[
https://issues.apache.org/jira/browse/TEZ-1748?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14207264#comment-14207264
]
Siddharth Seth commented on TEZ-1748:
-------------------------------------
[~zjffdu] - is there a reason this is required. The current code in
TaskAttemptImpl is structured to have a different Transition class per entity,
to allow any changes that are required for specific transitions. When this is
the same as others - it's just a simple override.
IMO, the current approach is a little cleaner than using a finished() method -
which ends up making assumptions on target state and choices based on that.
> Refactoring - Conslidate the Transition in TaskAttemptImpl
> ----------------------------------------------------------
>
> Key: TEZ-1748
> URL: https://issues.apache.org/jira/browse/TEZ-1748
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.5.1
> Reporter: Jeff Zhang
> Assignee: Jeff Zhang
> Attachments: TEZ-1748.patch
>
>
> 2 points:
> * Some common work needs to be done when TaskAttempt is completed
> (succeeded/killed/failed), like send event to Task, send event to
> AMScheduler( it depends on current state ), unregister taskHeartbeatHandler
> (it depends on current state), we could put them in one places like
> TaskAttemptImpl.finished(TaskAttemptState)
> * There're many kinds of TerminationTransition, most of them are very
> similar, can consolidate them into one or two Transition
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
