You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Dawid Wysakowicz (Jira)" <ji...@apache.org> on 2021/10/18 12:11:00 UTC
[jira] [Closed] (FLINK-24469) Incorrect calcualtion of the buffer
size in case of channel data skew
[ https://issues.apache.org/jira/browse/FLINK-24469?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Dawid Wysakowicz closed FLINK-24469.
------------------------------------
Fix Version/s: 1.14.1
1.15.0
Resolution: Implemented
Implemented in:
* master
** 6343e371c1d293c3df27c07cd0a4f65076a930b8
* 1.14.1
** 60d2a9cb9f04600e96f46f3362a01d65bcd8934b
> Incorrect calcualtion of the buffer size in case of channel data skew
> ---------------------------------------------------------------------
>
> Key: FLINK-24469
> URL: https://issues.apache.org/jira/browse/FLINK-24469
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Checkpointing
> Affects Versions: 1.14.0
> Reporter: Anton Kalashnikov
> Assignee: Anton Kalashnikov
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.15.0, 1.14.1
>
>
> If only a couple out of 100 input channels are active then there is a huge throughput drop is observed. I suppose it happens because of the line inside of SingleInputGate#getBuffersInUseCount:
> {noformat}
> Math.max(1, channel.getBuffersInUseCount());
> {noformat}
> Right now, I don't think that it makes sense to have min value as 1, perhaps zero buffer in use is also ok.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)