It seems that we can enter this state only if broker executes `registerControllerAndIncrementControllerEpoch()` and finds that the controller znode has already been created by itself. The question is, is this possible?
Previously if broker tries to create controller znode and node already exists, the broker will simply read the controller id from the controller znode and move on. This patches added quite a few new logic in `controllerNodeExistsHandler()`, e.g. uses zk session id to detect whether the controller znode is created by this broker, handles the scenario that the controller znode is created by this broker. So the new code is more complicated than the previous version. Can you explain a bit why we need these new logic? [ Full content available at: https://github.com/apache/kafka/pull/5101 ] This message was relayed via gitbox.apache.org for [email protected]
