RT ticket #1234:

We have setgid problem on glusterfs mount. Please refer the problem report submitted by our customer.

=============================
I've got an issue where GlusterFS is not honouring the SETGID bit being
raised on a directory. This is supposed to ensure that any directories
created underneath one that has the SETGID bit set are created with the
same GID as its parent. This is something that needs to work as I do too
much stuff that relies on proper permissions being set.

As an example:

[root@web01-dmts www]# ls
total 2696
drwxr-s--- 6 dmtsbiz apache 4096 Sep 1 19:14 . <--- SETGID bit
set
drwxr-s--- 3 root apache 41 Sep 1 14:37 ..
-rw-r--r-- 1 dmtsbiz apache 33 Sep 1 14:37 .bash_logout
-rw-r--r-- 1 dmtsbiz apache 176 Sep 1 14:37 .bash_profile
-rw-r--r-- 1 dmtsbiz apache 124 Sep 1 14:37 .bashrc
drwxr-s--- 2 dmtsbiz apache 6 Sep 1 14:37 cgi-bin
drwxr-s--- 2 dmtsbiz apache 6 Sep 1 14:37 images
drwxr-xr-x 10 root root 4096 Aug 9 13:59
phpMyAdmin-3.2.1-all-languages <--- Extracted archive wrong GID
-rw-r--r-- 1 root root 2739999 Aug 9 14:27
phpMyAdmin-3.2.1-all-languages.tar.bz2 <--- Downloaded file wrong GID
drwxr-x--T 2 root apache 6 Sep 1 14:37 usage

It should set the same group as the parent and the SETGID bit should be
set on each child directory created under the parent. I'm using XFS on
the backing bricks. I don't use LDAP on those nodes, however the nodes
that mount the GlusterFS export either use LDAP or have their passwd
files set to mirror each other (only the web servers use LDAP...things
like mail servers that only have 2 real UID/GID's I handle differently,
but still ensuring that everything maps properly...but this is outside
of that kind of issue I think as I can manually set permissions and
groups...but things like SETGID functionality is broken)

I'm running Gluster 2.0.6. Brick servers are CentOS 5.3 64 bit based as
are the machines mounting the exported file system.

This is fairly critical for me to get going as this is stopping me from
doing a migration to Gluster from NFS as the cleanup on the sheer number
of websites I'd have to perform is insane...

======================================