Docker RabbitMQ Example

TL;DR: Install Docker, then run docker-compose up. Control-C will cleanly stop everything.

This is a minimal example that runs RabbitMQ and two small Python messages, all in a single Docker environment. The publisher program publishes a message into RabbitMQ every 5 seconds; the consumer program prints every message it recevies to its stdout.

Once you've brought it all up, you should see the publisher and consumer printing messages on their stdout. You can also point a browser at http://localhost:15672/ (guest/guest) to see the RabbitMQ management console. (If you are using a Docker Toolbox or local Docker Machine environment, try http://192.168.99.100:15672/.)

Another Publisher

You can run parts of this environment separately if you'd like. With the whole environment running as above, try running the following:

virtualenv vpy
. vpy/bin/activate
pip install pika

cd publisher
export AMQP_URL=amqp://localhost
./publisher.py

The new publisher will start sending its own messages from "Message 0"; both sets of messages will appear on the consumer's output.

In much the same way (continuing to use the same Python virtual environment) you can run the consumer by hand. RabbitMQ supports multiple consumers reading from the same queue and only one of the consumers will print each message.

Development

You can run single parts of this; for instance, to run just RabbitMQ and run it in the background, you can

docker-compose up -d rabbitmq

Now you can run both the publisher and consumer on the console if you'd like.

As a general development approach, my recommendation would be:

• Develop on the host using your favorite text editor and a normally-installed Python, with libraries in a virtual environment

• Write as much non-network code as you can, and use a unit test framework like pytest to write tests for that

• Write simple mock message handlers that batch everything into an in-memory queue, and extend your pytest tests to test your message handlers.

• Use an environment variable like AMQP_URL to indicate the location of the broker (if any) and write bigger integration tests that pytest.skip() themselves if that variable isn't set.

• Only when all of this passes, docker build minimal images with your application and nothing else installed. Hand-test the combined application.