Completing a txn and starting a new one come with some cost, and hence is what we want to avoid generally.
On the other hand, we do not yet have a mechanism for "keep alive": with that, I think keeping a long lived EMPTY txn is okay, note that if the txn is not empty, then not committing it in time will increase the latency. Hence I'm only trying to optimize the case when the txn is empty. [ Full content available at: https://github.com/apache/kafka/pull/5428 ] This message was relayed via gitbox.apache.org for [email protected]
