[ 
https://issues.apache.org/jira/browse/TEZ-1259?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16094063#comment-16094063
 ] 

Siddharth Seth commented on TEZ-1259:
-------------------------------------

[~hitesh] - is this case handled currently in recovery?
>From the RecoveryParser - it seems like the latest Init / Started event will 
>be remembered. However, the old Finished event is not forgotten if the vertex 
>is restarted after completion. However I don't see that used. Does the 
>recovery for a Vertex rely on all the tasks completing?

If the recovery relies on all the tasks completing, I don't think this jira 
will cause problems - as long as recovery is dealing with task failures 
correctly.


> Vertex should move to RUNNING state without generating an extra VERTEX_START 
> event
> ----------------------------------------------------------------------------------
>
>                 Key: TEZ-1259
>                 URL: https://issues.apache.org/jira/browse/TEZ-1259
>             Project: Apache Tez
>          Issue Type: Improvement
>            Reporter: Siddharth Seth
>         Attachments: TEZ-1259.001.patch
>
>
> Currently, we end up generating a VERTEX_START event targeted at the same 
> vertex when a vertex is ready to start. Instead, the Vertex can directly move 
> to started state - which prevents unnecessary event queueing.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to