[
https://issues.apache.org/jira/browse/SPARK-53446?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Wenchen Fan resolved SPARK-53446.
---------------------------------
Fix Version/s: 4.2.0
Resolution: Fixed
Issue resolved by pull request 52646
[https://github.com/apache/spark/pull/52646]
> Improve broadcast block cleanup performance in BlockManager to prevent memory
> pressure and OOM
> ----------------------------------------------------------------------------------------------
>
> Key: SPARK-53446
> URL: https://issues.apache.org/jira/browse/SPARK-53446
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 3.4.3, 3.4.4
> Reporter: JacobZheng
> Priority: Major
> Labels: pull-request-available
> Fix For: 4.2.0
>
>
> When Spark applications continuously generate a large number of broadcast
> tasks, the BlockInfoManager’s blockInfoWrappers keeps accumulating broadcast
> block metadata. As a result, the BlockManager.removeBroadcast method becomes
> increasingly slower over time because it needs to iterate through a growing
> number of entries.
> This performance degradation can lead to severe memory pressure and
> eventually trigger an OOM error in long-running jobs with frequent broadcasts.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
