The phrasing of "unordered" and "consistent" is confusing to me.
Effectively, current behavior re-keys the destination topic if three conditions do not hold - the origin topic is using non-null keyed messages, a default partitioner is used by the origin producer, and the destination topic has the exact same partition amount. If those all are fine, then these changes aren't really necessary, right? At the very least, I would suggest renaming the config property to `preserve.partitions` and adding the appropriate documentation to mention the new property and say destination topic must have the same number of partitions if used. [ Full content available at: https://github.com/apache/kafka/pull/3151 ] This message was relayed via gitbox.apache.org for [email protected]
