Other SAS disks in T5140

This is fascinating. Could this be the fix for today's 4096 byte sector
drives that lie and screw up ZFS? Or is the problem higher up than this can
fix?

 

I had the same problem, iirc, with a batch of emc fc drives. They were all
Seagate, so was able to use the enterprise disk tools (seagate web site),
which will do a low level format at a different sector size. Only works
with seagate drives however

Thanks, That really *is* usefull info and a copy of the above is now in the
Solaris/general note book ...

Regards,

Chris

 

That makes sense. Still have a sun 3/60, 3/50 and 3/150 in store. The 3/60
was my first sun machine back in 1992'ish and arrived with no drives in
the disk
enclosure, though the scsi -> esdi card was present. Ended up building 2
x 766Meg
esdi drives, 60meg tape and controllers into a different enclosure and
used that
for a couple of years before getting a SS1+. Last time I fired the
system up,
the psu in the disk box had packed up, but it's on the to do list for
resurrection
at some stage.

Don't think I ever got to the stage of having a >1Gb drive in any of the
early
sun systems here. Too expensive at the time, whereas now, there's a
flood of
s/user kit to choose from ...

Regards,

Chris

 

It might be -- if the disk's on-board controller allows
reformatting to 512 byte sectors (which will cause a loss in total
capacity).

I know that I did this just so I could use the discs with ZFS,
and it worked well for that. (There is something about stumbling across
20 72 GB FC-AL disks at a hamfest for $5.00 each.

Enjoy,
DoN.

 

[ ... ]

Glad that it is where others can benefit from it.

Where can I get a copy?

Enjoy,
DoN.

 

My first Sun was a 2/120. Then a 3/140 (skinny box, stood on
end, with only three VME slots, IIRC, then a 3/60, then 3/160 and 3/180
(The latter a rack-mount beast), and then SPARC, starting with a clone
of the SS1+ -- Opus, IIRC. I still have pretty much all of them (along
with my first unix box -- the COSMOS CMS-16/UNX -- a v7 unix port by
UniSoft, on a 68010 in an Intel MultiBus. I also have my first box, an
Altair 680b (Motorola 6800 CPU, not the more common Intel 8080.)

Giving you the choice of putting two ESDI drives in it, or one
(or zero) disks, and a QIC-60 tape drive, with its own controller/SCSI
adaptor card.

With two disks, it was made of lead -- or perhaps gold in
weight.

Good luck with that.

Well ... I was still using the Sun3 systems when I got a couple
of 4GB SCSI drives, and then stumbled across a 9 GB one new (for $900
IIRC) at a hamfest. I wound up with four of those drives after the
price dropped. For a while I was running the 3/180 with four disks in a
rack-mount tray (1 GB was the smallest, IIRC), and I think the 9GB was
the largest in that tray. I actually bought the tray *new* from a
vendor who supplied similar ones to us at work.

Enjoy,
DoN.

 

That's very reassuring, sounds like more of a collector than I am, though
have a friend who shipped two 40ft containers of old kit from the uk to
NY at
one stage. He's currently doing a similar exercise, but this time from NY to
NZ .

A 2/120 would be like rocking horse whatever and even rarer in the uk, but
still have all the aforementioned sun 3's. Had a 3/160 donated at one stage
and it took 2 of us to lift the box down 2 flights of stairs. I think
the oldest
thing here at present is an old pdp11/05, but neither that or the old
sun kit has
been used in anger for years.

You must have had bigger budgets. The two 766's were cdc wren 6, fwir
and still
spun up to cylinder fine last time they were powered up. I only had 2Gb
drives
during the microvax and vax 8200 days. In fact, all those early machines
were
quite slow in real terms, but fast enough once you got the window system
loaded.
I think the 3/60 had the max memory of 24 x 1Mb sticks, not enough to do
anything
with today's software...

Regards,

Chris

 

Well ... I got each of these to *use* them, until something
better came along. Not really intended as a collection -- I just hate
getting rid of things which I have used.

Well ... I got it from someone who was moving up to a SS 1 or 1+
at the time.

