[
https://issues.apache.org/jira/browse/YARN-1210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13813109#comment-13813109
]
Omkar Vinit Joshi commented on YARN-1210:
-----------------------------------------
completely removed RECOVERED state. rest of the patch is same. Only major
difference is
* Before launching new appAttempt RM will check if any of the application
attempts were running before. If so then RM will wait instead of starting a new
application attempt. If no application attempts are found to be in running
(anything other than final state) state then it launch new application attempt.
* When Node manager receives resync signal it kills all the running containers
and then reports back the killed containers to RM during RM registration. On
receiving the container information RM checks if any of the reported container
is an AM container If so then sends container_failed event to the related app
attempt and eventually starts new application attempt.
> During RM restart, RM should start a new attempt only when previous attempt
> exits for real
> ------------------------------------------------------------------------------------------
>
> Key: YARN-1210
> URL: https://issues.apache.org/jira/browse/YARN-1210
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Vinod Kumar Vavilapalli
> Assignee: Omkar Vinit Joshi
> Attachments: YARN-1210.1.patch, YARN-1210.2.patch
>
>
> When RM recovers, it can wait for existing AMs to contact RM back and then
> kill them forcefully before even starting a new AM. Worst case, RM will start
> a new AppAttempt after waiting for 10 mins ( the expiry interval). This way
> we'll minimize multiple AMs racing with each other. This can help issues with
> downstream components like Pig, Hive and Oozie during RM restart.
> In the mean while, new apps will proceed as usual as existing apps wait for
> recovery.
> This can continue to be useful after work-preserving restart, so that AMs
> which can properly sync back up with RM can continue to run and those that
> don't are guaranteed to be killed before starting a new attempt.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
