Forcing TCP over NFS Kickstart Install

[Geoff Mayes <gmayes vmware com>]

Hello Kickstart List!

I would be most grateful for any help to the issue I'm having.

Here's the ultimate goal: to run kickstart installs over nfs using TCP
and not UDP on a RHEL3U5 system.

Here's what I've tried:
- It looks like the RHEL5 docs allow an --opts parameter to specify
client-side NFS mounting options.  This parameter didn't work on this
system / RHEL3U5/ anaconda version.
- Firewalling UDP on the NFS port, thinking that maybe anaconda would
fail over to TCP.
- Remounting /mnt/source as TCP in the %pre script portion, but it
complains that the device is busy (it looks
like the process "/usr/bin/python2.2 /usr/bin/anaconda --m
nfs://mnt/source"; is the culprit).  In the %pre script section is there
a way to free up /mnt/device from this process temporarily,
remount /mnt/source, and then start up the anaconda process again?
- Double-mounting on /mnt/source by mounting /mnt/source again as TCP in
the %pre script (not remounting, just mounting).

The double-mounting actually, kind-of, maybe worked (I saw that all
network traffic on the machine serving the source files was TCP and not
UDP), but the install took 10 times longer than when /mnt/source was
singly mounted. Is this due to a double-mount (some sort of file
contention?)?

What is the best way to accomplish the goal of doing TCP NFS kickstart
installs?

Many thanks, Geoff Mayes