[
https://issues.apache.org/jira/browse/SPARK-16984?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15414571#comment-15414571
]
Apache Spark commented on SPARK-16984:
--------------------------------------
User 'robert3005' has created a pull request for this issue:
https://github.com/apache/spark/pull/14573
> executeTake tries all partitions if first parition is empty
> -----------------------------------------------------------
>
> Key: SPARK-16984
> URL: https://issues.apache.org/jira/browse/SPARK-16984
> Project: Spark
> Issue Type: Bug
> Affects Versions: 2.0.0
> Reporter: Robert Kruszewski
>
> in executeTake if the number of rows returned by first partition is 0 we try
> all partitions next time. This can lead to pathological cases where your
> first partition is empty and rest have data. This unfortunately can happen
> with skewed data. Empirically observed it's better to make few roundtrips
> instead of potentially killing driver with big collect
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
