[
https://issues.apache.org/jira/browse/HIVE-10458?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14539299#comment-14539299
]
Rui Li commented on HIVE-10458:
-------------------------------
Hi [~xuefuz], I've looked at some of the failures. Most of them is due to how
we handle {{order by + limit}} queries. When order by is followed by a limit,
hive assumes that there's only 1 reducer. So we won't create an extra shuffle
to do the limit (otherwise there'll be an extra shuffle with only 1 reducer for
the limit clause).
To solve this, we can either disable parallel order by when there's a limit, or
we can force an extra shuffle for the limit when running on spark. What's your
opinion?
> Enable parallel order by for spark [Spark Branch]
> -------------------------------------------------
>
> Key: HIVE-10458
> URL: https://issues.apache.org/jira/browse/HIVE-10458
> Project: Hive
> Issue Type: Sub-task
> Components: Spark
> Reporter: Rui Li
> Assignee: Rui Li
> Attachments: HIVE-10458.1-spark.patch
>
>
> We don't have to force reducer# to 1 as spark supports parallel sorting.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
