[
https://issues.apache.org/jira/browse/HADOOP-3083?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12581701#action_12581701
]
Runping Qi commented on HADOOP-3083:
------------------------------------
Actually I think you should repair the leases so that the existing lease
holders can continue writing to their files.
Here is a use case.
say a client is writing to a file /xyz/parent_dir/dub_dir/file
Somebody performs a command like "hadoop dfs -rename /xyz/parent
/xyz/new_parent"
This operation should not cause the client to fail to continue writing the the
file.
Otherwise, "hadoop dfs -rename dir1 dir2" will be a very dangerous operations
> Remove lease when file is renamed
> ---------------------------------
>
> Key: HADOOP-3083
> URL: https://issues.apache.org/jira/browse/HADOOP-3083
> Project: Hadoop Core
> Issue Type: Bug
> Components: dfs
> Reporter: dhruba borthakur
> Assignee: dhruba borthakur
>
> When a file/directory is removed, the namenode should remove lease(s) for
> files that were affected by the rename.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
