| ... More on 8/16 bit SCSI disk compliance issue.
Using an existing (non-compliant) disk in a local BA356 on AlphaServer
1000A caused the system "init" to hang during the "Testing the disks" phase:
>>> BOOT (or power up)
.
.
Testing the System
Testing the Disks
Loading a non-compliant disk on a HSZ50 BA356 caused the remaining disks in
the BA356 to go offline failing into the FAILEDSET.
Based on the STARS article, the compliant FLEX assembly is part
#17-03503-01, minimum rev. C02. What's involved with upgrading the
FLEX? Is this a customer installable option?
Is there anyway to determine the disk's HW rev. via software. The
existing system is DEC 2300's running OpenVMS v6.2-1H3.
Note that the RZ26L-VA minimum rev is B01. Rev B01 tested OK, but
RZ26L-VA rev C01 failed. Please explain.
STARS Ref.:
ARTICLE: 15762 Logical DB: HARDWARE
Physical DB: FSADMN::$1$DUA70:[TIMA_STARS.HARDWARE] {served}
Identifier: 00990253-B6C78300-1C02A1 VEGA Item Number: 28164
Title: [StorageWorks] Wide (16 bit) vs. Narrow (8-bit)Configuratio
Author: 130503 OLSSON JON 10-MAY-1995
We're in the process of upgrading 100 customer sites from DEC2300
VMSclusters to HSZ50/AS1000A VMSclusters with WIDE BA356's. Please respond
ASAP, a $28M order begins shipping beginning at the end of this month.
|
| Becky,
You've some VERY OLD SBB's there. The last shipping rev of the
RZ26L-VA was E06 and E06 was ECO'd into existence in Feb 1995! The
RZ26L-VA was replaced by the RZ26N-VA sometime in 1995.
I checked the revision history and found some other dates for certain
revs: (The explanation gets pretty long so you may wish to print it!)
A01 changed to B01: on or before 1 Nov 93 (this is where the C02 flex
shows up)
B01 changes to C01: on or before 24 Feb 94 (this fixed a connector
spacing problem)
There have been many code changes since that time.
Let talk C02 flexes... There is no easy was to determine the flex
rev without taking the SBB apart or "ohm-ing out" some pins. The whole
purpose of the C02 flex is to remove grounding on certain pins in the
SBB 96 pin connector. When Storagework(tm) was designed, the SCSI
"bible" required ALL unused pins to be grounded. However, when 16-bit
drives were introduced, the "bible" was changed, and those unused pins
were now required for the upper 8 data lines. When an older SBB is
plugged into a BA356, that SBB grounds the upper 8 data lines. The
wide drives in the shelf "disappear" since they cannot handle having
these lines grounded!
The C02 flex removed the grounding of these pins. Now, there are two
versions of the C02 flex. For whatever reason (I learned about this
years later) the vendor built the initial C02 flexes with 3 lines that
weren't supposed to be there. While none of these were grounded in the
flex, at least one of the lines IS GROUNDED in the drive. So early C02
flexes still have the problem. (The purchase spec was correct, the
vendor just didn't follow the spec!) Again, for reasons unknown to me,
the inventory was NOT purged, so an unknown quantity of "bad" C02's
were shipped. (They work great in narrow only applications, but are
"bad" for wide applications.) The vendor corrected the flex, so later
C02's are correct. Since the spec was correct, no rev change was made.
However, the vendor did mark these differently - there is a small "+"
next to the C02. So we call these "C02+" flexes. With hindsight I say
that we should have written an ECO stating that early C02's did not
meet the spec, and we should have changed the rev to C03...
To write the "what's 16-bit compatible" memo, MCS asked for my input.
I checked the specs, found that C02 was correct (this was before I
learned about C02 vs. C02+), and that's what went out. I learned about
C02's not working as advertised from some flexes sent to me from the
field (Switzerland in this case!) I've never been able to track down
the exact dates that C02 SBB's were shipped vs. when C02+ SBB's started
shipping. I know that old C02 flexes were build into Nov 93. The
earliest C02+ flex I've found was build in April or May 94. I've not
seen any flexes between those dates. That's not to say they are not
out there.
Back to the B01 to C01 change.... The RZ26 was the first drive
installed in an SBB and the ID/LED flex was designed for it. The RZ26L
was the follow-on drive from the same source (in those days Digital!),
the pin out was the same. However, the connector was recessed farther
back on the RZ26L board that the RZ26. Sometimes when the flex was
installed, the connector was separated from the flex. This was fixed
with a short term (RZ26L-VA C01) fix and a long term (C02) fix.
So there should be no difference in functionality between B01 and C01
since the I/O flex is C02 for both and the drive is the same. Unless
of course the C01 drive has a C02 flex and the B01 has the C02+ flex.
So that's the background..... Hope that helps....
Bill
|
| Bill,
Thanks for the reply. Just one more thing...Is there anyway we can
tell, without pulling the drive out, what hardware rev the drive is?
It doesn't appear so to me, but I need confirmation.
Thanks!
Becky
PS. I'll look for the +...
|
| Becky,
Hardware revision - no, no way that I know of to tell without opening
up the SBB. (I can cross reference an SBB rev to a drive rev, but for
older drives this is tough, since the ECO history didn't always mention
the drive rev. Brick A01 SHOULD be in SBB's rev xxx and xxx, but
without opening the SBB, one cannot be certain.)
Unless....
Roger, any hooks in the firmware that gives hardware rev?
Bill
|