[Sheflug] NFS and fast machines

["Morris, David (Allvac, UK)" <david [at] xxxxxxxxxxxx>]

I'm just setting up our production LTSP server and have been testing it
with a single client.

Talk about problems. Repeated hangs when the client tries to do a
'pivot_root' prior to running init.

There are no security problems - nfs works, tftp works, the machines are
on their own little LAN. DHCP is fine.

Eventually, as a final measure of desparation, I started a 'tcpdump -I
eth0' on the server.

<PCMODE>Blow me</PCMODE> if the whole lot didn't spring into life and
fire up X on the client.

Stop tcpdump and reboot the client. Hang at pivot_root. Start tcpdump
and reboot and it works.

It looks like having something on the server (AMD XP2800+, 1MB 400MHz
memory dual-channel DDR on a Gigabyte MB, SuSE9.0, soft raid on a pair
of 73Gig drives, embedded gigabit nic into a 100 meg netgear switch
[testing only]) which slows the LAN down a bit has the effect of giving
everything enough time to boot.

I know there are some issues with fragmentation of UDP packets on NFS,
but I'm not sure how they would affect this installation, or even
exactly what they are.

Has anyone else got any clues?


-- 
David Morris, European IT Manager, Allvac Ltd
w: david [at] allvac.co.uk  http://www.allvac.co.uk
h: david [at] brassedoff.net  http://www.davidm.demon.co.uk 
ddi: +44 114 220-1289   m: 07973 530987

___________________________________________________________________

Sheffield Linux User's Group -
http://www.sheflug.co.uk/mailfaq.html

  GNU the choice of a complete generation.