Demo for reproducing a kafka-python bug

This is a setup that reproduces dpkp/kafka-python#1306.

How to prepare:

1. Make sure you have Docker installed.

2. Create and activate a Python 3 virtualenv.

3. pip install -r requirements.txt. This will install docker-compose>=1.13 (maybe older versions will work, maybe not).

4. docker-compose build. This may take several minutes and may download several hundreds of MB of stuff (but only the first time you run it).

Now let's go to reproduce the bug:

1. Make sure you have activated the virtualenv.

2. docker-compose up -d

3. In another window, docker-compose logs -f kafkaclient. Wait for 5 seconds or until you see a little output that indicates that kafkaclient connected to kafka.

4. docker-compose stop kafkaserver

At this point, after the kafkaserver goes down, one of two things will happen; either the kafkaclient will behave normally, or it will go into an endless loop.

If it behaves normally, the log will show a repeated couple of warning messages about a heartbeat problem, again and again and again. If you start the server again with docker-compose start kafkaserver, after about 10 seconds these messages will stop repeating themselves as the client will automatically reconnect. Go back to step (4) and retry, and eventually you'll get the endless loop.

If it goes into the endless loop, you will get an endless stream of some warning messages and a traceback, which will not stop when you restart the server.

When you're done, docker-compose down will stop the docker containers.