Description
Currently if you fail during controlled shutdown, you can get a message that says the socket.timeout.ms has expired. This config actually doesn't exist on the broker. Instead, we should explicitly say if we've hit the controller.socket.timeout.ms or the request.timeout.ms as it's confusing to take action given the current message. I believe the relevant code is here:
https://github.com/apache/kafka/blob/0.10.2/core/src/main/scala/kafka/server/KafkaServer.scala#L428-L454
I'm also not sure if there's another timeout that could be hit here or another reason why IOException might be thrown. In the least we should call out those two configs instead of the non-existent one but if we can direct to the proper one that would be even better.
Attachments
Issue Links
- links to