[
https://issues.apache.org/jira/browse/HIVE-22045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16897485#comment-16897485
]
Vineet Garg commented on HIVE-22045:
------------------------------------
[~gopalv] [~jcamachorodriguez] I am uploading a new patch which should restore
the behavior of non-blobstorage back to what it was in 3.1.
With this patch new behavior as compared to old behavior is as follow:
||Blob-storage||Old behavior||New behavior||
|SELECT|Avoid intermittent rename
and do final move file by file |*Avoid both intermittent rename*
*as well as final move*|
|CTAS/CMv|Avoid intermittent rename
and do final move file by file|Avoid intermittent rename
and do final move file by file|
|INSERT|Avoid intermittent rename
and do final move file by file|*Do intermittent rename and*
*do final dir rename (*_FASTER)_|
||Non-blobstorage||Old behavior||New behavior||
|SELECT|Do intermittent rename
and also final dir rename|*Avoid both intermittent rename*
*as well as final move*|
|CTAS/CMv|Do intermittent rename
and also final dir rename|Do intermittent rename
and also final dir rename|
|INSERT|Do intermittent rename
and also final dir rename|Do intermittent rename
and also final dir rename|
> HIVE-21711 introduced regression in data load
> ---------------------------------------------
>
> Key: HIVE-22045
> URL: https://issues.apache.org/jira/browse/HIVE-22045
> Project: Hive
> Issue Type: Bug
> Affects Versions: 4.0.0
> Reporter: Vineet Garg
> Assignee: Vineet Garg
> Priority: Major
> Attachments: HIVE-22045.1.patch, HIVE-22045.2.patch,
> HIVE-22045.3.patch, HIVE-22045.4.patch
>
>
> Better fix for HIVE-21711 is to specialize the handling for CTAS/Create MV
> statements to avoid intermittent rename operation but keep INSERT etc
> statements do intermittent rename since otherwise final move by file
> operation is significantly slow for such statements.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
