[
https://issues.apache.org/jira/browse/HDFS-1080?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jakob Homan updated HDFS-1080:
------------------------------
Attachment: HDFS-1080-trunk.patch
Patch for trunk. Exactly the same as the 20 patch, which we've tested manually
extensively here. This code is buried quite deep in the guts of the secondary
namenode; I think a new unit test isn't feasible here, without a serious,
potentially destabilizing refactor. I'd like to go ahead without one.
This issue occurs, and prevents the 2ndNN from functioning correctly, when a
machine is running on a secure cluster and using IP aliasing to run as a
different host than is returned from getLocalHost. The NN will attempt to
connect to the local host (say 192.168.0.1) rather than the alias (say
secure-2nn), and this will fail the Kerberized SSL authentication and prevent
the merged image from being downloaded.
> SecondaryNameNode image transfer should use the defined http address rather
> than local ip address
> -------------------------------------------------------------------------------------------------
>
> Key: HDFS-1080
> URL: https://issues.apache.org/jira/browse/HDFS-1080
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: name-node
> Reporter: Jakob Homan
> Assignee: Jakob Homan
> Attachments: HDFS-1080-trunk.patch, HDFS-1080-Y20.patch
>
>
> Currently when telling the NN where to get the merged image, SNN uses
> getLocalHost.getAddr(), which may not return the correct ip addr.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
