[
https://issues.apache.org/jira/browse/HADOOP-4618?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12646290#action_12646290
]
Steve Loughran commented on HADOOP-4618:
----------------------------------------
Even though this will be tricky for me to merge in with my lifecycle stuff, I
think it is good, hence the : +1
HADOOP-4383 would benefit from it, as it would not need to rummage in the
namesystem to get the relevant web port.
FSNamesystem.getNameNodeInfoPort() is public -if anyone is using it and its
sibling methods their app will break. And as NameNode.fsnamesystem is public,
this is effectively a bit of a public API that is being broken. I know my code
will break, and I'm happy with that, but I don't know who else is using it.
> Move http server from FSNamesystem into NameNode.
> -------------------------------------------------
>
> Key: HADOOP-4618
> URL: https://issues.apache.org/jira/browse/HADOOP-4618
> Project: Hadoop Core
> Issue Type: Improvement
> Components: dfs
> Affects Versions: 0.1.0
> Reporter: Konstantin Shvachko
> Assignee: Konstantin Shvachko
> Fix For: 0.20.0
>
> Attachments: MoveHttpServer.patch, MoveHttpServer.patch
>
>
> NameNode is responsible now for starting its RPC server. The (web UI) http
> server is started inside FSNamesystem class.
> I think it should be the NameNode's responsibility to deal with all issues
> intended for exposing information it holds to the outside world. This should
> include the RCP server as well as the http server.
> And FSNamesystem class should be a logical container of internal namespace
> data-structures and in general should not be accessed directly from the
> outside.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
