You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Thomas Graves (Jira)" <ji...@apache.org> on 2021/05/19 12:51:00 UTC
[jira] [Resolved] (SPARK-35093) AQE columnar mismatch on exchange
reuse
[ https://issues.apache.org/jira/browse/SPARK-35093?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Thomas Graves resolved SPARK-35093.
-----------------------------------
Fix Version/s: 3.2.0
3.1.2
3.0.3
Assignee: Andy Grove
Resolution: Fixed
> AQE columnar mismatch on exchange reuse
> ---------------------------------------
>
> Key: SPARK-35093
> URL: https://issues.apache.org/jira/browse/SPARK-35093
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 3.0.2, 3.1.1
> Reporter: Andy Grove
> Assignee: Andy Grove
> Priority: Major
> Fix For: 3.0.3, 3.1.2, 3.2.0
>
>
> With AQE enabled, AdaptiveSparkPlanExec will attempt to reuse exchanges that are semantically equal.
> This is done by comparing the canonicalized plan for two Exchange nodes to see if they are the same.
> Unfortunately this does not take into account the fact that two exchanges with the same canonical plan might be replaced by a plugin in a way that makes them not compatible. For example, a plugin could create one version with supportsColumnar=true and another with supportsColumnar=false. It is not valid to re-use exchanges if there is a supportsColumnar mismatch.
> I have tested a fix for this and will put up a PR once I figure out how to write the tests.
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org