On Friday 21 January 2005 11:37, Brian Cuttler wrote:
>Stefan,
>
>On Wed, Jan 19, 2005 at 10:48:54PM +0100, Stefan G. Weichinger wrote:
>> BC> Major imporvement though still below the acceptable threshold.
>>
>> BC> Definitely looking like it was never an amanda issue, this
>> looks to BC> be an underlying architecture problem.
>>
>> I hope that you spot the issue soon, sounds much better already
>> ;-)
>
>Always the simple stuff. Block size (still need to find out how to
>reset at boot time), SCSI bus length...
Put the appropriate commands, as you would exec them from a root
shell, in the /etc/rc.d/rc.local file. They'll work as long as there
is a loaded tape in the drive during the bootup.
>Confirmed with SGI engineer, with the Narrow device on the bus we
>are limited to 1.5 meters, we'd have been ok with WIDE devices, 3
> Meter limit but we are clearly over.
What??? Is??? He??? Smoking??? I want some of that! Its got to be
great stuff...
The scsi bus, properly setup and terminated, can be as long as 30 some
meters, and I have personally ran it well over 10 feet!
But with poorly matched stuff, you are often in the virgin sacrificing
business at 18 inches or less. This 'engineer' (note the lowercase,
its intended) obviously does not understand that the scsi bus is a
transmission line, and subject to the usual rules regarding control
of vswr. VSWR=Voltage Standing Wave Ratio, and is a measure of the
raltive magnitudes of the signal power going down the cable, and
coming back up the cable from the far end. Ideally, with proper
terms in place, there is no echo from either end as the terms
perfectly absorb the signal by presenting a load to the cable that
matches the cables impedance very closely, making the cable look to
be of almost infinite length.
The bus itself is wired OR, open collector style, meaning that any
device, anyplace on the bus, can exert a pulldown to a logic zero,
and thos signal will propagate both ways on the cable from a device
in the middle, with no echo's comeing back from either end of the
cable under ideal conditions.
In order to control echos or vswr as the rf folks call it, it must be
terminated with its characteristic impedance at both ends of the
cable. Note that this means the extra lines for a wide bus MUST be
terminated at the physical end of that wider portion of the bus,
preferably with whats known as an "active termination" as that
typically matches the cables impedance much better than the usual
powered resistor packages do. Then the rest of the cable, the narrow
part, needs to be terminated at its physical ends too, with no spare
cable going on by that point because of the echos generated by the
open end of the cable if the terms are not actually at the end of the
path.
The card normally terminates the src end of the cable, usually with
active terms today, but its toss a coin to see what you get from the
aftermarket folks unless you specify, and accept and pay for, no
less.
Let me get a view of what your logic one noise margin is Brian, grab a
multimeter and poke into an unused connector of the cable and read
the resting voltages of 3 or 4 of the lines and tell me what your
meter reads please. All lines on one side of the connector should be
grounded, so I'm interested in the 2.x volts you'll read on the other
lines.
>Will see if I can't get permission to remove SDLT/shoebox and see if
>the jukebox doesn't work properly. In the longer term I think we've
>convinced the machine owner to purchase/install an additional SCSI
> card.
They aren't THAT expensive, but non-technical bean counters often
cannot see the value obtained from the expense spent because they
don't understand the value received is often a do, or don't, and
there is no middle ground scenario. You often have to explain it in
terms of if you want it to work, we spend x dollars to make it work.
Otherwise its no workee. Then its simple. Give them a middle ground
where it might work, and they'll assume you walk on water and can
make it work everytime. We're quite often not that lucky in this
business.
Worth every penny in most cases.
Good luck.
--
Cheers Brian, Gene
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.32% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attorneys please note, additions to this message
by Gene Heskett are:
Copyright 2005 by Maurice Eugene Heskett, all rights reserved.
|
