Python backend: FastAPI/Gunicorn + SQLAlchemy/Postgres + Celery/Redis + nginx

Application template to quick start your API server.

Fully Dockerized local development environment.

In dev mode ($env=dev``) uses uvicorn` with live-reload (sources mounted to the container). And Celery worker with live reload.

Installation

Activate (and/or create) the Python environment

. ./activate.sh

Local development

Build and run all containers

./build.sh
./up.sh

To stop all containers but postgres use

./stop.sh

You can debug backend and celery tasks code locally, outside containers. Backend and Celery worker will connect to Postgres and Redis in containers. For that you need in your /etc/hosts:

127.0.0.1   postgres
127.0.0.1   redis

Working with DB

See docker/postgres/README.md.

Create migration script

# compare DB models and current DB and create DB upgrade script in alembic/versions
./alembic.sh revision --autogenerate -m "Schema changes."

# apply script to the DB so after that DB meta data will reflect DB models
./alembic.sh upgrade head

Testing

You only need Posgtres to run test container. We use fakeredis to emulate Redis, fastapi test client to emulate fastapi server.

./up.sh postgres

Run tests from container

./run.sh tests  # run all tests
./run.sh tests python -m pytest -v  # run tests `verbosely`

Run tests locally (as unit-tests without server running)

in /etc/hosts we need

127.0.0.1   postgres

in folder /backend execute

./test.sh -k token  # run tests with `token` in test name
./test.sh -m='unittest and not slow'  # run all fast unittests (locally)
./test.sh -m=benchmark  # run all tests that mesures speed using pytest-benchmark
./test.sh --markers  # see all markers that we can use with `-m` key
./test.sh --cov  # run tests with coverage report

Test local server

./up.sh backend
./test.sh --host 127.0.0.1

This command run local server and test it. It will skip unit tests that cannot be run for external server (marked as unittest).

Stress test

Run tests in parallel in loop as some kind of stress-test using nginx as proxy.

In folder backend/ run:

./stress.sh

Swagger / OpenAPI

Swagger UI available at localhost/docs after server start (./up.sh).

nginx proxy

Nginx proxy at 8001 port.

Without nginx gunicorn server will drop a lot of incoming connections. Because there are only <CPU number> + 1 workers in production mode. Or even only one worker in live reload ($env=dev) mode.

Nginx will buffer requests so your server will serve a lot of parallel clients.