[
https://issues.apache.org/jira/browse/HADOOP-2652?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tsz Wo (Nicholas), SZE updated HADOOP-2652:
-------------------------------------------
Attachment: 2652_20080124b.patch
*This is an incompatible change* since distcp may not be able to copy files
from cluster A (compiled with this patch) to cluster B (compiled with previous
versions). The reason is that distcp starts a Med/Reduce job in cluster B.
Then the job will use Hadoop library provided by cluster B to connect to
cluster A. If the connection scheme is hdfs://, it will work if they use the
same RPC protocal. If the connection scheme is hftp://, the job with old codes
does not provide a ugi. Then, the default web user account (see HADOOP-2614)
will be used in cluster A for further permission checking.
For other cases listed below, distcp is tested to be working for hftp.
- cluster (compiled with this patch) -> cluster (compiled with this patch):
*working* for both hftp and RPC
- cluster (compiled with previous versions) -> cluster (compiled with this
patch): *working* for hftp
- cluster (compiled with previous versions) -> cluster (compiled with previous
versions): *irrelevant*
> Fix permission issues for HftpFileSystem
> ----------------------------------------
>
> Key: HADOOP-2652
> URL: https://issues.apache.org/jira/browse/HADOOP-2652
> Project: Hadoop Core
> Issue Type: Bug
> Components: dfs
> Reporter: Tsz Wo (Nicholas), SZE
> Assignee: Tsz Wo (Nicholas), SZE
> Priority: Blocker
> Fix For: 0.16.0
>
> Attachments: 2652_20080117.patch, 2652_20080117b.patch,
> 2652_20080122.patch, 2652_20080124.patch, 2652_20080124b.patch
>
>
> Similar to HADOOP-2614, HftpFileSystem requires UserGroupInformation to
> access NameNode.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
