[
https://issues.apache.org/jira/browse/HDFS-15766?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17260536#comment-17260536
]
Ayush Saxena commented on HDFS-15766:
-------------------------------------
Thanx [~LiJinglun], Checked the test again. Firstly :
{code:java}
- routerProtocol.delete(mountPoint, true);
+ nnProtocol.delete(mountPoint, true);
{code}
Changing to {{nnProtocol}} isn't correct either, It returns {{false}] since
{{mountPoint}} which is {{/mntsnapshot}} doesn't exist in the Namenode side.
Hence it isn't of any use, (If you wrap this delete with assertTrue, it would
return false and assertion will fail).
I think the correct thing in the test is to do this instead:
1.
{code:java}
assertTrue(routerProtocol.delete(snapshotFolder, true));
{code}
2.
Remove:
{code:java}
routerProtocol.mkdirs(mountPoint, permission, false);
{code}
Since the mountPoint is already there, we added the mountPoint to resolver,
Correct?
3. In the end remove {{mountPoint}} from the resolver.
4. Even better if we do this cleanup in a {{finally}} block.
nit:
{code:java}
* @return Return zero-length list if the path is a mount point but there are
{code}
empty list sounds better than zero-length list
Other than that things looks ok
> RBF: MockResolver.getMountPoints() breaks the semantic of
> FileSubclusterResolver.
> ---------------------------------------------------------------------------------
>
> Key: HDFS-15766
> URL: https://issues.apache.org/jira/browse/HDFS-15766
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Jinglun
> Assignee: Jinglun
> Priority: Major
> Attachments: HDFS-15766.001.patch, HDFS-15766.002.patch
>
>
> MockResolver.getMountPoints() breaks the semantic of
> FileSubclusterResolver.getMountPoints(). Currently it returns null when the
> path is a mount point and no mount points are under the path.
> {quote}Return zero-length list if the path is a mount point but there are no
> mount points under the path.
> {quote}
>
> This is required by router federation rename. I found this bug when writing
> unit test for the rbf rename. Let's fix it here to avoid mixing up with the
> router federation rename.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
