[
https://issues.apache.org/jira/browse/KAFKA-8832?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16919784#comment-16919784
]
Guozhang Wang commented on KAFKA-8832:
--------------------------------------
[~LordChen] Did you override the `SocketSendBufferBytes` on broker side (by
default it is only 100KB)? With the nio writeTo implementation of FileRecords,
even if the records to write is huge, it would still be done asynchronously
where each time no more that that amount of data would be written to socket
buffer.
> We should limit the maximum size read by a fetch request on the kafka server.
> -----------------------------------------------------------------------------
>
> Key: KAFKA-8832
> URL: https://issues.apache.org/jira/browse/KAFKA-8832
> Project: Kafka
> Issue Type: Bug
> Components: core
> Affects Versions: 2.3.0, 2.2.1
> Reporter: ChenLin
> Priority: Major
> Labels: needs-kip
> Attachments: image-2019-08-25-15-31-56-707.png,
> image-2019-08-25-15-42-24-379.png, image-2019-08-29-11-01-04-147.png,
> image-2019-08-29-11-01-17-347.png, image-2019-08-29-11-02-01-477.png,
> image-2019-08-29-11-03-37-693.png, image-2019-08-29-11-21-49-998.png,
> image-2019-08-29-11-23-53-155.png, image-2019-08-29-11-25-52-242.png
>
>
> I found that kafka is not on the server side, limiting the amount of data
> read per fetch request. This may cause the kafka server program to report an
> error: OutOfMemory. Due to unreasonable client configuration,
> fetch.message.max.bytes configuration is too large, such as 100M, because the
> kafka server receives a lot of fetch requests at a certain moment, causing
> the server to report an error: OutOfMemory。So I think this is a bug。
> !image-2019-08-29-11-25-52-242.png!
> !image-2019-08-25-15-42-24-379.png!
> !image-2019-08-25-15-31-56-707.png!
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
