[
https://issues.apache.org/jira/browse/MAPREDUCE-3067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13112355#comment-13112355
]
Vinod Kumar Vavilapalli commented on MAPREDUCE-3067:
----------------------------------------------------
bq. [..] why the exit status does not need to be checked for map/reduce task
containers.
In MRV2 framework, successful maps and reduces directly talk to the AM via
{{TaskUmbilicalProtocol.done()}}. If the task JVM happens to crash after the
done() call, it doesn't really affect the real outcome of the task - so no
point in checking the exit code to be zero. For failed and crashed tasks, we
can just use the exit-code to let the user know via web-UI/command-line.
Other frameworks, like the DistributedShell you are writing, may not have this
direct communication between container and AM and so can leverage the exit-code
for figuring out the outcome of the containers.
> Container exit status not set properly to launched process's exit code on
> successful completion of process
> ----------------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-3067
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-3067
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: mrv2
> Affects Versions: 0.23.0
> Reporter: Hitesh Shah
> Assignee: Hitesh Shah
> Fix For: 0.23.0
>
>
> When testing the distributed shell sample app master, the container exit
> status was being returned incorrectly.
> 11/09/21 11:32:58 INFO DistributedShell.ApplicationMaster: Got container
> status for containerID= container_1316629955324_0001_01_000002,
> state=COMPLETE, exitStatus=-1000, diagnostics=
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
