[
https://issues.apache.org/jira/browse/TEZ-1773?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14221241#comment-14221241
]
Bikas Saha commented on TEZ-1773:
---------------------------------
bq. Would it be better change TaskAttemptImpl.trySetErrorCause to
trySetTerminationCause to make it consistent with DAG/Vertex
Will do
bq. TaskAttemptTerminationCause.CLUSTER_ACTION is not used, it is for future
usage ?
Will remove.
bq. Would it be better to remove errorCause from constructor, and set it
directly, since the terminationCause bind with the TerminationEvent
Not sure what you mean here. This is designed to be similar to diagnostic
event. Had to use interface because it already extends TaskAttemptEvent. Did
not make it part of TaskAttemptEvent since there are other events that dont
need to have a termination cause. If needed we can change that later.
bq. There's no terminationCause setting for Task, although we have
TaskTerminationCause
termination cause at the attempt level is whats most important, since thats the
real executable entity. Vertex and task are logical entities. Post-hoc analysis
of error trends or error causes would depend on looking at the attempts
termination cause to determine what is happening in the job or cluster. At some
point we can consider removing termination cause at the vertex/task level since
it may not be relevant.
> Add attempt failure cause enum to the attempt failed/killed history record
> --------------------------------------------------------------------------
>
> Key: TEZ-1773
> URL: https://issues.apache.org/jira/browse/TEZ-1773
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Bikas Saha
> Assignee: Bikas Saha
> Attachments: TEZ-1773.1.patch, TEZ-1773.2.patch, TEZ-1773.3.patch
>
>
> Having an enum in it, will provide a programmatic means to present, analyse
> and debug unsuccessful attempts.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
