[ 
https://issues.apache.org/jira/browse/HADOOP-14136?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15893304#comment-15893304
 ] 

Manoj Govindassamy commented on HADOOP-14136:
---------------------------------------------

[~xkrogen],

HADOOP-13055 brings in LinkMergeSlash support where by the root directory from 
any NN can be mounted. We can't mix the root mount and its sub directory mounts 
from the same Cluster though. But, we can always have another cluster sub 
directories mounted in a normal way in a different mount table. So, your use 
case example will turn like something below. Wouldn't this be sufficient ? 
Please share your thoughts.

{noformat}
fs.viewfs.mounttable.ClusterX.linkMergeSlash = hdfs://nn1/
fs.viewfs.mounttable.ClusterY./remote = hdfs://nn2/remote
{noformat}

> Default FS For ViewFS
> ---------------------
>
>                 Key: HADOOP-14136
>                 URL: https://issues.apache.org/jira/browse/HADOOP-14136
>             Project: Hadoop Common
>          Issue Type: New Feature
>          Components: fs, viewfs
>            Reporter: Erik Krogen
>
> It would be useful if ViewFS had the ability to designate one FileSystem as a 
> "primary"/"default" to fall back to in the case that an entry in the mount 
> table is not found. Consider the situation when you have a mount table that 
> looks like:
> {code}
> /data -> hdfs://nn1/data
> /logs -> hdfs://nn1/logs
> /user -> hdfs://nn1/user
> /remote -> hdfs://nn2/remote
> {code}
> {{nn1}} here is being used as the primary, with a specific directory 'remote' 
> being offloaded to another namenode. This works, but if we want to add 
> another top-level directory to {{nn1}}, we have to update all of our 
> client-side mount tables. Merge links (HADOOP-8298) could be used to achieve 
> this but they do not seem to be coming soon; this special case of a default 
> FS is much simpler - try to resolve through the VFS mount table, if not 
> found, then resolve to the defaultFS.
> There is a good discussion of this at 
> https://issues.apache.org/jira/browse/HADOOP-13055?focusedCommentId=15733822&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15733822



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to