Cherokee is a fast server with lots of interesting features. But it lacks a clean, text based configuration file; the full configuration is web based.

Fab is used to deploy this website. It makes it possible to deploy locally (on my laptop), on a staging server (a server I've running) and on production (the server you're connected to now). But it is quite a nuisance to update the configuration parameters of Cherokee on all these environments using a fabfile. So I decided a move from Cherokee was required.

First I've checked several servers, like lighttpd, litespeed and nginx. After trying and checking some reviews and benchmarks I stuck with nginx.

...and FastCGI

When deploying a Django project on a webserver, the server needs to be able to forward the Python requests to a Python environment. There are many gateway protocols, like CGI or FastCGI which do this, but they do have performance drawbacks.

The Python community came up with a standard interface between web servers and a Python environment. It is called WSGI. See PEP 333 and (for Python 3) PEP 3333 for a detailed technical explanation.

There are many implementations of this interface, but after some research I chose for uWSGI.

Then we went for nginx and uWSGI!

In this entry I'll explain how my nginx and uwgsi servers are installed. Hope it is handy for you.

Install nginx and uwsgi

To install nginx on a debian based system:

user@box:~$ sudo apt-get install nginx

To start it and to get its status:

user@box:~$ sudo service nginx start
Starting nginx: nginx.
user@box:~$ sudo service nginx status
* nginx is running

To install uwsgi (and its python plugin) also is very simple process:

user@box:~$ sudo apt-get install uwsgi uwsgi-plugin-python

Configure nginx

To configure nginx to handle your Django website and the related STATIC_ROOT, STATIC_URL, MEDIA_ROOT and URL directories, create this configuration file in /etc/nginx/sites-available.

server {

    listen 80;
    server_name elevenbits.com, www.elevenbits.com, elevenbits.org, www.elevenbits.org;

    access_log /var/log/nginx/elevenbits.access.log;
    error_log /var/log/nginx/elevenbits.error.log;

    location /static/ { # STATIC_URL
        alias /var/www/elevenbits/static/; # STATIC_ROOT
        expires 30d;
    }

    location /media/ { # MEDIA_URL
        alias /var/www/elevenbits/static/; # MEDIA_ROOT
        expires 30d;
    }

    location / {
        include uwsgi_params;
        uwsgi_pass 127.0.0.1:3031;
    }

}

You can name file as you like; I use localhost.conf when nginx runs on my local machine, elevenbits.conf when it runs on production.

I guess you'll understand the configuration file. The only hard part is the "location /" section where a link is given to uWSGI:

Configure uwsgi

The uWSGI starts one (or more) severs which will handle requests from nginx. The interface to these uWSGI servers is specified in the "location /" section of the nginx configuration.

The uWSGI configuration is very simple. This file needs to be placed in /etc/uwsgi/apps-available:

[uwsgi]
socket = 127.0.0.1:3031
uid = www-data
gid = www-data
chdir = /var/www/elevenbits
module = elevenbits.uwsgi

You can call this file whatever you like. I just called it django.ini.

Note that you can also specify this file in xml format, or yaml format. See the documentation for detailed information.

Lastly create some symbolic links and restart both servers

user@box:~$ ln -sf /etc/nginx/sites-available/elevenbits.conf /etc/nginx/sites-enabled/elevenbits.conf
user@box:~$ ln -sf /etc/uwsgi/apps-available/django.ini /etc/uwsgi/apps-enabled/django.ini

Best to make sure that the default site is removed from the nginx enabled folder. Now, when you restart uwsgi and nginx, your system will be online!

The cherokee-admin binds only to local loopback by default. There are some workarounds:

Tunneling

Create an SSH tunnel. This is the recommended way. In order to do so you must issue the following command:

user@box:~$ ssh -L 9090:localhost:9090 remote_IP

After that your terminal will be logged in the remote machine. There you can start the admin. From then on you can access this remote interface through http://localhost:9090. Every request will be forwarded to the remote IP running cherokee-admin.

The -b parameter

Launch cherokee-admin with the -b parameter in order to force it to listen to all the network interfaces.

Copy and paste

Finally you could always install cherokee on your local host, configure it there and then copy the generated cherokee.conf file to the device running the cherokee instance you wanted to set up.

To redirect http://a.mysite.com to http://mysite.com:8000 follow these steps in your Cherokee admin web environment:

• First go to the vServers tab and create a virtual server with name a.mysite.com and give it a document root, e.g. /var/www.

• Then link to the sources tab and create a new information source by clicking the 'plus' icon. Enter a.mysite.com as nick and mysite:8000 as connection.

• Go back to the vServers tab. After making sure that the a.mysite.com server is selected check that the host match method is set to 'match nickname'.

• In the behaviour tab select the default directory and in the handler tab select HTTP Reverse Proxy from the handler dropdown list. Then select Round Robin as balancer in the Back-end Server section.

• In the section Assign Information Sources section select a.mysite.com as application server.

• Then save and restart Cherokee.

Now the http://a.mysite.com will be redirected automatically to http://mysite.com:8000. And, when your site becomes more popular, you can add more application servers to mysite.com.