[
https://issues.apache.org/jira/browse/YARN-1210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13798620#comment-13798620
]
Omkar Vinit Joshi commented on YARN-1210:
-----------------------------------------
Summarizing current patch.
* After RMAppAttempts are recovered then all of the attempts are moved into
LAUNCHED state. After YARN-891 we will know the state of the earlier finished
application attempts; so then based on that we can decide where the current app
attempt should transition to. On RECOVER event
** It will move to LAUNCHED state if it is was the last running app attempt
** It will move to FAILED / KILLED /..other terminal application attempt state.
* When NM RESYNCs containers will be killed and then NM will re-register with
RM passing already running containers. On RM side if any of the container turns
out to be earlier AM container then we will fail that app attempt and
immediately start new app attempt. However if we don't get AM's finished
containerId during furture NM register then after some time AMLivelinessMonitor
will expire and will fail the running app attempt and start a new one.
> 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
>
>
> 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)
