You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Deepak Jaiswal (JIRA)" <ji...@apache.org> on 2018/06/11 05:42:00 UTC
[jira] [Resolved] (HIVE-17675) verify SMB join with multiple
inserts
[ https://issues.apache.org/jira/browse/HIVE-17675?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Deepak Jaiswal resolved HIVE-17675.
-----------------------------------
Resolution: Won't Fix
This only applies to ACID tables which sort the data using Row ID instead of sort column(s). For such tables SMB wont happen. For other tables a new insert goes into different files and SMB uses KeyValueMerger to merge data from those files.
> verify SMB join with multiple inserts
> -------------------------------------
>
> Key: HIVE-17675
> URL: https://issues.apache.org/jira/browse/HIVE-17675
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
> Assignee: Deepak Jaiswal
> Priority: Major
>
> Hive has a family of joins that interact with sorted and bucketed tables. Afaik one (all?) of them actually rely on the table being sorted, rather than sorting it.
> If one runs insert on such a table without merge more than once, there'd be 2+ files for every bucket that are individually sorted; but globally, the table would no longer be sorted.
> Would these joins work/disable themselves correctly in this case, or could it produce incorrect results? We might need a q file.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)