You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Godfrey He (Jira)" <ji...@apache.org> on 2023/01/29 07:55:00 UTC

[jira] [Commented] (FLINK-30814) The parallelism&maxParallelism of sort after a global partitioning is not forced to be 1

    [ https://issues.apache.org/jira/browse/FLINK-30814?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17681693#comment-17681693 ] 

Godfrey He commented on FLINK-30814:
------------------------------------

[~zhuzh] thanks for reporting this issue.

Currently, the parallelism of global sort is set as 1, but the max parallelism is not set because ExecNodeBase#inputsContainSingleton does not work for global sort.  This works fine for the default scheduler, but may occur wrong result if  adaptive batch scheduler changes the global sort parallelism.

 

I will fix it.

 

> The parallelism&maxParallelism of sort after a global partitioning is not forced to be 1
> ----------------------------------------------------------------------------------------
>
>                 Key: FLINK-30814
>                 URL: https://issues.apache.org/jira/browse/FLINK-30814
>             Project: Flink
>          Issue Type: Bug
>          Components: Table SQL / Planner
>    Affects Versions: 1.16.0
>            Reporter: Zhu Zhu
>            Priority: Major
>             Fix For: 1.17.0
>
>
> The parallelism&maxParallelism of sort after a global partitioning is not forced to be 1. The may lead to the parallelism to be changed by adaptive batch scheduler, which is unexpected.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)