Date:         Tue, 5 Aug 1997 18:22:03 -0700 (PDT)
Sender:       Vanagon Mailing List <vanagon@vanagon.com>
From:         "Christopher M. Smith" <csmith@sdsc.edu>
Subject:      Vanagon List at SDSC - What Happened


First, I want to thank Gerry Skerbitz at UMN for re-hosting the VANAGON
list while we repair our server at SDSC. And extend a personal apology to
the list. Since the list was less than one week old, I had not yet
established our back-up systems (list) for such a severe server failure.
This will be remedied as soon as our server is back on line.

Here's the scoop on the Vanagon List at SDSC:

Last Saturday, the 26th of July, the NT server ("gerry") that hosts the
VANAGON List at SDSC experienced a hardware failure. This is the
Hewlett-Packard Vectra XU dual-pentium machine that was donated to the
list. I was off-site til' Tuesday morning, July 29th. I spent most of
Tuesday morning on the phone with HP technical services trying to diagnose
the problem. We finally came to the conclusion that the problem was with
the motherboard or some component of the board. I immediately drove gerry
over to the nearest HP authorized repair center and was told the tech
would get to the machine sometime within the next 24-48 hrs. Upon my
arrival back at work, I was VERY surprized to get a call from the repair
tech. He not only had the machine, but had a preliminary diagnosis, the L2
memory cache chip was blown. A replacement chip from HP was an
exhorbitant $324. The tech was going to check with the maker of the
third-party chip (note: this is not the original HP memory cache chip) in
the machine to see what kind of warranty they had, and check on a few
leads for less expensive alternatives. I did the same.

The warranty had expired on the memory cache chip, he's leads and mine
bore no fruit. After trying for a day to get a less expensive chip,
Thursday morning I gave the go ahead to get a new chip
from HP and hurry-it-up with FEDEX next day delivery. The tech put in the
order. As standard business practice, HP sends all parts to their service
centers using UPS Overnight. Unfortunately, the HP parts distribution
center, on the East Coast, 1) did not send our chip FEDEX, and 2) made NO
accomodation for the impending UPS strike. They sent our chip UPS which
should have arrived sometime Friday. Well, you guessed it UPS went on
strike ! Need we say more ...

It was obvious on Monday, August 4, (yesterday) that our chip wasn't going
anywhere, whereever it was. So with Ron's urging, I ordered a new chip
personally from HP Direct Sales for FEDEX delivery to SDSC. It arrived
this morning at 10:31 am. I rushed the chip over to the repair center as
soon as I got in. I was fortunate in that the tech, Bill, working on our
server was there and he invited me in for the supposedly final touches to
the repair of our machine.

He swapped out the old with the new chip, turned the power on, AND nothing
happened. Obsolutely nothing! He removed the new memory cache chip and the
machine power'd up. So it looks like the motherboard (cache chip slot ?)
is the source of our problems. That's not to say that the memory cache
chip is also at fault. Well, a new motherboard costs $225. I ordered it1
Meanwhile, the tech was going to check with HP about their warranty of the
motherboard (given the serial number of the machine). With luck, and I'm
not counting on it, we could get a free motherboard replacement.

Worst case scenario: we will end up purchasing a new motherboard and
memory cache chip. If both the old and new cache chips work with the new
motherboard, then it's my opinion that the third-party chip may have
contributed to the motherboard failure. Based upon our experience looking
for a replacement cache chip for our machine, the cache chip we need is
VERY machine/model-specific (thus the third-party chip may have
contributed to the demise of our machine). Because of this, I would also
like to replace the "old" (third party) functioning memeory cache chip
with a new HP chip. Thereby, minimizing the potential for future cache
chip-induced motherboard failure. I'm not a physical systems expert, so if
someone has a comment (pro or con), I'd like to hear it!

	Est. Cost: $150   2-3 hours tech labor
                   $387   memory cache chip
		   $260   motherboard
                  ------
		   $797

We have approximately $550 in the reserves. Obviously, we need more
funds. One mechanism for this is to host another VW group from our server
and have them contribute to the equipment reserves pool. In the meantime,
I'll carry the excess charges. We need to get the LIST back on-ine.

I think the server we have is a fantastic machine. For the time it was
running it delivered mail exceptionally FAST and without a hitch. In
addition the set-up and configuration of the machine, NT OS, LSMTP, and
LISTSERV software was for the most part seamless and required very little
effort on my part. This is a $6-8K machine (excluding the 24 Mg extra RAM)
and 4 GB HD), and was donated. In condsideration of the fact that it has
been in storage for some time, I   was very surprized that it was NOT a
problem from the beginning. It wasn't! Although I was surprized at the
degree and timing of the hardware failure, it was not entirely unexpected.

Once we make the appropriate repairs, I feel that we'll have a robust
machine that will serve us for many years, with little, if any, future
hardware problems (but don't quote me on it!).

Cheers, Chris
Vanagon List Co-Admin

San Diego Supercomputer Center
P.O. Box 85608
San Diego, California  92186-5608
(619) 534-8370 (office)		
csmith@sdsc.edu