[
https://issues.apache.org/jira/browse/HIVE-6759?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Owen O'Malley updated HIVE-6759:
--------------------------------
Attachment: HIVE-6759.patch
This patch fixes the problem by using the supplied length rather that the stat
from the NameNode.
> Fix reading partial ORC files while they are being written
> ----------------------------------------------------------
>
> Key: HIVE-6759
> URL: https://issues.apache.org/jira/browse/HIVE-6759
> Project: Hive
> Issue Type: Sub-task
> Reporter: Owen O'Malley
> Attachments: HIVE-6759.patch
>
>
> HDFS with the hflush ensures the bytes are visible, but doesn't update the
> file length on the NameNode. Currently the Orc reader will only read up to
> the length on the NameNode. If the user specified a length from a
> flush_length file, the Orc reader should trust it to be right.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
