[ 
https://issues.apache.org/jira/browse/TEZ-2633?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14644917#comment-14644917
 ] 

Bikas Saha commented on TEZ-2633:
---------------------------------

Attaching a patch with the changes. The patch is mainly a refactoring and 
updating some functionality based on that refactoring so that new tests can be 
added to verify the changes. VertexManagers can now receive attempt info on 
their callbacks and can forward scheduling dependencies (if based on attempts) 
when they schedule tasks. This will be consumed later on in TEZ-2646. Backwards 
compatibility has been maintained. [~rajesh.balamohan] please review, specially 
the changes in the different vertex managers.

The other thing that needs to be more widely discussed here is how we expose 
vertex/task/attempt identification. We could directly expose our internal ids 
or we could wrap them and use their functionality without exposing them 
directly. This may be related to some of the needs to expose similar 
information to other plugins created under TEZ-2003. I have gone with the 
wrapper approach for the following reasons
1) It is generally not a good idea to expose internal ids and implicit logic 
encoded in them. This can be a compatibility and maintainability issue going 
forward.
2) Creating an interface for users and wrapping internal ids under them allows 
reuse of the all the functionality in the ids that we want to not duplicate.
3) External interface can expose vertex and dag names, with are useful from the 
user code point of view (because they specify and use those names in their 
code. They may not know the exact integer identifier for the dag or vertex 
before runtime.
[~hitesh] [~sseth] [~rajesh.balamohan] [~pramachandran] Thoughts?

> Allow VertexManagerPlugins to receive and report based on attempts instead of 
> tasks
> -----------------------------------------------------------------------------------
>
>                 Key: TEZ-2633
>                 URL: https://issues.apache.org/jira/browse/TEZ-2633
>             Project: Apache Tez
>          Issue Type: Sub-task
>            Reporter: Bikas Saha
>            Assignee: Bikas Saha
>         Attachments: TEZ-2633.1.patch
>
>
> If the same event is sent from an attempt and its retry then there is no way 
> to differentiate between them.



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

Reply via email to