[
https://issues.apache.org/jira/browse/HADOOP-13823?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15822547#comment-15822547
]
Ashutosh Chauhan commented on HADOOP-13823:
-------------------------------------------
[[email protected]] You mentioned in description that you intend to have same
behavior for s3a rename as HDFS. However, both HDFS and azure returns false
when dest file already exists, but in this patch you are instead throwing
exception in that case.
Further, this changed semantics doesn't help Hive since even with this fix we
need to handle HDFS and S3a differently because of return false vs throw
exception. Was that intentional?
> s3a rename: fail if dest file exists
> ------------------------------------
>
> Key: HADOOP-13823
> URL: https://issues.apache.org/jira/browse/HADOOP-13823
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 2.7.3
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Blocker
> Fix For: 2.8.0, 3.0.0-alpha2
>
> Attachments: HADOOP-13813-branch-2-001.patch,
> HADOOP-13823-branch-2-002.patch
>
>
> HIVE-15199 shows that s3a allows rename onto an existing file, which is
> something HDFS, azure and s3n do not permit (though file:// does). This is
> breaking bits of Hive, is an inconsistency with HDFS and a regression
> compared to s3n semantics.
> I propose: rejecting the rename on a file -> file rename if the destination
> exists (easy) and changing the s3a.xml contract file to declare the behavior
> change; this is needed for
> {{AbstractContractRenameTest.testRenameFileOverExistingFile}} to handle the
> changed semantics.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
