This test periodically fails, often from either not consuming, or not receiving a publication confirmation for the final, 10th message.
Is the broker not sending it for some reason? Is it being sent to some OTHER consumer running in another test? It smells of an off-by-one error since it almost always seems to be the final message that get lost. The overhaul of the relay synchronization logic dud not fix it.
Implementing #10 might be useful here to isolate if this is an integration problem or an underlying problem.
peake100
commented
3 years ago
This test periodically fails, often from either not consuming, or not receiving a publication confirmation for the final, 10th message.
Is the broker not sending it for some reason? Is it being sent to some OTHER consumer running in another test? It smells of an off-by-one error since it almost always seems to be the final message that get lost. The overhaul of the relay synchronization logic dud not fix it.
Implementing #10 might be useful here to isolate if this is an integration problem or an underlying problem.