[ 
https://issues.apache.org/jira/browse/MAPREDUCE-6706?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15500791#comment-15500791
 ] 

Junping Du commented on MAPREDUCE-6706:
---------------------------------------

Hi [~drankye] and [~asuresh], thanks for proposing the effort. I think we need 
more discussions here before we moving ahead.
In my understanding, TaskUmbilicalProtocol is used only between AM and tasks 
for updating the progress. There is no version compatible issue during cluster 
upgrade because we already support MR library deploy over distributed cache so 
MR (and TEZ also) job can keep exactly the same version for AM and all tasks 
during cluster upgrade. Can you guys clarify what benefit we can achieve by 
doing this effort?
In the mean time, I am not sure if Tez (or other apps)'s buit-in RPC protocol 
are still using WritableRPCEngine or not. cc [~hitesh] and [~sseth] If so, I am 
concern to completely remove WritableRPCEngine for the same reason proposed by 
[~ste...@apache.org] at HADOOP-12579.

> Update TaskUmbilicalProtocol to use ProtobufRPCEngine
> -----------------------------------------------------
>
>                 Key: MAPREDUCE-6706
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6706
>             Project: Hadoop Map/Reduce
>          Issue Type: Improvement
>            Reporter: Arun Suresh
>            Assignee: Wei Zhou
>
> Currently, {{TaskUmbilicalProtocol}} still uses {{WritableRPCEngine}}, which 
> should be moved to {{ProtocolRPCEngine}} so that the former can be deprecated 
> as tracked by HADOOP-12579



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

---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org

Reply via email to