You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Abdullah Yousufi (JIRA)" <ji...@apache.org> on 2016/08/20 18:40:20 UTC
[jira] [Updated] (HIVE-14272) ConditionalResolverMergeFiles should
keep staging data on HDFS, then copy (no rename) to S3
[ https://issues.apache.org/jira/browse/HIVE-14272?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Abdullah Yousufi updated HIVE-14272:
------------------------------------
Assignee: Sergio Peña (was: Abdullah Yousufi)
> ConditionalResolverMergeFiles should keep staging data on HDFS, then copy (no rename) to S3
> -------------------------------------------------------------------------------------------
>
> Key: HIVE-14272
> URL: https://issues.apache.org/jira/browse/HIVE-14272
> Project: Hive
> Issue Type: Sub-task
> Reporter: Sergio Peña
> Assignee: Sergio Peña
>
> If {{hive.merge.mapfiles}} is True, and the output table to write is on S3, then Hive will generate a conditional plan where smaller files will be merged into larger sizes.
> If the output files written by the initial MR job are small, then a second MR job is run to merge the output into larger files (a copy from S3 to S3 in the current code).
> If the original output files are large enough, then the conditional task is followed by a move/rename which is very expensive in S3.
> We should keep staging data on HDFS previous to copying them to S3 as final files.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)