[
https://issues.apache.org/jira/browse/HDFS-7702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14297712#comment-14297712
]
Ray commented on HDFS-7702:
---------------------------
Hi [~clamb],
Good questions.
For the failure scenarios, I will document more on it, as you can image , it's
a transaction, if one step failed the whole transaction will roll back.
Administrator has to kick off again, and the whole transaction should not take
long time.
I choose kryo rather than protobuf, because namespace structure is too complex
to convert so many java objects to PB (may be I am wrong here), and I have to
dynamically transfer some namespace data. I use kyronet, because it's seemless
with kryo, it's possible to use IPC instead if have a good reason.:) And I
found deadlock while trying to serialize INode object, Kryo seems easy to
handle customized serialization as well.
I am sorry for other uncleared statement, I will try to make them more clear.
> Move metadata across namenode - Effort to a real distributed namenode
> ---------------------------------------------------------------------
>
> Key: HDFS-7702
> URL: https://issues.apache.org/jira/browse/HDFS-7702
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: Ray
> Assignee: Ray
>
> Implement a tool can show in memory namespace tree structure with
> weight(size) and a API can move metadata across different namenode. The
> purpose is moving data efficiently and faster, without moving blocks on
> datanode.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
