[
https://issues.apache.org/jira/browse/KAFKA-21311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18129012#comment-18129012
]
Sean Quah commented on KAFKA-21311:
-----------------------------------
This might need updates to the group coordinator's error handling if we change
how `MemoryRecordsBuilder` handles buffer expansion
> Throw when uncompressed records need more buffer space
> ------------------------------------------------------
>
> Key: KAFKA-21311
> URL: https://issues.apache.org/jira/browse/KAFKA-21311
> Project: Kafka
> Issue Type: Sub-task
> Reporter: Evan Zhou
> Assignee: Evan Zhou
> Priority: Major
>
> For uncompressed batches, RecordAccumulator sizes the batch buffer before
> each append, so a write should never run past the buffer's capacity. In the
> case that this does happen, we should throw an exception instead of expanding
> so that we don't hide the underlying issue.
> https://github.com/apache/kafka/pull/23483/changes/BASE..a036ed99be4f2f38efc6d5def2961c72aed2af94#r4224404030
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
