[
https://issues.apache.org/jira/browse/KNOX-1413?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16567334#comment-16567334
]
Kristopher Kane commented on KNOX-1413:
---------------------------------------
I'm working on a fix for the master branch that inspects the 403 for
FileAlreadyExists.. but https://issues.apache.org/jira/browse/KNOX-1414 will
need to be resolved in order to make use of it.
> Knox masks WebHDFS 403 errors which are not NameNode HA or safe mode related
> ----------------------------------------------------------------------------
>
> Key: KNOX-1413
> URL: https://issues.apache.org/jira/browse/KNOX-1413
> Project: Apache Knox
> Issue Type: Bug
> Components: Server
> Affects Versions: 0.12.0
> Reporter: Kristopher Kane
> Assignee: Kristopher Kane
> Priority: Minor
>
> WebHdfsHaDispatch checks return status for 403 for StandbyException and
> SafeModeException and raises context accurate exceptions.
> FileAlreadyExistsException is also surfaced from WebHDFS as a 403 but is not
> captured explicitly and the error handling falls through to catching
> IOException which then conducts failoverRequest as if it is a
> StandbyException.
> The strange thing is 403 with the FileAlreadyExistsException does return
> sometimes. I think this has to do with where the rounds of NN HA failover
> end up lying. With a default of three, it can fall on a standby namenode and
> 500 is returned. Other times the correct 403 are returned after the default
> of three failover attempts.
>
> This was seen on 0.12.0 but master branch code hasn't change enough to alter
> this behavior.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
