The current state of deploying Django applications

written on Monday, December 16, 2013

I've recently tried to release a Django application in a small intranet. It was hell all over again. I'm not very experienced with nginx, uwsgi and gunicorn, so I always have to read a lot.

The fact is: Deploying Django in 2013 is hard! As a guy who writes 2-3 small web apps a year, I don't want to fight with nginx and gunicorn. I just want to use something like deploy in my Django directory.

So I decided to write my own (and yet) little deployment software. If people like it, I'm willing to invest some of my time. I would be very happy to receive feedback on how we could create an API that makes deploying simple things really easy!

PHP Goodness?!

Let's face it, PHP is horrible. But the PHP eco system has always done one thing fairly well: Deploying is easy. You can argue that Nginx is better than Apache, but still all the new sexy systems (Django/Rails/Node/Flask) are far more complicated to set up than apt-get install php5, if you want to deploy to your own server.

Now - I don't want (old) PHP deployment back. There's often no easy way of making a difference between debug and production. People screw up dependencies and permissions. But let's face it. PHP has something really valuable there for beginners and "app programmers" that don't like dealing with systems, especially if they don't like Linux (e.g. WAMP people).

Solution: Travis-ci?

Anybody using Travis-ci? Awesome, right? It just solves testing for you. That is how I imagine depl to be. Travis also does something like a deployment. Your code is completely running on their servers and can be tested. In comparison to e.g. Heroku (which is also pretty cool), travis uses a more explicit way to deploy. I like travis way better, because travis yaml files allow for a more flexibility and complete control. I also like to know what's going on.

My solution provides defaults for typical use cases. It leaves more complex use cases to awesome software like ansible. If you really need 10 different databases, custom paths, etc, you shouldn't be using depl.

Some Deployment Options

There are a variety of tools to deploy, but non of them really fit the simple use case. I don't want dozens of files just for deploying my very simple Django project. I want at most one (something like travis). But to give you an overview:

  • ansible is awesome for deploying big projects, you could check out this for a Django deployment with ansible.
  • chef/puppet do the same thing. But I like ansible with its .yml files better somehow. It's a cleaner approach IMHO.
  • docker is awesome, but you still have to deploy your software into docker containers - that's where depl could step in.
  • vagrant is cool(for development), but again - how do you deploy in the first place. And also I wouldn't trust virtualbox for deploying things.

I have analyzed those options and Platform as a Service providers briefly. I realized that there is nothing to deploy Django easily (in an in-house setting), so I started creating depl. Not just for Django. For all the frameworks out there that still struggle with deployment.

On a Side Note: Similar ideas to depl

I've found two projects that are similar to what I imagine a good deploy tool should be like. The first one is capistrano for ruby/rails. My problem with capistrano is that I don't really understand it - but I also don't really understand the ruby world, so I won't complain here. Something just smells bad, because the tool can not be explained in one page.

Quite with the same problems struggles the second tool "wercker". It's a little bit too hard to explain, because it also includes a testing/deploying/whatever platform. It seems to be a very complete and complicated model - too complicated for a normal open source developer. Might be interesting for professionals deploying web applications very often.

However both really haven't inspired me. The only positive inspiration is travis and the fact that all things should be easy in life. :-)

Thank you for reading! You can follow me on github or twitter. If you like my OS projects, you could consider a donation on gittip.

This entry was tagged depl, deploy, django and python

blog comments powered by Disqus