[
https://issues.apache.org/jira/browse/KAFKA-13630?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nilesh Kumar reassigned KAFKA-13630:
------------------------------------
Assignee: Nilesh Kumar
> Client producer reduce RecordAccumulator locks in producer thread
> -----------------------------------------------------------------
>
> Key: KAFKA-13630
> URL: https://issues.apache.org/jira/browse/KAFKA-13630
> Project: Kafka
> Issue Type: Improvement
> Components: clients, producer
> Reporter: Jason Koch
> Assignee: Nilesh Kumar
> Priority: Minor
>
> Kafka Java client RecordAccumulator uses a thread to manage and send batches
> as they are ready. The algorithm in both ready() and drain() use slightly
> more locking than strictly necessary, which holds up the producer thread from
> writing to batch.
> By reducing the code executed in the lock, we see a 11% reduction in
> spin-loop lock-related CPU in the producer threads (as they wait for an
> available batch to append).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
