[comp.protocols.tcp-ip] bizarre PC/NFS <-> UNIX <-> PVCS problem encountered

sblair@synoptics.COM (Steven C. Blair) (07/06/90)

We use a rather straightforward Source Code Control System here
for the PC based developers. They use PVCS from some company in
the USA, with PC/NFS from Sun onto our Sun servers.

I.E. Their setup looks like this:




--------------------Ethernet-----------------------------------------
			|				|
			|				|
		|===============|			|
		|		|			|
		|		| Sun 4/390		|
		|		| 7053 Controller	|
		|		| 2x 892Mb hitachi's	|
		|		| SUNOS 4.0.3		|
		|		| Heavy NFS Traffic	|
		|		|			|
		|		|			|
                |===============|			|
						|===============|
						|		|
						|		|
						|===============|
					ComPaq 386/33 PC DOS PC/NFS 3.0.1
				    ****>PVCS Source Code Control System<****
				        ****>used on each local node<****


Tonight, I'm sitting here waiting for the exabyte to finish seeking out to
the 10th record on the tape so I can restore their files, which somehow
"mysteriously" are now corrupted.

Things I know are fine:
=======================
1) network has no errors or crc problems on the Sun's for many days now(+21).
2) PC users flagrantly warm reboot their PC's for no good reasons(or few sometimes).
3) dmesg, fsck, etc, show no errors on the Sun's disks. Including a slow reboot to
do a second pass of fsck -p's( first was with fsck, followed by fsck -b 32).
4) Sometimes I've observered developers checking out whole trees, and getting frustrated
in the middle which leads them to #2. This can and has probably caused me to be doing
a level -0- restore(every night onto 8mm for s/w development to save my time).


The Software manager is convinced "beyond reasonable doubt" that the Sun's, PC/NFS or
the network is to blame. I'm not quite so sure to make that assessment/assumption yet.

I thought that I'd inquire onto the net before I'm convinced that anything other might
be at fault here(including human error). Since I use RCS for my code development, and
not the PC based tool, a call to the manufacturer was futile. The vendor said "We have
no problems in our code --**Your Network Is At Fault**". Yeah right, PVCS is obviously
staffed by rocket scientists from goverenment agencies.

Does anyone else have stories with PVCS, or the combination of it with some type of
"networked" disk service(s) that have had these type of problems??

If so, please contact me via the email addr below