Status

Current state: ["Under DiscussionCancelled"]

Discussion thread: here [Change the link from the KIP proposal email archive to your own email thread]

...

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Resolution

After the discussion, we decided not to introduce the reconnect timeout config and just do infinite retry of the creation of the Zookeeper object. This way, the broker can be brought down at the time that the admin determines. There is no need to deprecate this property once ZOOKEEPER-2184 is fixed.

Motivation

When a broker's ZK session expires, the broker will try to reconnect to ZK in a session by instantiating a new Zookeeper object. The reconnection could fail if the ZK host can't be resolved (e.g., DNS issue). Currently, when such a failure occurs, the broker just logs the error. This means that even though the broker is still alive, it will never be re-registered in ZK even after the ZK host can be resolved subsequently. An admin will need to restart the broker manually in order for the broker to be added back to the Kafka cluster.

...