nfs server not responding problems

[Daniel A. Duval]

Hi,

I have some remote workstations that have
started to get "nfs server not responding"
errors and timeouts. This is preventing
people from logging into the workstations
(I have NFS mounted home directories) until
the vue process is killed. They can usually
login the next time with no problems.

I can simulate these "hangs" by using the
escape-K key sequence (for file completion
in ksh) as I trudge my way down through the
directory structure of the NFS mount. I
usually have to cntl-c to get out of it.

The clients shows timouts and a badcall or
two while the hanging is going on (via the
nfsstat -rc command)

the server is a new L2000 with 11.00 on it,
but it only seems to happen from one of
my subnets (I currently have clients in 3
other subnets).

I have fooled with setting the timeout option
in the mount command with no effect on the
problem. I have played with the wsize and
rsize options for the mount command and got
no failures at 1024, one or two failures at
2048, a couple more at 4096.

I have since made my personal workstation a
NFS server (C240 with 10.20) and mounted it
to the problem clients and had no failures
even with the rsize and wsize settings at
the defaults (8192 I think).

I am about to go after the NFS patches for
the L2000 and am looking for any insight as
to whether or not I am barking up a wrong tree.

The L2000 has had very little done to the
kernel, I have adjusted max_thread_proc to
256, ncallout to 2064, and nkthread to 2048
per a response center paper.

Packet traces while the login process is
hanging shows little to no traffic between
the machines either.

Any advice would be much appreciated!
Thanks!

[Vincent Stedema]

Hi,

Which NFS version is used between the L2000 and the clients? How high is the retrans count on the clients? Also, have you checked the network configuration for the specific subnet (e.g. switch ports and/or NICs set to half in stead of full duplex)?

Regards,

Vincent

[Daniel A. Duval]

Hi,

The clients are nfs version 2, the server is
version 2 and 3.

2 of the machines are at 100mb full duplex and
the other client is 10mb half. the server was
100mb full and is now in a different port at
10mb half. The problem happens when I am
mounted to the fddi address, and it also
happens when I use the ethernet address also.

The retrans stays at 0 on the clients. While
using the packet capture tool yesterday, we
say little to no traffic between the machines
as it was hanging. We didn't get to look
at the packets themselves (not even sure what
to be looking for), but will try again today.

Thanks.
Danny Duval

[Vincenzo Restuccia]

Try http://docs.hp.com/cgi-bin/fsearch/framedisplay?top=/hpux/onlinedocs/B1031-90045/B1031-90045_top.html&con=/hpux/onlinedocs/B1031-90045/00/00/42-con.html&toc=/hpux/onlinedocs/B1031-90045/00/00/42-toc.html&searchterms=problem&queryid=20010628-040729