[jira] [Commented] (HBASE-4420) MasterObserver preMove() and postMove() should throw IOException instead of UnknownRegionException

["Ted Yu (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-4420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13106695#comment-13106695 ] 

Ted Yu commented on HBASE-4420:
-------------------------------

+1 on the proposed approach.

> MasterObserver preMove() and postMove() should throw IOException instead of UnknownRegionException
> --------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-4420
>                 URL: https://issues.apache.org/jira/browse/HBASE-4420
>             Project: HBase
>          Issue Type: Bug
>          Components: coprocessors
>            Reporter: Gary Helmling
>
> We've standardized on IOException as the main way for coprocessors to communicate errors back out of the Observer hooks.  All Observer hooks throw IOE except for MasterObserver.preMove() and MasterObserver.postMove(), which throw UnknownRegionException, since that's what HMasterInterface.move() declares.  In hindsight, making these two MasterObserver methods inconsistent seems like a mistake.
> I think we should change MasterObserver.preMove() and MasterObserver.postMove() to throw IOException for consistency with the other methods.  We could deprecate the existing HMasterInterface.move() method to have it switch over to throwing IOException as well, but this would require creating a version with a new name, which seems unnecessarily ugly.  So I'd suggest we just have HMaster.move() handle the IOException and use it to init an UnknownRegionException.  Wonky as that is, it seems the lesser evil.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira