This post is filled with examples ranging from a simple Dockerfile to multistage production builds for Python apps. Here’s a quick summary of what this guide covers:
- Using an appropriate base image (debian for dev, alpine for production).
gunicornfor hot reloading during development.
- Optimising for Docker cache layers — placing commands in the right order so that
pip installis executed only when necessary.
- Serving static files (bundles generated via React/Vue/Angular) using
flaskstatic and template folders.
- Using multi-stage
alpinebuild to reduce final image size for production.
- #Bonus — Using gunicorn’s
--reload_extra_filesfor watching on changes to files (html, css and js included) during development.
If you’d like to jump right ahead to the code, check out the GitHub repo.
If you’d like to jump to
git push and deploying on the Hasura platform: hasura/hello-python-flask
- Simple Dockerfile and .dockerignore
- Hot Reloading with gunicorn
- Running a single python script
- Serving Static Files
- Single Stage Production Build
- Multi Stage Production Build
Let’s assume a simple directory structure. The application is called python-app. The top level directory has a
Dockerfile and an
The source code of your python app will be in the
src folder. It also contains the app dependencies in
requirements.txt file. For brevity, let’s assume that server.py defines a flask server running on port 8080.
python-app ├── Dockerfile └── src └── server.py └── requirements.txt
1. Simple Dockerfile Example
For the base image, we have used the latest version
During image build, docker takes all files in the
context directory. To increase the docker build’s performance, exclude files and directories by adding a
.dockerignore file to the context directory.
.dockerignore file should be:
.git __pycache__ *.pyc *.pyo *.pyd .Python env
Build and run this image:
$ cd python-docker $ docker build -t python-docker-dev . $ docker run --rm -it -p 8080:8080 python-docker-dev
The app will be available at
Ctrl+C to quit.
Now let’s say you want this to work every time you change your code. i.e. local development. Then you would mount the source code files into the container for starting and stopping the python server.
$ docker run --rm -it -p 8080:8080 -v $(pwd):/app \ python-docker-dev bash [email protected]:/app# python src/server.py
2. Hot Reloading with Gunicorn
gunicorn is a Python WSGI HTTP Server for UNIX and a pre-fork worker model. You can configure gunicorn to make use of multiple options. You can pass on
--reload to the gunicorn command or place it in the configuration file. If any files change, gunicorn will automatically restart your python server.
We’ll build the image and run gunicorn so that the code is rebuilt whenever there is any change inside the
$ cd python-docker $ docker build -t python-hot-reload-docker . $ docker run --rm -it -p 8080:8080 -v $(pwd):/app \ python-hot-reload-docker bash [email protected]:/app# gunicorn --config ./gunicorn_app/conf/gunicorn_config.py gunicorn_app:app
All edits to python files in the
appdirectory will trigger a rebuild and changes will be available live at
http://localhost:8080. Note that we have mounted the files into the container so that gunicorn can actually work.
What about other files? If you have other types of files (templates, views etc) that you want gunicorn to watch for code changes, you can specify the file types in
reload_extra_files argument. It accepts an array of files.
3. Running a single python script
For simple single file scripts, you can run the python script using the python image with docker run.
docker run -it --rm --name single-python-script -v "$PWD":/app -w /app python:3 python your-daemon-or-script.py
You can also pass some arguments to your python script. In the above example, we have mounted the current working directory, which also allows files in that directory to be passed as arguments.
4. Serving Static Files
The above Dockerfile assumed that you are running an API server with Python. Let’s say you want to serve your React.js/Vue.js/Angular.js app using Python. Flask provides a quick way to render static files. Your html should be present inside the
templates folder and your css, js, images should be present inside the
Check out a sample hello world static app structure here.
In your server.py,
if __name__ == '__main__': app.run(host='0.0.0.0')
Note the host,
0.0.0.0 — this allows your container to be accessible from outside. By default, if you don’t give a host, it binds only to the
5. Single Stage Production Build
Build and run the all-in-one image:
$ cd python-docker $ docker build -t python-docker-prod . $ docker run --rm -it -p 8080:8080 python-docker-prod
The image built will be ~700MB (depending on your source code), due to the underlying Debian layer. Let’s see how we can cut this down.
6. Multi Stage Production Build
With multi stage builds, you use multiple
FROM statements in your Dockerfile but the final build stage will be the one used, which will ideally be a tiny production image with only the exact dependencies required for a production server.
This will be really useful when you are using system dependent modules or ones that requires compiling etc.
numpy are good examples of this type.
With the above, the image built with Alpine comes to around ~90MB, a 8X reduction in size. The
alpine variant is usually a very safe choice to reduce image sizes.
Note: All of the above Dockerfiles were written with
python 3 — the same can be replicated for
python 2 with little to no changes. If you are looking to deploy your
django app, you should be able to deploy a production ready app with minimal tweaking to the Dockerfiles above.
Any suggestions to improve the ideas above? Any other use-cases that you’d like to see? Do let me know in the comments.