Mailing-List: contact hdfs-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-issues@hadoop.apache.org
Message-ID: <24905253.43601272405699949.JavaMail.jira@thor>
Date: Tue, 27 Apr 2010 18:01:39 -0400 (EDT)
From: "Allen Wittenauer (JIRA)" <jira@apache.org>
To: hdfs-issues@hadoop.apache.org
Subject: [jira] Commented: (HDFS-1113) Allow users with write access to a
 directory to change ownership of its subdirectories/files
In-Reply-To: <4142786.41691272402513650.JavaMail.jira@thor>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HDFS-1113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12861571#action_12861571 ] 

Allen Wittenauer commented on HDFS-1113:
----------------------------------------

I do know that tar and likely other archiving technologies rely upon this behavior in POSIX.  The code for tar p basically tries to set the user/group to match the archive.  If it fails, it just throws away the error and continues unarchiving.

Needless to say, such a change would have large ramifications for something like FUSE.  [and eliminate HFDS as a backend replacement for POSIX-style fs's such as pNFS?]

While I have no hard evidence that this is such a bad change, I'd prefer this to be a toggle.  Preferably a per-dir-like thing similar to how quotas were done.  [Of course, such an option just exacerbates the reporting and operational problems that HDFS has.]  I'm concerned about user home directories where it would be an administrative nightmare.

There is also the security aspects to be concerned about.  If I chown a file to someone else, they'll be able to chmod it and now they have write access to a dir.  What if they were never supposed to have access to this cluster to begin with?

While I can appreciate why this would be useful in certain contexts, it definitely sets my spidey sense off.

> Allow users with write access to a directory to change ownership of its subdirectories/files
> --------------------------------------------------------------------------------------------
>
>                 Key: HDFS-1113
>                 URL: https://issues.apache.org/jira/browse/HDFS-1113
>             Project: Hadoop HDFS
>          Issue Type: New Feature
>          Components: name-node
>         Environment: All
>            Reporter: Milind Bhandarkar
>            Assignee: Sanjay Radia
>
> owner and group of a file/directory, and namespace/diskspace quota for a directory are mutable attributes. If I have writable access to a directory, say /team/MyTeam, and if there are subdirectories underneath, such as /team/MyTeam/TeamMember1, /team/MyTeam/TeamMember2, then I should be able to chown, chgrp, setQuota, clrQuota on TeamMemeber{1|2} subdirectories. Currently in HDFS (and in Posix), it requires me to be a superuser to perform these operations.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.