[
https://issues.apache.org/jira/browse/MAPREDUCE-6449?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14953851#comment-14953851
]
Anubhav Dhoot commented on MAPREDUCE-6449:
------------------------------------------
Changes look good overall
AMWebServices and MRApps pair seems safe to replace. This is of the variety
where you are replacing an existing try catch pair of YarnRuntimeException with
MRRuntimeException.
TypeConverter#toYarn also looks like a candidate.
CachedHistoryStorage, ClientCache, DefaultSpeculator, HistoryFileManager,
LocalContainerAllocator seems of the type where this is no existing catch block
in the callers and is basically replacing one uncaught RuntimeException with
another. So seems fine.
YarnRunner change seems unnecessary.
Minor comment, you can remove unused imports of YarnRuntimeException as you go
along for eg in MRApps.
> MR Code should not throw and catch YarnRuntimeException to communicate
> internal exceptions
> ------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-6449
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6449
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Anubhav Dhoot
> Assignee: Neelesh Srinivas Salian
> Labels: mapreduce
> Attachments: MAPREDUCE-6449.001.patch, MAPREDUCE-6499-prelim.patch
>
>
> In discussion of MAPREDUCE-6439 we discussed how throwing and catching
> YarnRuntimeException in MR code is incorrect and we should instead use some
> MR specific exception.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
