[PlanetCCRMA] NFS client trouble with my ccrma box

Fernando Lopez-Lezcano nando@ccrma.Stanford.EDU
Tue Mar 15 10:12:01 2005


On Mon, 2005-03-14 at 17:33, Chris Synan wrote:
> Hi,
> 
> I have a debian server with NFS share, and two FC3 clients, one
> vanilla FC3 and one ccrma.  I keep both relatively up to date, i
> probably update each box weekly.
> 
> On the vanilla box I have no problems accessing the share. 
> 
> On the ccrma box I am having a lot of problems accessing the share.
> 
> The cat5 running between these machine is probably not over 20 yards,
> and I can always ping the server fine from either box.
> 
> Sometimes I transfer large files from my ccrma box onto the share.  It
> is not typical that these files are >4 GB.  Sometimes I leave the
> share mounted for long periods of time with files open or terminal
> window open. 
> 
> The problem arises when the terminal window quits responding to
> requests to access that share.  If I perform 'umount /mnt/pubshare -f'
> it just says "device or resource busy" or something to that effect,
> and I am unable to unmount the share.  There are no open apps that I
> can see that are using part of the mounted resource, and I am not in
> that directory when I try to unmount.  If I perform 'fuser -v
> /mnt/pubshare' it hangs.  At that point, running 'umount' from a
> different terminal will release the hang, but it will complain of
> Input/output error rather than giving me result of fuser, and it still
> will not let me umount.

Do you see anything pertinent to nfs in the output of dmesg or in
/var/log/messages after this happens?

> Meanwhile from the other FC3 box the NFS share is responding just
> fine.  
> 
> Any ideas?  Is there anything about the ccrma setup that should make
> this situation different from my vanilla FC3?

Yes, of course, the kernel. I presume there is some bug in the newer
kernels that shows up in nfs, this could be in the standard kernel or in
the realtime patches that I add. 

> Is there some troubleshooting I can do, or some log file that will
> show the problem I am having with the nfs share?  if I run nfsstat,
> the output looks normal, even when the box is having the connection
> problem.  
> 
> The only way out for now is to reboot.

Yuck. See if there are any messages in /var/log/messages that could
point to the origin of the problem. 

-- Fernando