[jira] [Comment Edited] (TEZ-2576) It is not necessary to send NodeFailureEvent to task attempt of completed DAG

["Hitesh Shah (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TEZ-2576?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15435738#comment-15435738 ] 

Hitesh Shah edited comment on TEZ-2576 at 8/24/16 9:25 PM:
-----------------------------------------------------------

\cc [~aplusplus] 

Can this simply be addressed by ignoring this event if the attempt is complete and the vertex has completed? Or something along the lines that the TA should only be marked as failed after completion on fetch failure kind of retrospective errors? 


was (Author: hitesh):
\cc [~aplusplus] 

Can this simply be addressed by ignoring this event if the attempt is complete and the vertex has completed? 

> It is not necessary to send NodeFailureEvent to task attempt of completed DAG
> -----------------------------------------------------------------------------
>
>                 Key: TEZ-2576
>                 URL: https://issues.apache.org/jira/browse/TEZ-2576
>             Project: Apache Tez
>          Issue Type: Bug
>            Reporter: Jeff Zhang
>            Assignee: Jeff Zhang
>            Priority: Critical
>
> When node fails, it would send NodeFailureEvent to all the task attempts on this node. It is not necessary to send this to the task attempts that belong to the completed dags. 
> {code}
>  for (TezTaskAttemptID taId : container.failedAssignments) {
>         container.sendNodeFailureToTA(taId, errorMessage, TaskAttemptTerminationCause.NODE_FAILED);
>       }
>       for (TezTaskAttemptID taId : container.completedAttempts) {
>         container.sendNodeFailureToTA(taId, errorMessage, TaskAttemptTerminationCause.NODE_FAILED);
>       }
> {code}



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