Coda File System

Re: Coda+NFS

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Tue, 8 Jul 2003 10:49:30 -0400
> > > On Mon, Jul 07, 2003 at 10:45:47AM +0200, Dick Kniep wrote:
> > > 2. If NOT possible, can we mount a Coda partition thru NFS?
> > >
> > > If the second is also not possible, are there any alternatives?
> 
> > It is possible, I used the userspace nfsd before, but never tried the
> > knfsd. NFS doesn't like to export a 'network filesystem', possibly to
> > avoid recursively exporting itself, and needs a special option like
> > --reexport or something.
> 
>     I am sorry, but i don't understand in which combination (exactly)
> i can use (safely) Coda with NFS.

Well, the problem is mainly that the NFS 'layer' makes it impossible to
use cfs and repair and such on the NFS clients. So as long as you stick
to read-only operation it should be fine, but when you write anything
there is a chance that there will be a conflict.

Now the NFS daemon likes to cache filehandles, which gives 2 additional
problems. First of all the Coda client can't really manage it's local
cache when all files are held 'open' and it will not get any useful
access information and might purge an often accessed object from the
cache as soon as it is closed because it only saw one open/close
operation, while another object that wasn't used as much didn't really
get much caching from the NFS daemon and as such has seen many
open/close calls (and seems to be more valuable to the Coda client).

The second problem is that when a conflict is detected, we can't turn
an 'open' object into a dangling symlink. So users will notice access
problems but as the administrator (who will be the only one who can
potentially repair the conflicts) you will not be able to find conflicts
in /coda until the nfs daemon is shut down. And shutting down the NFS
daemon ofcourse affects all users and not just the one who was affected
by the conflict.

Jan
Received on 2003-07-08 10:51:24