[ 
https://issues.apache.org/jira/browse/HDFS-4320?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Mostafa Elhemali updated HDFS-4320:
-----------------------------------

    Attachment: HDFS-4320.trunk.2.patch

OK I just realized the error of my ways: the original patch I uploaded was 
incorrect, since it was trying to overload the dfs.namenode.servicerpc-address 
configuration which is just meant for providing a second endpoint for services 
as opposed to clients of NameNode.

I have a second proposed fix for this problem: a new configuration key 
dfs.namenode.address that specifies the namenode's address if present. If it's 
absent, we fallback on today's fs.default.name logic. Would that work?
                
> NameNode tries to get its address by looking at fs.default.name instead of 
> its own config keys
> ----------------------------------------------------------------------------------------------
>
>                 Key: HDFS-4320
>                 URL: https://issues.apache.org/jira/browse/HDFS-4320
>             Project: Hadoop HDFS
>          Issue Type: Bug
>    Affects Versions: 1.1.0
>            Reporter: Mostafa Elhemali
>         Attachments: HDFS-4320.patch, HDFS-4320.trunk.2.patch, 
> HDFS-4320.trunk.patch
>
>
> When NameNode starts up, it tries to find out its address by looking at the 
> fs.default.name configuration key instead of using its own keys (namely 
> dfs.namenode.servicerpc-address). This breaks scenarios where we try to 
> configure a Hadoop cluster that uses a different default file system than 
> DFS, but still try to prop up namenode and datanode services as a secondary 
> file system.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to