[
https://issues.apache.org/jira/browse/HIVE-20620?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16624148#comment-16624148
]
Sergey Shelukhin commented on HIVE-20620:
-----------------------------------------
[~ashutoshc] can you take a look? a small change.
Unfortunately try as I might I cannot force a local repro of the original issue
I see on a cluster, so the test doesn't fail without the fix.
On a cluster, the final stage had 16 reducers, but was writing into 5 buckets.
No matter what I do in q files, Tez always generates the correct number of
reducers and taskId in filesinkoperator never changes, each FSO writes its own
files in an orderly manner; in the original repro each reducer wrote files for
multiple different buckets.
[~gopalv] [~djaiswal] do you know by any change how to force Hive/Tez to have
a number of reducers different from the number of buckets for a SMB table, in
tests?
> manifest collisions when inserting into bucketed sorted MM tables with
> dynamic partitioning
> -------------------------------------------------------------------------------------------
>
> Key: HIVE-20620
> URL: https://issues.apache.org/jira/browse/HIVE-20620
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
> Assignee: Sergey Shelukhin
> Priority: Major
> Attachments: HIVE-20620.patch
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
