[
https://issues.apache.org/jira/browse/SPARK-38401?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
XiDuo You updated SPARK-38401:
------------------------------
Description:
It has several issues with method `ShuffledRowRDD#getPreferredLocations`.
* it does not respect the config `spark.shuffle.reduceLocality.enabled`, so we
can not disable it.
* it does not respect `REDUCER_PREF_LOCS_FRACTION`, so it has no effect if DAG
schedule task to an executor who has less data. In worse, driver will take more
memory to store the useless locations.
was:
It has several issues in the method `getPreferredLocations` of `ShuffledRowRDD`.
* it does not respect the config `spark.shuffle.reduceLocality.enabled`, so we
can not disable it.
* it does not respect `REDUCER_PREF_LOCS_FRACTION`, so it has no effect if DAG
schedule task to an executor who has less data. In worse, driver will take more
memory to store the useless locations.
> Unify get preferred locations for shuffle in AQE
> ------------------------------------------------
>
> Key: SPARK-38401
> URL: https://issues.apache.org/jira/browse/SPARK-38401
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 3.3.0
> Reporter: XiDuo You
> Priority: Major
>
> It has several issues with method `ShuffledRowRDD#getPreferredLocations`.
> * it does not respect the config `spark.shuffle.reduceLocality.enabled`, so
> we can not disable it.
> * it does not respect `REDUCER_PREF_LOCS_FRACTION`, so it has no effect if
> DAG schedule task to an executor who has less data. In worse, driver will
> take more memory to store the useless locations.
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
