[
https://issues.apache.org/jira/browse/MAPREDUCE-5533?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13777927#comment-13777927
]
Xuan Gong commented on MAPREDUCE-5533:
--------------------------------------
bq.Should we allow some lag time before marking the history data as
no-progress? I am concerned if this method can cause many false positive
speculation. Can you please make some inputs if this is unlikely to happen?
Originally, between the two continous scheduler TaskAttempUpdate events,
estimatedRunTime will not change. The changes that I have made will keep
estimatedRunTime update everytime that Speculator calls.
I agree that we may need to add some lag time between each changes.
> Speculative execution does not function for reduce
> --------------------------------------------------
>
> Key: MAPREDUCE-5533
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5533
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: applicationmaster
> Reporter: Tassapol Athiapinya
> Assignee: Xuan Gong
> Fix For: 2.1.1-beta
>
> Attachments: MAPREDUCE-5533.1.patch
>
>
> We have sort job where reduce attempt does not send heartbeat in timely
> manner to application master. The AM should kick off another attempt to let
> job succeeds. What we find is the job fails and there is no speculation
> happening.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
