6 May
2013
6 May
'13
7:51 p.m.
On Mon, 6 May 2013, David Riley wrote:
On May 5, 2013, at 5:01 PM, Tothwolf <tothwolf at
concentric.net> wrote:
How long is the SCSI chain? The first thing that
came to mind for me
was the possibility of a SCSI bus issue since I've seen similar
behavior with other systems. In fact, since you just mentioned the Jaz
drive, it could very well be the culprit. I had major compatibility
issues with Jaz drives in a non-PC application back when they were
current products and I ended up having to connect them to a PC to
update their firmware and change their internal settings. A quick
Google search turned up this link too:
http://www.linux-m68k.org/faq/howjaz.html
I'm familiar with the general problems with Jaz drives, but this
particular one has never given me much trouble. Having workable disks
around is a problem, though; they all seem to die over time, even
disused (unlike Zip disks, which have always been fine for me). I got
10 off eBay and they all failed a long format, which I hope doesn't
actually mean the drive in the Mac I was formatting them has gone off
(not impossible). At this point, I don't think I even have any Jaz
disks that pass the long format.
In any case, the SCSI chain isn't hideously long; 4 devices, all
connected with 3-foot cables, so it should be well within spec. Nice,
thick cables as well. Disconnecting the Jaz drive doesn't seem to have
solved the problem; next I'll try the Zip drive. It's also altogether
possible that the update to the latest firmware on the CQD-220 has
introduced some instability in the card itself, since the error logs
seem to be indicating that the controller needs to get reset.
Specifically, I'm getting this error once in a while: MESSAGE TYPE 0010
IMMEDIATE MODE COMMAND TIMEOUT
_ CONTROLLER RESET
It's followed by a few other entries documenting
the reset and init sequence.
My thinking is that the controller shouldn't be timing out on MSCP commands
just because of some balky devices on the SCSI bus, though having seen some
of the failure modes of the firmware of the CQD-220, I wouldn't be entirely
surprised.
Resets are common when there is data corruption or noise on a SCSI bus.
With single-ended Fast SCSI, the recommended max cable length is 3 meters
(~10ft) so you are already close to the length limit, if all the devices
are well behaved. You might try using active terminators and/or shorter
cables and see if that helps too.
You don't happen to have a wide (16 bit) device attached to the SCSI bus
with an adapter do you?