NFS Question

[Jerry Feldman]

I stayed out of this discussion because I think it's been answered 
adequately. 
A few years ago, while working for HP at Raytheon, I tested the robustness 
of some of HP's large v-class servers. One of the stress tests I used was 
to run many processes (and many threads). When the file I was writing to 
(and locking) was NFS exported, the system would hang at some very low 
number. Since the NFS server was an NFS appliance, I tested it completely 
on HP-UX and ran into the same problem. When the file was native, the test 
would run to a proper completion. I filed the appropriate problem report to 
HP. The problem more specifically was a client side buffer overflow 
problem. Don't know if it is fixed or not..BTW: I was using fcntl as the 
locking mechanism. A similar (or may have been the same) program (I think 
using threads) would occasionally return 0 from open(2) on an existing 
file. Since 0 is a valid file descriptor, the thread would hang trying to 
read the file. Again, I filed a bug report. This one was given a very high 
priority because it was a real race condition in the kernel. 
-- 
Jerry Feldman <gaf at blu.org>
Associate Director
Boston Linux and Unix user group
http://www.blu.org PGP key id:C5061EA9
PGP Key fingerprint:053C 73EC 3AC1 5C44 3E14 9245 FB00 3ED5 C506 1EA9