Run production webservers such as pyuwsgi (aka uWSGI) or gunicorn as a Django management command.
-
Install a variant:
pip install 'django-webserver[pyuwsgi]'or
pip install 'django-webserver[gunicorn]'or
pip install 'django-webserver[uvicorn]' # Python 3.5+ onlyor
pip install 'django-webserver[waitress]' -
Add to
INSTALLED_APPS:INSTALLED_APPS = [ # ... "django_webserver", # ... ]
-
Run:
manage.py pyuwsgi --http=:8000 ...or
manage.py gunicornor
manage.py uvicornor
manage.py waitress --port=8000
Pyuwsgi already knows the Python interpreter and virtualenv (if applicable) to use from the Django management command environment. By default, it will run with the following flags (using settings.WSGI_APPLICATION to determine the module):
--strict --need-app --module={derived}
If you have STATIC_URL defined with a local URL, it will also add --static-map, derived from STATIC_URL and STATIC_ROOT.
You can pass any additional arguments uWSGI accepts in from the command line.
But uWSGI has a lot of flags, and many of them, you want every time you run the project. For that scenario, you can configure your own defaults using the optional setting, PYUWSGI_ARGS. Here's an example you might find helpful:
PYUWSGI_ARGS = [
"--master",
"--strict",
"--need-app",
"--module",
":".join(WSGI_APPLICATION.rsplit(".", 1)),
"--no-orphans",
"--vacuum",
"--auto-procname",
"--enable-threads",
"--offload-threads=4",
"--thunder-lock",
"--static-map",
"=".join([STATIC_URL.rstrip("/"), STATIC_ROOT]),
"--static-expires",
"/* 7776000",
]Don't forget to also set something like --socket=:8000 or --http=:8000 so your app listens on a port. Depending on your setup, it may make more sense to pass this in via the command line than hard-coding it in your settings.
Same as the standard gunicorn configuration, but the application will be set for you from settings.WSGI_APPLICATION.
Note: Unlike the other servers, you have to configure gunicorn with environment variables or via sys.argv. If you use it with Django's call_command, keep in mind any additional arguments you pass will not be applied.
Same as the standard uvicorn configuration, but the application will be set for you from settings.WSGI_APPLICATION as well as --wsgi.
Same as the standard waitress-serve arguments, but the application will be set for you from settings.WSGI_APPLICATION.
Unlike the other servers, waitress is supported on Windows.
Default:
WEBSERVER_WARMUP = TrueTypically, when a WSGI server starts, it will bind to the necessary ports then import/setup your application. On larger projects, it's normal for startup to take multiple seconds. During that time, it is unable to respond to incoming requests.
To avoid that downtime, this app imports your WSGI module before starting the relevant server. If, for some reason this behavior is undesirable, you can set WEBSERVER_WARMUP = False in your settings.
This is not enabled by default. It requires WEBSERVER_WARMUP = True.
WEBSERVER_WARMUP_HEALTHCHECK = "/-/health/"Internally calls the provided URL prior to starting the server and exits with a failure if it does not return a 200.
It can be helpful to have your app exit immediately if it is unable to successfully respond to a healthcheck. Your process or container manager should immediately show the service failed instead of waiting for a load balancer or some other monitoring tool to notify catch the problem.
In some scenarios, it is beneficial to distribute a Django project with a single entrypoint for command-line interaction. This can come in handy when building Docker containers or self-contained Python apps with something like shiv.
Pre-warming the application and running a healthcheck can also open the door for some zero-downtime deployment scenarios that previously weren't possible due to the issues described in "Pre-warming your app". For example, you could use the --reuse-port option in uWSGI or gunicorn to bring up a new version of your app on the same port, knowing it is already warmed-up and healthy. After a successful startup, the old version can safely be torn down without dropping any traffic.