You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Lianet Magrans (Jira)" <ji...@apache.org> on 2026/05/11 13:28:00 UTC
[jira] [Resolved] (KAFKA-20535) Investigate potential improvements to async consumer CPU usage under low max.poll.records
[ https://issues.apache.org/jira/browse/KAFKA-20535?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Lianet Magrans resolved KAFKA-20535.
------------------------------------
Fix Version/s: 4.4.0
Resolution: Fixed
> Investigate potential improvements to async consumer CPU usage under low max.poll.records
> ------------------------------------------------------------------------------------------
>
> Key: KAFKA-20535
> URL: https://issues.apache.org/jira/browse/KAFKA-20535
> Project: Kafka
> Issue Type: Task
> Components: clients, consumer
> Reporter: Lianet Magrans
> Assignee: sanghyeok An
> Priority: Major
> Fix For: 4.4.0
>
>
> As series of improvements were already made to reduce CPU usage on the async consumer for low max.poll.records (e.g, 5).
> Changes to improve CPU usage https://issues.apache.org/jira/browse/KAFKA-18376
> That showed good improvements (async CPU usage aligned with the classic consumer), but it introduced some correctness issues that had to be fixed (https://issues.apache.org/jira/browse/KAFKA-20332 ). After these correctness fixes, low max.poll.record scenarios CPU went from 18% to 25% +- (classic remains at 18%) in some validations (scenarios with default configs remained unaffected)
> This task is to look into the low max.poll.records case and CPU usage for the async consumer, to identify what else could be improved there (without sacrificing any correctness guarantees)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)