12.2.3.3 Gunicorn Runtime

Gunicorn runtime serves a Python WSGI application in production through multiple worker processes, providing the kind of robust, concurrent request handling a development server (like Flask's built-in one) isn't designed to provide, making it the standard choice for actually running a Python web application's container in production.

Configuring Gunicorn as the Container's Runtime Command

Gunicorn replaces a framework's built-in development server as the actual process running inside the production container.

RUN pip install gunicorn
CMD ["gunicorn", "--bind", "0.0.0.0:8000", "--workers", "4", "app:app"]

This runs the application using four worker processes, providing genuine concurrent request handling appropriate for production traffic.

Why a Development Server Is Inappropriate for Production

A framework's built-in development server is typically explicitly documented as unsuitable for production use, lacking the performance, concurrency, and stability characteristics a production-grade WSGI server like Gunicorn provides.

flask run

gunicorn --bind 0.0.0.0:8000 app:app

The second command represents the appropriate production runtime, in contrast to the first, development-oriented command.

Choosing an Appropriate Number of Worker Processes

The number of worker processes should generally reflect the container's allocated CPU resources, following a common guideline of roughly twice the CPU count plus one.

gunicorn --workers 5 --bind 0.0.0.0:8000 app:app

This count might be appropriate for a container allocated 2 CPU cores, following this common guideline.

Combining Gunicorn With an Async Worker Class for I/O-Bound Applications

For an application with substantial I/O-bound work, an async worker class can provide better concurrency than Gunicorn's default synchronous worker.

gunicorn --worker-class gevent --workers 4 --bind 0.0.0.0:8000 app:app

Why Gunicorn Runtime Matters

Using a production-grade WSGI server like Gunicorn, rather than a framework's built-in development server, is essential for a Python web application's container to actually handle production traffic reliably and with appropriate concurrency.