[jira] [Commented] (HIVE-11719) acid insert with dynamic partitioning doesn't create empty buckets

["Eugene Koifman (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-11719?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15220683#comment-15220683 ] 

Eugene Koifman commented on HIVE-11719:
---------------------------------------

Above is not quite correct.  I won't break Bucket based joins but disable them.

>From [~ashutoshc]
bq. This may result in loss of performance on MR, because MR join optimization rule for BMJ/SMBJ specifically checks for number of buckets before attempting to transform a join to BMJ/SMBJ. No loss of performance on Tez because Tez has no such requirement.

> acid insert with dynamic partitioning doesn't create empty buckets
> ------------------------------------------------------------------
>
>                 Key: HIVE-11719
>                 URL: https://issues.apache.org/jira/browse/HIVE-11719
>             Project: Hive
>          Issue Type: Bug
>          Components: Transactions
>    Affects Versions: 1.0.0
>            Reporter: Eugene Koifman
>            Assignee: Eugene Koifman
>
> {code:sql}
> CREATE TABLE T(a INT, b STRING)
>       PARTITIONED BY(ds string)
>       CLUSTERED BY(a) INTO 2 BUCKETS
>       STORED AS ORC TBLPROPERTIES ('transactional'='true')
> insert into T partition (ds) values (1, 'fred', 'today'), (2, 'wilma', 'yesterday')
> {code}
> See TestCompactor.dynamicPartitioningUpdate()
> This will currently create 1 bucket file in each partition.  This may break Bucket based joins on MR since they expect to always have a full complement of buckets.
> Should not be an issue on Tez.
> See FileSinkOperator.createBucketForFileIdx()
> Also, double check that compaction properly handles empty buckets,i.e. does delta/base have full complement of bucket files



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)