NFS problem with AIX server

NFS problem with AIX server

Post by Derek Fountai » Wed, 23 May 2001 21:21:14

I'm trying to access an NFS directory on an AIX server using 2.4.3. I've
tried NFSv2 and NFSv3 but they both have the same problem: when doing a
series of file copies the copying process will hang up. Ctrl-C doesn't
budge it. The machine is otherwise fine, but any attempt to access the NFS
directory from another shell (even a cd) will also block. I have to reboot
to free it up. I've tried a stock 2.4.0 kernel and that shows the same

But, I also tried the 2.4.0 kernel which comes with SuSE-7.1 and that has
no problem. This suggests a patch exists, which has been applied by SuSE,
which fixes the problem. I asked Trond Myklebust (NFS maintainer) but he
figured it was more likely the network driver. I'm using the eepro100
driver which hasn't been changed in nearly a year, and the network is
otherwise perfectly alright.

I'm stumped. Anyone got any ideas?


1. Wierd AIX/NFS client, Sun/NFS server problem

AIX 3.1.6 and 3.1.5 on 320s, SunOS 4.1.1 on SS2s, thinnet on the IBMs,
both on the suns.

Suddenly (why do things always happen suddenly?), NFS client speed
from IBM -> Sun is *amazingly* slow.   As in 8 minutes to copy a
400K file from the remote filesystem to /tmp.  AIX<->AIX speed
is per usual, and Sun<->Sun speed is usual.

The only thing I've changed recently is to move the gateway for our
subnet to a new IP address.  I changed my rc files to reflect the change.
I don't remember the AIX machines slowing down then, either, as I was
using them to build software on the remote file systems immediately after the

I've applied nfsextfix, and that didn't help.  etherfind(1) on the sun
shows that the 320s are asking and getting nfs requests w/ no exraneous
packets inbetween requests.  It's just that there's a 5-10 second lag
between each nfs request (which are 8K in size, like they've always been.)

There are no routers or bridges between the clients and the server.  The
arp table information on the AIX clients is correct.  The routing tables
as reported by netstat -rn seem to be just fine.  The percentage of bad
packets reported by nfsstat, etc, are all well below tolerances (and
are below the rates for the sun clients!).

Any clues/hints?


vox: (713) 749-2126  '91 CB750, DoD# 0378

2. Strange error scsi msg. on console

3. Awful NFS client perf. w/AIX NFS server

4. Maelstrom for Linux/X ??????????

5. AIX nfs server problems - rpc.mountd dies

6. XPG3 standard: where?

7. NFS mounting problem from AIX 4.1.4 to AIX 3.2.5


9. linux NFS server problem - can't ls in the NFS mounted directory.

10. Need info on PC/NFS and Unix NFS servers, server/client ratio

11. problem with LINUX (NFS server) and Solaris 2.7 (intel ) NFS client.

12. Problems with NFS Maestro (Hummingbird) and Solaris/EMC Celerra NFS servers

13. FC/2 NFS Client problem with Linux NFS Server