django-herokuapp is a set of utilities and a project template for running Django sites on Heroku.
Heroku provides a guide that describes a reasonable Django project setup. The django-herokuapp project suggests a more advanced approach with the following benefits:
- waitress is used as an app server instead of gunicorn. Gunicorn is not recommended for use as a public-facing server, whereas waitress is hardened for production use.
- Amazon S3 is used to serve static files instead of django-static. Django does not recommend serving static files using a Python app server.
- Various minor security and logging improvements.
If you're creating a new Django site for hosting on Heroku, then you can give youself a headstart by running
the herokuapp_startproject.py script that's bundled with this package from within a fresh virtual environment.
$ mkdir your/project/location $ cd your/project/location $ git init $ virtualenv venv $ source venv/bin/activate $ pip install django-herokuapp $ herokuapp_startproject.py <your_project_name>
The rest of this guide describes the process of adapting an existing Django site to run on Heroku. Even if
you've started from scratch using herokuapp_startproject.py, it's still worth reading, as it will
give you a better understanding of the way your site has been configured.
- Install django-herokuapp using pip
pip install django-herokuapp. - Add
'herokuapp'to yourINSTALLED_APPSsetting. - Read the rest of this README for pointers on setting up your Heroku site.
A site hosted on Heroku has to handle traffic without the benefit of a buffering reverse proxy like nginx, which means that the normal approach of using a small pool of worker threads won't scale in production, particularly if serving clients over a slow connection.
The solution is to use a buffering async master thread with sync workers instead, and the waitress project provides an excellent implementation of this approach.
Simply create a file called Procfile in the root of your project, and add the following line to it:
web: waitress-serve --port=$PORT <your_project_name>.wsgi:application
Heroku provides an excellent Postgres Add-on that you can use for your site. The recommended settings for using Heroku Postgres are as follows:
import dj_database_url
DATABASES = {
"default": dj_database_url.config(default='postgres://localhost'),
}
This configuration relies on the dj-database-url package, which is included in the dependencies for django-herokuapp.
You can provision a starter package with Heroku Postgres using the following Heroku command:
$ heroku addons:create heroku-postgresql:dev
A pure-python webserver like waitress isn't best suited to serving high volumes of static files. For this, a cloud-based service like Amazon S3 is ideal.
The recommended settings for hosting your static content with Amazon S3 is as follows:
# Use Amazon S3 for storage for uploaded media files.
DEFAULT_FILE_STORAGE = "storages.backends.s3boto.S3BotoStorage"
# Use Amazon S3 for static files storage.
STATICFILES_STORAGE = "require_s3.storage.OptimizedCachedStaticFilesStorage"
# Amazon S3 settings.
AWS_ACCESS_KEY_ID = os.environ.get("AWS_ACCESS_KEY_ID", "")
AWS_SECRET_ACCESS_KEY = os.environ.get("AWS_SECRET_ACCESS_KEY", "")
AWS_STORAGE_BUCKET_NAME = os.environ.get("AWS_STORAGE_BUCKET_NAME", "")
AWS_AUTO_CREATE_BUCKET = True
AWS_HEADERS = {
"Cache-Control": "public, max-age=86400",
}
AWS_S3_FILE_OVERWRITE = False
AWS_QUERYSTRING_AUTH = False
AWS_S3_SECURE_URLS = True
AWS_REDUCED_REDUNDANCY = False
AWS_IS_GZIPPED = False
# Cache settings.
CACHES = {
# Long cache timeout for staticfiles, since this is used heavily by the optimizing storage.
"staticfiles": {
"BACKEND": "django.core.cache.backends.locmem.LocMemCache",
"TIMEOUT": 60 * 60 * 24 * 365,
"LOCATION": "staticfiles",
},
}
You can set your AWS account details by running the following command:
$ heroku config:set AWS_ACCESS_KEY_ID=your_key_id \ AWS_SECRET_ACCESS_KEY=your_secret_access_key \ AWS_STORAGE_BUCKET_NAME=your_bucket_name
This configuration relies on the django-require-s3 package, which is included in the dependencies for django-herokuapp. In particular, the use of django-require to compress and serve your assets is recommended, since it allows assets to be precompiled during the project's build step, rather than on-the-fly as the site is running.
Heroku does not provide an SMTP server in its default package. Instead, it's recommended that you use the SendGrid Add-on to send your site's emails.
