Re: SS-10 funky SCSI problems...

Robert Dinse (nanook@eskimo.com)
Wed, 3 Nov 1999 09:04:36 -0800 (PST)


On Thu, 4 Nov 1999, Anton Blanchard wrote:
>
> esp0: IRQ 36 SCSI ID 7 Clk 25MHz CCF=5 TOut 167 NCR53C9x(esp236)
> esp1: IRQ 53 SCSI ID 7 Clk 40MHz CCF=8 TOut 167 NCR53C9XF(espfast)
>
> My 4m/690 has two esp cards and the disks work hard but I don't see this
> problem. Do you have any other non-ross CPUs to try and isolate the problem?

I don't but here is another data point. I took those same CPU's out of
the SS-10 and put them back in a 4/670 system. I took the same SCSI/10-base-T
combo board out of the SS-10 and put it in a 4/670MP. I took the CG-SIX video
board out of the SS-10 and put it back in the 4/670MP. I put 384MB of RAM in
the 4/670MP, and I put all of the same disks that were on the SS-10 onto the
4/670MP with the exact same SCSI cables, cases, power supplies, etc. In other
words I'm not using the internal drive bays on the 4/670MP, everything is the
same except it's plugged into the 4/670MP system board instead of the SS-10.

The 4/670MP is running completely stable, it's been up three days without
a crash save for one incident where I accidentally turned the power off on it
(had it plugged into a switched outlet). No errors, no spinlocks, nada, just
hummin' along like a computer ought to.

So whatever is hosed is something specific to the SS-10. Just another
funny datapoint; SunOS on the 4/670MP when we were using it for a shell server
would crash about once every 2-3 weeks on average with a swtch message of some
sort, and IT runs completely stable on the SS-10's with Hypersparc while under
a substantial load.

uptime
9:02am up 52 days, 4:45, 74 users, load average: 7.73, 7.54, 6.50

I talked to a tech over at Solar Systems and found out I did not have the
proper PROM for supporting the Ross Hypersparc CPU's in that machine. So I
have the proper PROM now for it and friday one more time I will try moving
everything back. But I am not optimistic because the web server does have the
proper PROM and it's not stable either, though I only very infrequently see
disk errors on it and only on the Quantum drive (and I've had nothing but
problems with Quantums so the drive may just be on the verge of death). But I
still get the spinlocks on the web server that I am not seeing on the 4/670MP.

> If Sun wasn't so anal and actually helped us out a bit then we concentrate
> on fixing things. As it stands I have to scrounge around at uni to find
> hardware to work on.
>
> Anton

The rumours I've heard is that if you can find out what the correct ISBN
numbers, nearly all the documentation can be ordered, but that apparently is a
closely guarded secret. I do miss the Sun 3 days, I did develop a driver for a
funky terminal back when SunOS 3.5 was the current release and Sun sent me
everything I wanted free, books with all the chips, the register addresses,
etc. Granted, the documentation wasn't alwasy correct, but at least they made
the effort. Now all you can get is glossy sales literature. I share your
frustration with this.

I realize my remark about NT was a bit snide; frustration level here is
just getting a bit high. My electricity bill dropped considerably when I went
from the 4/670MP's to SS-10's but so did my ability to sleep through the night
without getting paged about a server being down.

I am curious if your 4/690MP news server is running a full news feed and
if so if it's having any difficulty keeping up?

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/