This is a test application illustrating how to utilize asyncio with a Flask application to call multiple HTTP endpoints within a single request. A web application server.py calls a "3rd-party" service api.py either in serial or concurrently and returns timing information.
server.py provides a few endpoints to test different strategies for making multiple HTTP calls. api.py simulates doing work by accepting a sleep parameter and waiting for that number of milliseconds before returning. By default server.py will call api.py with delays of 1, 5, and 2 seconds.
To develop locally set up a virtual environment and install the local dependencies:
> python3 -m venv venv
> source venv/bin/activate
> pip install -r requirements.txtThen run each web app:
> python3 api.py
> python3 server.pyHowever when Flask is self-hosted it only handles one request at a time so everything runs serially anyway. Hence this test hosts both web applications in docker containers to take advantage of uWSGI's multi-threading capability.
To launch the app run:
> docker-compose upAnd test it by curling to server.py
> curl -i http://localhost:9090/serial
> curl -i http://localhost:9090/serial?timeout=1.2
> curl -i http://localhost:9090/gather
> curl -i http://localhost:9090/gather?timeout=1.2
> curl -i http://localhost:9090/async
> curl -i http://localhost:9090/async?timeout=1.2serial will invoke all api requests in order, gather will invoke api requests using asyncio.gather(), and async will invoke api requests using the async/await pattern.
To load test server.py fire up the locustfile as:
>locust --host=http://localhost:9090Go to http://localhost:8089 in a browser and kick off the load test there.