SS2 upgrade from SS1, Wren Drives, Fujitsu Drives, HP drives

[Bruce Allen]

Thanks for those of you who responded to my request for information about
the SS2 upgrade that I have received, and for suggestions regarding the
Wren V drive.

Here is what I have learned:

1. Upgrade only requires you to send back chassis and memory.  You get to
   keep extra keyboard/mouse/documentation/spare connectors.

2. The load resistor in the place of one internal disk can be removed if
   you have a frame buffer card.  This resistor is only needed if you are
   running as a "compute server", ie diskless and frame-buffer-less, with
   minimum memory.

3. The second internal fan can be disconnected if you have no internal
   drives.

4. If you don't return the NVRAM chip, Sun will be angry.  (But could
   someone please tell me if this SS1 chip works in the SS2??  This way you
   keep your hostid.)

5. OK, disk drives.  The SS2 runs faster than the SS1.  This makes SCSI
   bus timing more critical.  The HP drives work fine on either SS1 or SS2.
   The Fujitsu 2263S works OK on either SS1 or SS2.  The Fujitsu 2266 NEEDS A
   FIRMWARE UPGRADE FROM FUJITSU; contact them or your vendor.  The Wren V
   (and presumably IV,VI,VII) does not appear to work correctly on the SS2
   although it is perfectly OK on the SS1, SS1+ and IPC, both in asynch and
   synch mode. 

My vendor has spoken with his technical liason at Seagate/CDC and they are
going to try to test some of these drives on an SS2 next week, to
determine what is wrong.  My Wren V, which does not work on the SS2, was
made in 1989 and is serial number 035620 model number 94181-702 lot number
O-02-8921-1.  The drive works perfectly on a SS1 and IPC running 4.0.3 and
4.1 with the kernel patched into synch mode.  On the SS2 one can read the
label using the probe-scsi command, but one gets continual error messages
saying "esp0: SCSI bus MESSAGE IN phase parity error" repeating over and
over.  Jumpering parity on actually makes things worse; it changes the
word "IN" to "OUT" in the above error message, and makes the disk label
unreadable.

My short-term solution was to take the Fujitsu 2263 off my IPC at home and
swap it for the Wren V on my SS2 at work.  After only ten wonderful hours
of backing everything up and getting the #$#%@ing networking set back up,
everything seems to be working fine.

And the SS2 is exactly a factor of two faster at my favorite numerical
simulation (and might be better if I recompile everything) so it was worth
the hassle.

I would be interested in any further news concerning Wren V's on SS2's.

	Bruce Allen
	Department of Physics
	University of Wisconsin - Milwaukee
	ballen at dirac.phys.uwm.edu