[OpenAFS] AFS from NFS Mount

Tino Schwarze tino.schwarze@informatik.tu-chemnitz.de
Fri, 11 Apr 2003 15:10:23 +0200


On Thu, Apr 10, 2003 at 11:01:39AM -0400, Derrick J Brashear wrote:

> > So the only way I could run (theoretically) a fileserver over nfs is by
> > using a namei fileserver?
> >
> > Also, any good info out there on what namei is and why we should use it?
> 
> It's not clear that "you should use it" (to me), but I suspect you mean
> "what advantages does it have"... yes?
> 
> It just means that it's an all userspace daemon instead of needing to
> lookup inodes directly via a kernel interface, and particularly that the file
> names are significant.
> 
> There's some overhead because you're doing a lookup in the filesystem
> instead of just opening the inode you want by number. How much this is has
> been argued, and of course will vary from filesystem to filesystem anyhow.
> But it means the fileserver is portable, because no longer do you need
> kernel code, or need to worry that your filesystem (cough cough reiserfs)
> doesn't bother to have unique inode numbers.

You also don't need to worry about a special fsck because the standard
one will happily delete your AFS data. IMO it is also bad style to hide
files this way on a disk. 

IMO, a non-temporary file should have a name.

Bye, Tino.

-- 
             * LINUX - Where do you want to be tomorrow? *
                  http://www.tu-chemnitz.de/linux/tag/