hive-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eugene Koifman (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-17562) ACID 1.0 + ETL strategy should treat empty compacted files as uncovered deltas
Date Wed, 20 Sep 2017 18:31:03 GMT

    [ https://issues.apache.org/jira/browse/HIVE-17562?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16173638#comment-16173638
] 

Eugene Koifman commented on HIVE-17562:
---------------------------------------

The general idea makes sense but it's not clear to me that the tests actually prove that this
is working.

Also, build bot failed with a compilation error

> ACID 1.0 + ETL strategy should treat empty compacted files as uncovered deltas
> ------------------------------------------------------------------------------
>
>                 Key: HIVE-17562
>                 URL: https://issues.apache.org/jira/browse/HIVE-17562
>             Project: Hive
>          Issue Type: Bug
>          Components: Transactions
>    Affects Versions: 2.4.0
>            Reporter: Prasanth Jayachandran
>            Assignee: Prasanth Jayachandran
>         Attachments: HIVE-17562.1.branch-2.patch
>
>
> In branch-2, with ACID 1.0, following sequence will result in incorrect results
> 1) Set split strategy to ETL
> 2) Insert some rows
> 3) Delete all rows
> 4) Alter table compact MAJOR
> 5) Insert some rows
> 6) Select * query will not return any rows that is written at last (step 5)
> The reason for that, compaction essentially voids the first insert in step 2. Now when
ETL split strategy is chosen, there will not be any stripes in the base files. So no split
gets generated and any subsequent deltas gets ignored.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message