NFS hangs

NFS hangs

am 03.11.2004 17:41:20 von Greg Kilfoyle

Hi,

I'm experiencing strange NFS hangs which force me to reboot my desktop
system regularly.

Mine is one of several basically identical systems running RedHat
Enterprise Linux (WS) 3.0. The kernel is 2.4.21-20EL. We even tried
replacing the hardware (Dell) but the same problem occurs.

What happens is I'll either be editing in xemacs or running make and the
application or command will hang. Other applications will run fine until
I try to access a file via NFS to a NetApp (Network Appliance) file
server. Any NetApp NFS file access (that is not locally cached) hangs.

I ran an ethernet trace on a small test where I had a one line text file
in my home directory (which is on the NetApp). Start the trace, cat the
small file (hangs), stop the trace. The trace shows that the file is
being requested from the NetApp and in fact delivered fully to the
desktop, but not given back to the application.

I've looked at a Network Appliance document entitled "Using the Linux
NFS Client with Network Appliance Filers" and tried a couple of things
in there, to no avail.

I've tried the following mount options:

hard/soft
udp/tcp
intr

Even with a 'soft' mount there is never an error returned to the
application. Looking at the NFS debug on the Linux desktop I notice that
EBADCOOKIE is being returned by find_dirent() around the time of the
hang, but I've noticed other instances of this which do not cause a
hang.

I'm looking at giving Fedora Core 2 a try which will bring in a 2.6
kernel and may have some extra NFS fixes, although I don't know if all
of our applications will run on FC2.

Mine is the only system (out of about 10) that has this problem. Either
my environment or an application I'm running appears to be inducing
this, but others run the same apps. I could start will a brand new home
directory and rebuild my environment, I guess.

Can't think of what else to do. Yesterday I had to reboot about 6 times.

Any ideas?

Cheers, Greg.
--
Greg Kilfoyle

-
To unsubscribe from this list: send the line "unsubscribe linux-admin" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html