The original question was:
>I often get SCSI bus timeouts while reading one of my SCSI disks during
>a full dump to an EXB-8200 on the same adapter. I thought this was due
>to the cable length, which is pushing 8 meters (if you count the cable
>inside each of the boxes, which makes it hard to say, electrically, if
>it's really too long). But, after shortening it as far as possible, I
>no longer get timeouts while *using* any of these disks, just when
>dumping them.
[...]
>I had to re-run my weekly full dump several times last week, and it's
>getting to the point where I'm about to spend $1000 to buy a VME SCSI
>adapter just for the $2500 tape drive. Do I have to? What would you
>do? Whose fault is this? Exabyte? Sun? Mine?
I got several responses, but no satisfying "click", if you know what I
mean. Several people suggested that the maximum bus length is 6m, not
8m (even shorter for successful sync operation), which might go a ways
toward explaining it. Another said that moving the EXB closest to the
host, i.e. first in the chain, fixed such a problem for him. But he
said "desktop", and this is a 4/380; also the EXB-8200 is *already* the
first thing in the physical cable chain.
Kevin Sheehan (who ought to know, since he wrote most of the SCSI
drivers) told me that I'm not *really* doing sync SCSI because the 4/300's
sm0 can't do it, and that's why I don't see messages to that effect at boot.
Well, harrumph! I demanded that my salesman promise that this worked before
I bought it. (Lambert: Comment? :-) The disks did seem to work more quietly
than on the 3/180+sc0, and they *sounded* like they did when on my 4/65,
where they were *definitely* doing sync. Hard to say. Will someone step
forward with a believable, absolutely definitive answer to this?
He also corrected my terminology: I had an "sc0", not an "si0" Multibus
SCSI board lying around idle, and told me that Sun no longer supports
it. Experiments showed that the 4/380 liked it, and the disk on it
showed up in the boot probe, but format(8) couldn't find it, so I gave
up on that idea. I might be able to configure a kernel that makes it
work, but I decided it's not worth the effort. (Anyway, the problem is
on "sm0" in the 4300; the "sc0" is just a side-project.)
The "Jumbo ST Patch" (100280-02) was suggested, but faintly.
So, it's all a hodgepodge: no clearcut answers, only suggestions for
further study and experiment. *sigh* Well, it helped me make up my
mind that I need to pop for that $1000 "second SCSI board" for my
4/380. Maybe *that* will do sync, and I'll put all the disks on it,
and put the tape, all alone, on sm0. Very disappointing.
Many thanks to:
Kevin Sheehan <kevins@Aus.Sun.COM>
Mark Seiden <mis@seiden.COM>
Mike Raffety <miker@sbcoc.com>
Daniel Strick <dan@breeze.bellcore.com>
Marc Phillips <phillips@athena.Qualcomm.COM>
P.S. Sorry this is so late; I've been called away from this sort of
thing lately, because I had to get some "actual work" done. :-)
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:19 CDT