Strange happenings w/NFS

[Chris Cappuccio <chris_(_at_)_empnet_(_dot_)_com>]

I have a P-90 (i386) with some large but big and old/noisy drives which I
am using temporarily as a sort of backup server until I get my tape drive
fixed...

The P-90 is an NFS server /home/backup -maproot=backup 192.168.19.4

192.168.19.4 is a P2-400 with Adaptec 2940 UW SCSI (Middle august obsd
snapshot..) It is running nfsiod and has /home/backup nfs mounted from the
P-90...

When I do tar cf - /etc | gzip -9 >/home/backup/19.4-etc.tar.gz on the
P2-400 machine, /home/backup/19.4-etc.tar.gz is written with NFS to the
P-90 'backup server'.  This goes OK because /etc is pretty small...

I have several web pages totalling 2GB with logs in /home/www on the
P2-400 and if I do something like tar cf - /home/www | gzip -9
>/home/backup/19.4-www.tar.gz, the load on the P2-400 goes up to 15 or so
and the machine becomes extremely slow, and has periods of three to ten
seconds where it does not respond to any sort of network traffic at all.
I'm not anywhere near the console so I don't know that that is like but I
imagine that dies too..

The P-90 has a Buslogic PCI controller which OpenBSD is using as ISA so
it's rather slow, so I have 40% buffer cache on it in the kernel config,
so the disk access isn't bad at all..  the P2-400 has 20% cache, no soft
updates, very standard hardware...  BX chipset, adaptec 2940uw, intel
etherexpress.. 

Any ideas as to why the P2-400 dies out like I've described?  The CPU
utilization if I have top running seems to be around %10 during a backup
process, while it isn't locked up... Could this have to do with the
adaptec driver? 

---
"Microsoft is to quality software what McDonalds is to gourmet cooking"