hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Suresh Srinivas (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HDFS-1359) HDFS federation: Add BlockPoolID to block
Date Thu, 02 Sep 2010 23:37:34 GMT

     [ https://issues.apache.org/jira/browse/HDFS-1359?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Suresh Srinivas updated HDFS-1359:
----------------------------------

    Attachment: HDFS-1359.1.patch

Nicholas, thanks for the comments.

New patch address all but the following:
bq. You may have mentioned it somewhere earlier. What is the reason that ExtendedBlock includes
Block as a field but not extending Block?
# Datanode and Namenode store local part of block without pool ID. The ExtendedBlock is sent
only over protocols. It is optimal to build ExtendedBlock using the Block stored in NN and
DN.
# Though names Block and ExtendedBlock suggest "is a" relationship, in reality the right names
are LocalBlock and Block. Better relationship is ExtendedBlock "has a" local Block in it.

bq. In ExtendedBlock.write(..), it should call Block.write(..). Similar to ExtendedBlock.readFields(..)
This is what I want to do. However, {{ReplicaRecoveryInfo}}, which is a subclass of Block
and implements {{write()}} method. ExtendedBlock built using it and sent in InterDatanodeProtocol.updateReplicaUnderRecovery()
results in RPC issues on the server, since while serializing ReplicaRecoveryInfo is used and
while deserializing Block is used.

The alternative ExtendedBlock building its own Block using the Block supplied does not work
for Mock objects.

I am not happy about current serialization/deserialization of Block in ExtendedBlock, due
to the intimate knowledge it has about what Block is constituted of. However that is the case
in constructors, get methods any way.



> HDFS federation: Add BlockPoolID to block
> -----------------------------------------
>
>                 Key: HDFS-1359
>                 URL: https://issues.apache.org/jira/browse/HDFS-1359
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>            Reporter: Suresh Srinivas
>            Assignee: Suresh Srinivas
>             Fix For: Federation Branch
>
>         Attachments: HDFS-1359.1.patch, HDFS-1359.patch
>
>
> This is a subtask to HDFS-1052 to add BlockPoolID to Block.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message