[
https://issues.apache.org/jira/browse/HIVE-3593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13478419#comment-13478419
]
Kevin Wilfong commented on HIVE-3593:
-------------------------------------
Mildly related, if only one partition of the big table is used as input to the
SMB join, there is no need to prefix the file name with the partition spec.
> Output files of SMB join grow indefinitely
> ------------------------------------------
>
> Key: HIVE-3593
> URL: https://issues.apache.org/jira/browse/HIVE-3593
> Project: Hive
> Issue Type: Bug
> Components: Query Processor
> Affects Versions: 0.10.0
> Reporter: Kevin Wilfong
> Assignee: Kevin Wilfong
>
> The output files of a SMB join are prefixed by the big table's partition spec
> that was used to create them. The length of the bucket number portion of the
> file name is updated to be the same length as the length of the task ID.
> Since the task ID is the name of the file, this means that if the output of a
> SMB join is used as the big table of another SMB join, the output files will
> increase by the size of the original partition spec. Compound this and the
> file size can grow indefinitely.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
