[ 
https://issues.apache.org/jira/browse/HDFS-7411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14254213#comment-14254213
 ] 

Ming Ma commented on HDFS-7411:
-------------------------------

Couple more comments:

* dfs.namenode.decommission.blocks.per.node together with 
dfs.namenode.decommission.nodes.per.interval can control how long 
DecommissionManager will hold FSN write lock. Should we just define 
dfs.namenode.decommission.blocks.per.interval instead? If 
dfs.namenode.decommission.blocks.per.interval is defined in the configuration, 
use it. If only dfs.namenode.decommission.nodes.per.interval is defined, use 
the run time "average block count per node" to estimate.

* It seems exceededNumBlocksPerCheck is called only when DecommissionManager 
moves to to check the next DN. If a DN has lots of blocks, the check won't stop 
earlier.

* This patch has make "dfsadmin -refreshNodes" asynchronously for decommission; 
e.g. startDecommission no longer calls checkDecommission which takes FSN write 
lock. But it has to wait for DecommissionManager's next check to kick off the 
replication process. https://issues.apache.org/jira/browse/HDFS-5757 and 
https://issues.apache.org/jira/browse/HDFS-7521 provide asynchronous 
notification so replication can start right away.

* We can also make "dfsadmin -refreshNodes" asynchronously for recommission. 
But it doesn't have be done by this jira. We can let other jiras handle this.

> Refactor and improve decommissioning logic into DecommissionManager
> -------------------------------------------------------------------
>
>                 Key: HDFS-7411
>                 URL: https://issues.apache.org/jira/browse/HDFS-7411
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>    Affects Versions: 2.5.1
>            Reporter: Andrew Wang
>            Assignee: Andrew Wang
>         Attachments: hdfs-7411.001.patch, hdfs-7411.002.patch, 
> hdfs-7411.003.patch, hdfs-7411.004.patch, hdfs-7411.005.patch
>
>
> Would be nice to split out decommission logic from DatanodeManager to 
> DecommissionManager.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to