[
https://issues.apache.org/jira/browse/HIVE-7302?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14045662#comment-14045662
]
Vikram Dixit K commented on HIVE-7302:
--------------------------------------
I discussed this with Gopal today and I think that is the right place to
replace it. I tested this on the bucket map join tests I am currently modifying
and getting the right behavior.
Showing this info in the explain extended plan is a good idea. One issue would
be if during execution we actually change the number of reducers in tez, the
number shown may not have any real meaning apart from saying there was an
estimation.
> Allow Auto-reducer parallelism to be turned off by a logical optimizer
> ----------------------------------------------------------------------
>
> Key: HIVE-7302
> URL: https://issues.apache.org/jira/browse/HIVE-7302
> Project: Hive
> Issue Type: Bug
> Components: Tez
> Affects Versions: 0.14.0
> Reporter: Gopal V
> Assignee: Gopal V
> Attachments: HIVE-7302.1.patch
>
>
> Auto reducer parallelism cannot be used for cases where a custom routing
> VertexManager is used.
> Allow a tri-state for ReduceSinkDesc::isAutoParallel to allow "allow,
> disable, unset" mechanics.
> The state machine for this setting will now be
> Allowed transitions
> unset -> allow
> unset -> disable
> allow -> disable
> with no transition case for
> disable -> allow.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