I was seriously using the machines, and the 1 GB SCSI drives
were around $200.00 each (3-1/2" ones, so I had to make mounting
adaptors to use them in the larger Sun3 machines. These were Fiji
drives, IIRC, from a friend who was a dealer in computers at the time,
so the price was somewhat better.

But finding a 9GB drive for under $1000 was just too tempting.
The later three of them were after the price had fallen to about $300.00
each or less.

BTW I discovered something with the Sun3 machines which had
me puzzled for a while. I could boot from tape and install the
OS on the drive with no trouble, but I could not boot from the
drive.

It turned out that while a booted OS could handle either parity
or no parity on the drive, to *boot* it had to have parity off,
at least on the boot drive. And the SPARC systems *required*
parity checking turned on, to add to the fun.

Certainly -- and, of course, I didn't know what speed was
possible then. Comparing the Sun-3 machines to the Sun Blade 2000 with
two 1.2 GHz CPUs makes me wonder how I got anything done back then.

Agreed. IIRC, there were two different system boards for the
3/60 -- one such as you had, which used the 1Mx9 SIMMs, and the other
which used only 256 K SIMMs. I never had one of those to play with. I
also never wound up with any of the Sun3x systems, which needed a
different kernel, because they were using the 68030's built-in memory
management instead of Sun's own design.

Enjoy,
DoN.

 

Well, that does sound like collecting to me ;-)

I had one years ago. I think the problem was that only some RAM
slots would accept 1M SIMMs while the rest needed 256K SIMMs.

Those machines are pretty interesting. The 3/80 isn't any faster than
a 3/60, but has internal disks. The 3/470 is a beast as far as 68k goes.
I have got a 3/470 and it's more than twice as fast as a 3/60. It's a
much better design than a 3/80.

Dennis

 

Just not *intended* as a collection.

That sounds about right. I would have to dig up my old FEH to
verify that, but it does sound right.

Hmm ... I wonder whether that was the use of the 68030's memory
management, or some other factor. Did the 3/470 have more than one CPU?

Enjoy,
DoN.

 

The 3/470 was just a single CPU box.
The main difference was that the 470 had some external CPU cache
and the CPU was clocked at 33MHz. It was a 030 finally done right.
I guess the 3/80 was too much of a competition to the new SPARCstations,
so they made it like a cheaper version of a 3/60. 20MHz 020 and 20MHz
030 doesn't make much of a difference.


Dennis

 

I guess others can benefit from it here short term, but am
working on a website and it could go there at some stage. Just
a copy on the server here at present though.

Ars longa, vita brevis and all that ...

Regards,

Chris

 

Especially since it is using the same compiler, and by default
avoiding the extra CISC instructions which the 68030 has that the 68020
does not. (IIRC, it has an instruction which will do the full address
calculations of access into a two dimensional array of arbitrary sized
objects -- and fetch the data (or at least start for multi-word sized
data objects).

I had a COSMOS CMS-16/UNX (which was a 68000 cpu at 8 MHz on an
Intel Multibus, with a port of v7 unix done by Unisoft.) Once I needed
it (or some other system) to do a lot of password crypt crunching (to
recover the root password on a system which had a guest account without
password, and which had a hidden fuse in the power supply blown -- this
so I could actually use the system in question.

I combined the /usr/dict/words files from that box, from the one
being cracked, and from a AT&T 7300 (Unix-PC -- 68010 10 MHz), and wrote
a program to call the crypt routine with a fixed salt stepping through
the words file. I started it on the 68000, got impatient, and started
it running on the 7300, and it did not take long for it to pass the
v7 box -- a lot quicker than should have happened based on the 8 MHz vs
10 MHz clock difference.

At a later time, I asked both compilers to stop at the assembly
language source of the same code, and discovered that the code in the
7300 was a *lot* tighter. Looking at the v7 system's code, I would have
to say that they were having the compiler treat the 68000 as a PDP-11
and never use any instructions which were not similar to that in the
PDP-11 -- except that they *did* at least use the LINK and ULINK
instructions which created and destroyed the stack frame on the 68000.
So -- anything from that compiler would be very much slower (and a lot
more instructions) than that on the 7300.

Enjoy,
DoN.