[
https://issues.apache.org/jira/browse/HDFS-2112?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13069478#comment-13069478
]
Tsz Wo (Nicholas), SZE commented on HDFS-2112:
----------------------------------------------
- For {{replthread}},
-* Please remove public from replthread. You may add a getReplicationThread()
method in BlockManagerTestUtil for the tests.
-* Set it to {{final}}.
-* Rename it to replicationthread.
-* Change the comment to javadoc.
i.e. declare it as following.
{code}
/** Replication thread. */
public final Daemon replicationthread = new Daemon(new ReplicationMonitor());
{code}
- For the similar reason, let's keep corruptReplicas package private and remove
getCorruptReplicas().
- Similarly, remove public from computeDatanodeWork and add a getter in
BlockManagerTestUtil.
- Let's set replicationRecheckInterval to {{final}}
- Add private to ReplicationMonitor class and @Override to run().
- Remove namesystem.getBlockManager() since we are already inside BlockManager.
{code}
//ReplicationMonitor.run().
+ namesystem.getBlockManager().processPendingReplications();
{code}
- Remove the following comments since it is no longer valid.
{code}
/////////////////////////////////////////////////////////
//
// These methods are called by the Namenode system, to see
// if there is any work for registered datanodes.
//
/////////////////////////////////////////////////////////
{code}
> Move ReplicationMonitor to block management
> -------------------------------------------
>
> Key: HDFS-2112
> URL: https://issues.apache.org/jira/browse/HDFS-2112
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: name-node
> Reporter: Tsz Wo (Nicholas), SZE
> Assignee: Uma Maheswara Rao G
> Attachments: HDFS-2112.patch
>
>
> Replication should be handled by block manager instead of name system.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
