It seems that in some cases, the producer in MM2 does not want to shut down gracefully when rolling updates are happening etc. This does not seem to cause many serious issues wince it seems to happen during shutdown, but it might be good to look into it and fix it.

Increasing the task.shutdown.graceful.timeout.ms seems to get rid of the error. But we do not expect it solves the issue, it just makes the error disappear (possibly because the container is force-killed by Kubernetes before it times out).

More logs / metrics will be needed to figure out what is the issue => mmaison@redhat.com will think about what we might ask the user / customer to provide.