# Email settings.
EMAIL_HOST = "smtp.sendgrid.net"
EMAIL_HOST_USER = os.environ.get("SENDGRID_USERNAME", "")
EMAIL_HOST_PASSWORD = os.environ.get("SENDGRID_PASSWORD", "")
EMAIL_PORT = 25
EMAIL_USE_TLS = False
You can provision a starter package with SendGrid using the following Heroku command:
$ heroku addons:create sendgrid:starter
The smaller the size of your compiled project, the faster it can be redeployed on Heroku servers. To this end,
django-herokuapp provides a suggested .slugignore
file that should be placed in the root of your project. If you've used the herokuapp_startproject.py script
to set up your project, then this will have already been taken care of for you.
Ideally, you should not store your site's SECRET_KEY setting in version control. Instead, it should be read
from the Heroku config as follows:
from django.utils.crypto import get_random_string
SECRET_KEY = os.environ.get("SECRET_KEY", get_random_string(50, "abcdefghijklmnopqrstuvwxyz0123456789!@#$%^&*(-_=+)"))
You can then generate a secret key in your Heroku config with the following command:
$ heroku config:set SECRET_KEY=`openssl rand -base64 32`
It's also recommended that you configure Python to generate a new random seed every time it boots.
$ heroku config:set PYTHONHASHSEED=random
Heroku provides a free piggyback SSL service for all of its apps, as well as a SSL endpoint addon for custom domains. It order to detect when a request is made via SSL in Django (for use in request.is_secure()), you should add the following setting to your app:
SECURE_PROXY_SSL_HEADER = ("HTTP_X_FORWARDED_PROTO", "https")
If you intend to serve your entire app over SSL, then it's a good idea to force all requests to use SSL. The
django-sslify app provides a middleware for this. Simply pip install django-sslify,
then add "sslify.middleware.SSLifyMiddleware" to the start of your MIDDLEWARE_CLASSES.
By default, Django does not log errors to STDERR in production, which is the correct behaviour for most WSGI apps. Heroku, however, provides an excellent logging service that expects to receive error messages on STDERR. You can take advantage of this by updating your logging configuration to the following:
LOGGING = {
"version": 1,
"disable_existing_loggers": False,
"handlers": {
"console": {
"level": "INFO",
"class": "logging.StreamHandler",
},
},
"loggers": {
"django": {
"handlers": ["console"],
}
}
}
Because your site is setup to read some of its configuration from environmental variables stored on
Heroku, running a development server can be tricky. django-herokuapp provides a configuration utility
that should be added to your project to load the heroku config dynamically. Simply add
the following lines to your manage.py script, at the top of the run block:
if __name__ == "__main__": # << This line will already be present in manage.py
# Load the Heroku environment.
from herokuapp.env import load_env
load_env(__file__, "your-app-name")
Django management commands can then be run normally:
$ python manage.py runserver
Accessing the live Heroku Postgres database is a bad idea. Instead, you should provide a local settings file, exclude it from version control, and connect to a local PostgreSQL server. If you're on OSX, then the excellent Postgres.app will make this very easy.
A suggested settings file layout, including the appropriate local settings, can be found in the django-herokuapp template project settings directory.
Once you've completed the above steps, and are confident that your site is suitable to deploy to Heroku,
you can validate against common errors by running the heroku_audit management command.
$ python manage.py heroku_audit
Many of the issues detected by heroku_audit have simple fixes. For a guided walkthrough of solutions, try
running:
$ python manage.py heroku_audit --fix
Things don't always go right first time. Here are some common error messages you may encounter:
The Heroku CLI looks up your app's name from a git remote named heroku. You can either specify the app
to manage by adding -a your-app-name every time you call a Heroku command, or update your git repo with a
Heroku remote using the following command:
$ git remote add heroku git@heroku.com:your-app-name.git
Many django-herokuapp commands need to know the root of the project's file stucture. Django 1.6 provides
this setting automatically as settings.BASE_DIR. If this setting is not present in your settings file,
it should be added as an absolute path. You can look it up dynamically from the settings file like this:
import os.path # Assumes the settings file is located in `your_project.settings` package. BASE_DIR = os.path.abspath(os.path.join(__file__, "..", ".."))
Downloads and bug tracking can be found at the main project website.
The django-herokuapp project was developed by Dave Hall. You can get the code from the django-herokuapp project site.
Dave Hall is a freelance web developer, based in Cambridge, UK. You can usually find him on the Internet in a number of different places: