753 not seen in Sun-3/280 (not as long, torturous)

[Dan Davison]

I thought I had gone crazy.  I *had* a 753 running in a 3/280, and one day
the controller no longer saw the drive, a Fujitsu M2372K, just as Mr.
Butt's:

Hooking up the cables was a little weird since we had not done it on
Micro-SMD devices such as the Xy753 ctlr but other than that it was not
particularly difficult.  We went with one disk to be safe and thought that
our task was done.  Power on....controller shows  up but no
disks....panic. 

I didn't panic yet.  I had another controller too!

It was at this time we did the smartest thing of all, namely to call in a
hardware type person who came armed with a scope and logic-analyzer.
After much logic analysis of the drive we moved our probe to the first
diode on the 753which is supposed to be live at -12V.  There seemed to be
no juice. OH NO!!!! We ripped off the front panel of the Sun and quickly
went to work probing its power supply.  YEP!! NO -12V DC !!!  It had used
the +5V to drive its "brains" and had actually come up as a legit device
but it needed -12V to control our disks.  We changed power supplies and
our Sun worked like a champ.  Farooq Butt 

Both our controllers didn't work either...  What power supply did you
change?  Can you supply part numbers, etc?

For the record, here's our experience (frighteningly familiar...)

Recently, I tried to install a XY753 (3MB/sec version, from Software
Associates) in two Sun-3/280s.  The first 3/280 was already had a 753, and
the only change was to add another drive.  The second 3/280 was getting
its first 753.  It did not go well.  Both machines saw their board at the
proper address (xdc0 at ... ee80) but the drives attached to the board
were not seen.  Cables and drives were switched, such that 3 different
drives (Fujitsu 2333, 2372, and a Hitachi DK815-10) were sequentially
connected to one controller.  Three sets of cables were also tried,
although probably not in all permutations.  Backplane jumpers were
repeatedly checked.

On one occasion, which was not reproducible, the controller did
see the drive in "format" but any attempt to do anything (read the
label or find the defect list) there was a error as follows:
	Block error 0/0/0 : disk sequencer error
(that's the approximate, not exact, message) and the operation would
abort.

What is especially curious is one of these boards *had* been working; CNLS
folks borrowed it and didn't report any problems.  Also, the other was in
a 3/280 *and working*.  The only change was to attach the second command
cable connector to another drive.  Both drives then disappeared (xdc0 at
... ee80 reported, but no xd0 or xd1 reported) and when the initial
configuration was restored the controller would *not* see the drive it had
been connected to, and working with, for weeks!

The jumper settings on the xy753 are fairly straightforward, and both
controllers were configured as the first xd controller in their systems
(xd0c at ee80).  As mentioned, one of these had been working for some 9
months and the only change was to attempt to add a second disk to the
controller.

We were wearing grounding straps at all times, the machine room itself is
temperature- and humidity-controlled, and the machine room floor is very
well grounded.

Does anyone have any clues as to what would be going on here?  Suggestions
as to things to test?  All machines are running SunOS 4.0.1, and have at
least one xy controller already.  The machine which *ran* with the xy753
had two 451s in it at the same time.


dan davison
theoretical biology and biophysics
t-10 ms k710
los alamos national laboratory
los alamos, nm 87545 USA
dd at lanl.gov
505-665-1355