strange diskless linux problem

paulsch@us.ibm.com
Fri, 16 Jul 1999 11:06:24 -0500


Hello all...

I've come across a very strange problem serving diskless clients via NFS from
AIX 4.2.1 with all patches applied and AIX 4.3.2 with no patches...

First, my setup... I took RedHat 5.2 and installed the base system on a RH 5.2
server in a NFS exported directory and fixed it up to work diskless.. I
compiled my client kernel (2.0.36) with the root fs on NFS option and everything
works a-okay...

The problem came when I tried to move this to an AIX server.. I mounted the NFS
exported directory on the AIX machine and did a 'cp -a /my/base/client/tree/*
/aix/nfs/export/dir'... After I set things up for my diskless client to use the
AIX server everything worked fine except it was very, very slow...

The problems seems to be with the devices in the /dev directory... When I did
my first copy of the base client tree I did it with a 2.2.5 kernel on my
server.. I then tried to recopy only the /dev directory from another machine
with a base RH 5.2 install with the stock 2.0.36 kernel and everything worked
perfectly.. So I rebooted my first Linux server with the original 2.0.36 kernel
and recopied the /dev tree and it also worked perfectly.. I tried the copy from
my first RH 5.2 server with the following kernels and the client suffered the
same slowdown: 2.2.5, 2.2.10, and 2.3.10... I also tried copying the /dev
directory from stock TurboLinux 3.6 and RH 6.0 machines which use a 2.2.x kernel
and they caused a slowdown also...

This is very strange because the device files work in either case be it a copy
from a 2.0.36 kernel based server or the 2.2.x, and 2.3.10, but the 2.0.36 copy
doesn't cause performance problems... When doing an ls -l /dev from my diskless
client after either copy everything looks fine.. Correct permissions and
correct major, minor device numbers... There seems to be no difference
whatsoever...

I also set up another RH 5.2 server and used it to NFS serve my root client
directory and did a copy to it from my server with a 2.2.10 kernel and that
worked fine... So this problem only appears to happen with AIX NFS...

So the question is (hoping I described the above clearly)... Does anybody have
any idea as to why this may be happening???

Thanks in advance...Paul...

Paul Schroeder
Software Engineer
NCSD Division, IBM Corp
Office: 512-838-9675
Lab: 512-838-8305
paulsch@us.ibm.com

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/