[linux-lvm] NFS

[linux-lvm] NFS

[ardy]

Are there any issues regarding exporting lvm mounted filesystems
over NFS?

Re: [linux-lvm] NFS

[Andreas Dilger]

Andy writes:
> Are there any issues regarding exporting lvm mounted filesystems
> over NFS?

No, I've been doing it a long time.  LVM is a block device like a disk,
while NFS is a filesystem interface.  As long as the filesystem can
support NFS you are OK.  This limits you pretty much to ext2 or ext3 only,
unless you have reiserfs with up-to-date patches.

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert

Re: [linux-lvm] NFS

[Urs Thuermann]

Andreas Dilger <adilger@turbolinux.com> writes:

> No, I've been doing it a long time.  LVM is a block device like a disk,
> while NFS is a filesystem interface.  As long as the filesystem can
> support NFS you are OK.

??  Has a file system to be written so that it supports NFS?  I
thought all file systems beneath the VFS provide the same interface to
the NFS server and can be exported via NFS.

> This limits you pretty much to ext2 or ext3 only, unless you have
> reiserfs with up-to-date patches.

I have also exported iso9660 and msdos file systems on local block
devices via NFS.


urs

Re: [linux-lvm] NFS

[Eric M. Hopper]

[-- Attachment #1: Type: text/plain, Size: 1500 bytes --]

On Fri, Mar 02, 2001 at 10:56:15AM +0100, Urs Thuermann wrote:
> 
> ??  Has a file system to be written so that it supports NFS?  I
> thought all file systems beneath the VFS provide the same interface to
> the NFS server and can be exported via NFS.

	NFS is such an inane sharing protocol...  It expects directories
to be arranged as if they were just a linear file of names with inode
numbers tacked on the end.  It doesn't treat them any differently than
regular files.  This causes problems with filesystems that are in
essence databases, like reiserfs.

	Reiserfs needs to generate some kind of 'position' cookie that
can be used to refind the spot in the directory that NFS was scanning.
I think the position cookie has to increase as more entries are read,
but I'm not sure.  NFS is stateless, so this position cookie has to work
for some indeterminate amount of time.

> I have also exported iso9660 and msdos file systems on local block
> devices via NFS.

	Yeah, but modeling directories in those filesystems as a string
of bytes is accurate.  You just have to have some means of magically
coming up with an inode number, and that's not too hard.  In FAT, for
example, you can just use the starting cluster number as the inode #.

Have fun (if at all possible),
-- 
The best we can hope for concerning the people at large is that they
be properly armed.  -- Alexander Hamilton
-- Eric Hopper (hopper@omnifarious.mn.org http://www.omnifarious.org/~hopper) --

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [linux-lvm] NFS

[Terje Kvernes]

"Eric M. Hopper" <hopper@omnifarious.mn.org> writes:

> Reiserfs needs to generate some kind of 'position' cookie that can
> be used to refind the spot in the directory that NFS was scanning. I
> think the position cookie has to increase as more entries are read,
> but I'm not sure. NFS is stateless, so this position cookie has to
> work for some indeterminate amount of time.

  now this... I'm not saying anything about the way it's been done...
  but I am exporting three reiserfs-filesystems (all inside a VG by
  the way :) to my local network. the _only_ issue I have had now and
  then are stale NFS-messages. but I've had those on a ext2-system as
  well. I think it's an issue with a loaded and _slow_ IDE disk. or at
  least so it seems.
 
> Have fun (if at all possible),

  oh, we do try. :)

-- 
Terje