earn the White PyBites Ninja earn the Yellow PyBites Ninja earn the Orange PyBites Ninja right arrow earn more PyBites Ninja belts and certificates
The best way to learn to code in Python is to actually use the language.

Our platform offers effective Test Driven Learning which will be key to your progress.


Join thousands of Pythonistas and start coding!


Join us on our PyBites Platform
Click here to code!

Deploying a Django App to PythonAnywhere

Posted by Bob on Sun 23 July 2017 in Django • 3 min read

After Julian's great article on deploying a Flask app to Heroku, let's look at how we can deploy a Django app to PythonAnywhere (PA). In this article I share a few things I learned deployed our first Django app.

Good docs + nice interface

One of my first tries was the API / helper script, but unfortunately it did not make it to the end. Yet the Deploying an existing Django project on PythonAnywhere using the Manual option worked great for me:

pa choose manual option

Our first Django app is online:

pa site online

22nd of Jan 2018 update: we ended up hosting this on Heroku.

I really like the infrastructure of browser consoles and intuitive GUIs. Also config files like wsgi.py were clearly commented so setting it up was quick and almost painless.

Git + venv

Important steps of the deployment steps are git pulling your code and creating a virtual env, this worked very well.

pa supports git and venv

After this step I could just do a pip install -r requirements.txt to pull down Django and feedparser.

Now when I make changes to my app I can just do a git pull in the repo dir and restart the app in the browser. It just takes seconds :)

Scheduled tasks ... OK time to upgrade

Our app pulls in new articles from Planet Python as explained here. Planet was not on PA's whitelist and I wanted this task to be run every hour (instead of once a day). At this point I had to upgrade from Free to the Hacker tier. This has additional benefits, check out pricing.

Similar to other providers, at PA you pay for what you need/ consume. You can add apps on the fly. I'd hoped to get a small PostgreSQL DB with Hacker's tier, but that requires further upgrading ...

The interface to set up a scheduled task is nice and easy:

pa scheduled tasks

Notice that I activate the venv in the command because it needs to load env variables.

Less obvious

  • I had to add our PA domain pybites.pythonanywhere.com to ALLOWED_HOSTS (Django settings).

  • Django encapsulation. As detailed in this excellent article you want to hide your SECRET_KEY, DB credentials, etc from version control. I also followed the settings best practices described in the article. So make sure you do some work upfront. Make sure you check Django's checklist.

  • Env variables. You need to set them in 2 places. As the help doc admits this is not ideal. At the virtual env level it only seemed to work adding them to the activate script, not postactivate.

  • Handling static files was a bit of a pain. I ended up using collectstatic to get them all in one place:

    (venv) 12:27 ~/pybites-django/pybites (master)$ python manage.py collectstatic
    62 static files copied to ...
    

    And in the web GUI config I set:

    pa static files solution

    Not sure if this is the best solution because I need to rerun this when static files change ...

Multiple apps

As the Hacker tier gives me one app I made this container project / repo to host multiple apps in. It's a nice exercise in Django's architecture of one project -> multiple apps that can be moved around. As said, PA's pricing structure is pretty flexible, so we can always add apps if necessary.

Conclusion

So far I am happy with PA. In spite of some minor issues it is easy to deploy a Django app and it performs well. The help docs are well written.

They are also quite proactive: there was a file server issue last weekend which they updated via Twitter. And they just posted a full post mortem!

It was easy to reach out to support and I got a quick response. I also like the short email tutorial updates.

I will compare with Heroku pricing for PostgreSQL because adding one bumps the monthly price up from 5 to 12 USD (at this time of writing). Or I go with their default MySQL which is free.


Keep Calm and Code in Python!

-- Bob


See an error in this post? Please submit a pull request on Github.