Getting Django 1.7 to work on Google App Engine

Question

Can anyone help to point us to instructions on how to get Django >1.5 working on Google App Engine? I have seen a number of people claim they have Django 1.6 working. We'd like to get 1.6 or 1.7 running. I have searched here for instructions on how to set this up. No luck so far.

Update: In our development machine we have Django 1.7 installed (both /user/local and on virtualenv). However, if we modify GAE yaml to use Django 1.7 we get the following error messages:

google.appengine.api.yaml_errors.EventError: django version "1.7" is not supported, use one of: "1.2", "1.3", "1.4", "1.5" or "latest" ("latest" recommended for development only)   in "./app.yaml",

The version 1.9.12 GoogleAppEngine sdk install in our /Applications/GoogleAppEngineLauncher.app/Contents/Resources/GoogleAppEngine-default.bundle/Contents/Resources/google_appengine/lib directory shows the following Django versions listed:

1. django-0.96 django-1.2 django-1.3 django-1.4 django-1.5

My question is related to how to get our development environment setup correctly for Django 1.7 on Google App Engine and how to make sure we successfully deploy our app with Django 1.7 when we deploy to Google App Engine in production. How do we get the Django 1.7 to deploy on GAE when we deploy our app?

Answer 1

You can use any pure Python third party libraries in your Google App Engine application. In order to use a third party library, simply include the files in your application's directory, and they will be uploaded with your application when you deploy it to our system. You can import the files as you would any other Python files with your application.

I have application using Django 1.7 this way and everything is working fine. However, sometimes you may need to sort of hack something due to the App Engine limitations and its specifics. But it depends on your use cases.

I would also suggest to use virtual environment for your project. Install each library that is not supported by App Engine directly via pip and then create a symlink in your application directory pointing to the given library.

This way you can keep all required packages in a file (e.g. requirements.txt) that can be stored in SCM system (e.g. Git) along with your source files and other team members can quite easily replicate your working environment.

Provided that you use virtual environment and install all needed libraries (Django, ...) via pip, here is the directory layout that should work for you.

• virtual-env-root
  • .Python
  • bin
  • include
  • lib
  • app-engine-project-root
    • app.yaml
    • django-project-root
    • django-app-root
    • symlink-to-django -> lib/python2.7/site-packages/django
    • symlink-to-another-lib -> lib/python2.7/site-packages/...

Such a layout can be easily deployed with the below command.

$ appcfg.py update app-engine-project-root

Or tested with App Engine development server.

$ dev_appserver.py app-engine-project-root

UPDATE

Since App Engine Python SDK version 1.9.15 you can use the vendoring mechanism to set up third party libraries. You do not have to create symlinks in your application directory pointing to the Python lib folder anymore.

Create lib directory directly in your application root and tell your app how to find libraries in this directory by means of appengine_config.py file.

from google.appengine.ext import vendor

# Add any libraries installed in the "lib" folder.
vendor.add('lib')

New directory layout follows.

• virtual-env-root
  • .Python
  • bin
  • include
  • lib
  • app-engine-project-root
    • lib
    • app.yaml
    • appengine_config.py
    • django-project-root
    • django-app-root

Use pip with the -t lib flag to install libraries in this directory.

$ pip install -t lib [lib-name]

Or

$ pip install -t lib -r requirements.txt

Answer 2

App Engine's Python environment currently knows how to provide Django up to version 1.5 via the libraries: configuration mechanism. This doesn't mean that later versions of Django won't work, only that they aren't yet built in. (I'm not sure why the latest built-in version is 1.5. It may have something to do with AE's historical policy of bundling each supported version of Django with the SDK, which probably needs to be revised to keep the SDK from getting too large.)

You can try to include Django 1.7 with your application files. I haven't tried this with 1.7 specifically yet, but it's worked with previous versions. Some adjustments to sys.path will be needed in your main.py.

Note that there is a limit of 10,000 application files. If you're concerned about this limit, one option is to use Python's zipimport and include Django as a zip archive. https://docs.python.org/2/library/zipimport.html

Answer 3

You cannot - GAE only supports 1.5, and even that is marked as experimental. If you need django 1.7, perhaps you should use Google Compute Engine, which is Google's brand name for virtual machines that you can spool up.

If you are not married to Google App Engine, Heroku supports django 1.7 without issues.

---

Do you have specific a guide on how to move a Django 1.7 project to Google Compute Engine? There is a bunch of Google stuff without any guides on how to make them work.

Here are the steps, but they are the same had you deployed on any other server because GCE just gives you a linux instance:

1. First, make sure your developer account has a billing method attached to it.
2. Go to the developer console
3. Create a new project by clicking on Projects, then Create Project.
4. Wait as the project is being created (you'll see a progress window on the bottom right of your screen).
5. Once the project is finished creating, the console will automatically shift to that project's settings:

6. You can create a new instance, or deploy a ready-made template from the second column. You can see there are popular stacks and software applications for which templates are created.

7. As there is no django template yet, you will start by creating an instance.

8. Billing is controlled on a per-project basis, so you'll have enable billing at this point if you haven't done so already.

9. The next page is where you configure the instance. The fields are self-explanatory. You set the type of machine you like (how many virtual CPUs and memory), where (physically) you prefer the machine to be located, if you want both HTTP and HTTPS ports open, and then a disk image from which the instance will boot:

Once you have configured the machine, it will be brought online booted up and then you'll have access to the terminal via SSH.

From this point forward, you should treat this instance like any linux server. Install whatever you need to make your project work using the normal packaging tools; upload your files, etc.

For Amazon, the process is a bit simpler as there is a large library of AMIs that you can use for a one-click deployment process. AMI is Amazon Machine Image - a template from which you can deploy an instance.

For Heroku, as its a PaaS, you don't have to worry about the hardware components; however as with most PaaS platforms, you don't have write access to the filesystem. So to manage your static assets you have to do some extra work. The easiest option is to create a S3 bucket on Amazon and use that with django-storages. The official django tutorial at heroku suggests the use of dj-static to serve files directly from Heroku. This works fine for testing, but if you want to start uploading files, then you need to handle those correctly.

However, once you sort that out the steps are even simpler:

Pre-requisites:

• git
• heroku toolbelt
• dj-database-url Python package
• gunicorn Python package

The basic steps:

1. Create a git repository (if you have not done already) in your source code directory with git init.
2. Create a requirements.txt at the root of your project. pip freeze > requirements.txt should do it if you are using a virtual environment. Otherwise, you can create a text file and list the packages you need.
3. Adjust your settings.py, by adding this line at the very bottom: import dj_database_url DATABASES['default'] = dj_database_url.config()
4. Create a Procfile (case is important). This is how you tell Heroku what kind of dyno (process) you need for your application. For django, you need a web dyno so in this file the following line should do: web: gunicorn yourproject.wsgi --log-file -

Create an app on Heroku and deploy. You should run these commands from your source code directory:

1. heroku create --buildpack https://github.com/heroku/heroku-buildpack-python
2. heroku addons:add heroku-postgresql:dev
3. git push heroku master
4. heroku run python yourproject/manage.py migrate --noinput
5. heroku run python web/manage.py collectstatic

You only do the first two steps once, then whenever you need to update your application simply git push heroku master to create a new revision on Heroku.