You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Bhagya Lakshmi Gummalla (JIRA)" <ji...@apache.org> on 2018/06/28 14:55:00 UTC
[jira] [Commented] (KAFKA-6980) Recommended MaxDirectMemorySize for
consumers
[ https://issues.apache.org/jira/browse/KAFKA-6980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16526392#comment-16526392 ]
Bhagya Lakshmi Gummalla commented on KAFKA-6980:
------------------------------------------------
[~jlu717] Hi, does this need to be worked on? Is this still an issue?
> Recommended MaxDirectMemorySize for consumers
> ---------------------------------------------
>
> Key: KAFKA-6980
> URL: https://issues.apache.org/jira/browse/KAFKA-6980
> Project: Kafka
> Issue Type: Wish
> Components: consumer, documentation
> Affects Versions: 0.10.2.0
> Environment: CloudFoundry
> Reporter: John Lu
> Priority: Minor
> Labels: consumer, docuentation
>
> We are observing that when MaxDirectMemorySize is set too low, our Kafka consumer threads are failing and encountering the following exception:
> {{java.lang.OutOfMemoryError: Direct buffer memory}}
> Is there a way to estimate how much direct memory is required for optimal performance? In the documentation, it is suggested that the amount of memory required is [Number of Partitions * max.partition.fetch.bytes].
> When we pick a value slightly above that, we no longer encounter the error, but if we double or triple the number, our throughput improves drastically. So we are wondering if there is another setting or parameter to consider?
>
>
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)