hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Konstantin Shvachko (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-3935) Extract classes from DataNode.java
Date Thu, 14 Aug 2008 00:06:46 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-3935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12622401#action_12622401
] 

Konstantin Shvachko commented on HADOOP-3935:
---------------------------------------------

Johan, thanks for doing this. Patch looks much better now. A few more things.
- Throttler is still public and we should make it package private. 
I just removed public from everywhere in the Throttler class, and moved TestBlockReplacement
into package
org.apache.hadoop.hdfs.server.datanode and everything compiled alright. I think this test
belongs to the datanode package anyway.
- Why I want to rename Throttler to something less generic? Because if you read it together
with the package name it says
"hadoop datanode throttler". Which by intention it is not. It throttles block transfers for
the block scanner and for the balancer
but not the data-node. Another consideration is that we have many other throttling mechanisms
in different places and may have
more in the future so we will have to distinguish them.
My first proposition was not good.  How about BlockTransferThrottler or please feel free to
use your own variant.
- I also noticed that DataBlockScanner can be made package private if we move TestInterDatanodeProtocol
into the same package as above. 
This is optional for this patch because it is not directly related to splitting data-node,
we can do it separately.

> Extract classes from DataNode.java
> ----------------------------------
>
>                 Key: HADOOP-3935
>                 URL: https://issues.apache.org/jira/browse/HADOOP-3935
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: dfs
>            Reporter: Johan Oskarsson
>            Assignee: Johan Oskarsson
>            Priority: Trivial
>         Attachments: HADOOP-3935.patch, HADOOP-3935.patch
>
>
> DataNode.java is becoming hard to navigate with over 3000 lines of code. I suggest moving
some of the classes out into their own files in the same package. This will also make it easier
to see how the classes depend on each other and to see what code belongs where.

-- 
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