[
https://issues.apache.org/jira/browse/HIVE-15421?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15745885#comment-15745885
]
Daniel Dai commented on HIVE-15421:
-----------------------------------
+1
> Assumption in exception handling can be wrong in DagUtils.localizeResource
> --------------------------------------------------------------------------
>
> Key: HIVE-15421
> URL: https://issues.apache.org/jira/browse/HIVE-15421
> Project: Hive
> Issue Type: Bug
> Components: Tez
> Affects Versions: 2.2.0
> Reporter: Wei Zheng
> Assignee: Wei Zheng
> Attachments: HIVE-15421.1.patch, HIVE-15421.2.patch,
> HIVE-15421.3.patch
>
>
> In localizeResource once we got an IOException, we always assume this is due
> to another thread writing the same file. But that is not always the case.
> Even without the interference from other threads, it may still get an
> IOException (RemoteException) due to failure of copyFromLocalFile in a
> specific environment, for example, in a kerberized HDFS encryption zone where
> the TGT is expired.
> We'd better fail early with different message to avoid confusion.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
