[
https://issues.apache.org/jira/browse/MAPREDUCE-5562?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13785807#comment-13785807
]
Zhijie Shen commented on MAPREDUCE-5562:
----------------------------------------
I've done the local test by imitating two scenarios of unregister() throwing
exceptions:
1. FinsihApplicationMaster is normally done on RM, but somehow the connection
between AM and RM breaks. In this case, everything is fine: RM concludes the
app without starting the 2nd AM (no need, RM has already marked the application
is finished), client sees the final status from JHS. Since AM crashes at
unregister(), the remaining finishing work (e.g., clean staging dir will not
happen)
2. Internal errors happens on RM when processing FinsihApplicationMaster
request, or even before sending the request, AM has thrown exception (e.g.,
cannot connect RM). RM will detect the crash of AM via heartbeat, and the 2n AM
runs, and finishes. The client get the status of the 2nd AM.
> MR AM should exit when unregister() throws exception
> ----------------------------------------------------
>
> Key: MAPREDUCE-5562
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5562
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Zhijie Shen
> Assignee: Zhijie Shen
> Attachments: MAPREDUCE-5562.1.patch, MAPREDUCE-5562.2.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.1#6144)
