[
https://issues.apache.org/jira/browse/KAFKA-20687?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18123503#comment-18123503
]
Lianet Magrans commented on KAFKA-20687:
----------------------------------------
This was already delivered in 4.4 (fully unused chunks released when closing
for appends)
[https://github.com/apache/kafka/blob/c2fbf6f52f78ecc1c11c6511242aade6a2c50b71/clients/src/main/java/org/apache/kafka/clients/producer/internals/ChunkedByteBufferOutputStream.java#L229]
> Consider improving unused chunks reuse
> --------------------------------------
>
> Key: KAFKA-20687
> URL: https://issues.apache.org/jira/browse/KAFKA-20687
> Project: Kafka
> Issue Type: Sub-task
> Reporter: Lianet Magrans
> Assignee: Lianet Magrans
> Priority: Major
>
> Chunks return to the pool when batch completes (ensure no data corruption
> across retries etc), but we can consider returning chunks that hold no data
> sooner.
> Initially we could return them when the batch is drained (not waiting until
> batch close with a broker response) . Should be an improvement for cases
> where compressions over-reserves or when concurrent producers allocate
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
