For unit tests, I'm wanting to start and stop instances of my containerized kafka server, repeatedly in the same test suite.
Unfortunately, I'm noticing that there appears to be strange ordering dependencies, my tests pass or fail depending on which order they run in the process. I'm starting and stopping a brand new instance (with a new temp log dir) each time, and starting and stopping a fresh zk instance each time (with it having a fresh temp log/data/dirs each time). Is there some global state dependency? Essentially, what I'm seeing, is if the first test sends and receives a message, then everything subsequently succeeds. However, if the first test starts everything up, but then never sends a message, then shuts down, then subsequent tests which send events never succeed (e.g. messages never get received). Should this sort of thing work ok? Jason
