[SunRescue] Proper use of VME scsi controllers?

BSD Bob bobkey at weedcon1.cropsci.ncsu.edu
Tue Feb 8 16:21:09 CST 2000


> The sun2 controller is Multibus controller (thus requiring one of the
> many Multibus-VME adapters) that is extremely brain-dead, somewhat
> slower than the sun3 controller, and probably not 100% SCSI compliant.
> It does work well enough for the 1/4" MT02 (SCSI->QIC36) interface and
> it makes a lot of sense to put tape drives on a separate bus from the
> disks they are likely to be used to backup, especially on older and
> slower SCSI subsystems.  Mixing disks and tape, especially on a sun2
> controller, will cause serious performance degradation.

Both the sun2 and sun3 controllers I have appear to be standard 2 plug
vme things.  They are in one of several different vme to ``multibus''
adapters.  I was thinking multibus was card edge and not the 3row vme
plug thing.  Maybe that was sun's adaptation?

For my built-up adapters, I used the internal/external adapter frames
from sun2 controllers, and dropped in sun3 controller boards.  That
gave me both internal and external I/O.  It seems to work fine, but
does not like long bus lengths, especially on older drives that load
the bus down.

I flipped controller boards around, and both the sun2 and sun3 boards
worked equally well at running one internal tape and two external drives,
but not more than that.  It could be that my drive cabling externally
is not as good as it could be (old sun cables).  Generally, there were
fewer problems with the sun2 controllers.  But, on the 3/260 and 4/260
I left the  sun3 controllers in, and put the sun2 controller in my
3/110 in a 12 slot case with tape and 2 internal ESDI drives.

My feeling is that the older controllers don't handle bus loading
very well.  It seems to take more power to drive the transitions
on the bus with the older full height drives.

> Sun has a long history of avoiding stupid user-support nightmares by
> simply saying that odd-ball and un-tested configurations are "unsupported".

OK, I was suspecting something like that.

> > 1.  What are the approved ways to run both internal and external devices
> >     on the scsi bus on say a 12 slot deskside cabinet?  How many devices
> >     and how should they be hung?
> 
> Anything's possible, within the realm of the SCSI bus physical specs.

Theoretically then I should be able to run a 6 meter cable with 7 devices.
It has a hard time with that on my machines or with my drives.

> > 2.  What are the approved ways to run two controllers, where one runs
> >     the tape, like a sun2 controller, internally, and the other, like
> >     a sun3 controller with external connector hacks runs the external
> >     bus.  Or, maybe run the tape and a pair of internal drives on one
> >     controller and external drives on the second controller.
> 
> Depends on what your other available peripheral hardware is, physically.

For the sake of discussion.....

1.  Tape in top bay on one controller, and 4 drives externally on the
    second controller.  What specific part numbers for the frame adapter
    should I use on the second controller to keep from blowing it or the
    backplane, and what slot/busjumpering is required?

2.  Tape and two drives in top bay all scsi, then 4 drives and external
    tape on the second controller.  How?


> Ideally you would put the sun2 controller in slot-7 (i.e. with a
> 501-1149 adapter) and hook the tape drive up internally.  Then put an
> external-only sun3 controller in another (supported) slot (that means
> getting the right 6u-9u adapter, probably a 501-1217, that does *not*
> connect to the backplane; and connect it to an external disk chassis of
> some (any!) description.

On my internal/external hacks, only the tape is inside, and the drives
are outside in shoeboxes.  The terminators have been removed on the board
and the tape has one end terminator and the end of the drive chain has
the other.   Total length about 16 feet of bus, within the 25 foot (6M)
standard.  It could be that the cabling is pushing things lengthwise
because of old less than perfect connections and cabling.  I was thinking
of cobbling up a metal slide plate with mounts for 4 drives internally
and then using the last slot for that as a place holder.  It would take
up about 2-3 slots or lose me 10-12.  That would take the cable run off
the internal header with a relatively short ribbon cable.  Power is
there aplenty for 3.5 inch drives from the top bay.  The other end
through the backplane would catch the internal tape.  Would something
lie that be reasonable?

> > 3.  How does one get an internal tape and 4 external drives on one
> >     sun3 controller?  I get maxed out at 2-3 drives and tape before I
> >     start getting scsi bus busy errors.  ONE time I got 4 drives and
> >     tape up, but it was not real happy.  There has gotta be a way
> >     to keep the bus happy, and lots of devices hung thereon.
> 
> I never had that problem on my 3/260s.  I had the internal tape, an
> internal drive, two external disk drives, an external CD-ROM, and an
> external tape (i.e. all target units were occupied).
> 
> You do have to follow the SCSI physical specifications very closely
> though.  I.e. you must not ever exceed the total cable length
> limitations (3m).  Avoid as many connectors as possible (i.e. use a
> large external chassis with a single internal ribbon cable).  Avoid
> using stub cables on devices (even though the spec. allows some).

I thought it was 6 meters?    I am beginning to think it cabling
and poor connectors more than anything else.
 
> Ideally you'll want to provide termination power from one or more of the
> drives in the external chassis and always be extremely careful to use a
> high-quality active (or forced-perfect) terminator on both ends of the
> bus.  So long as your external chassis power supply matches your
> internal one in voltages the use of multiple termination power sources
> on the SCSI bus should not harm anything.

I have been using passive termination.  I am sensing  that is less than
optimal?

> I used the on-board (active) termination on the internal drive (a wee
> quantum 3.5" drive) and a Sun-original active terminator externally on
> my SCSI chassis.  I had only the one internal ribbon cable from the
> backplane to the tape and disk, one external sun-original connector
> cable with the EMI filters, plus the internal ribbon in the external
> chassis.  This worked like a charm on my 3/260, and then on the 3/280 it
> was transformed into, and finally on the SS2 I use today.

Never seen one of those sun-active terminator things, unless that D shell
I have on the end is active.  I was thinking it was passive?

> > 4.  What can the kernel stand in the way of multiple scsi controllers?
> >     And, how should they be set up?  I sense two, but somewhere I thought
> >     I saw in one fellow's hackers notes that it could handle 4, with
> >     proper jumpering and setup.
> 
> Dunno.  I had trouble getting a sun2 and sun3 SCSI controller working
> simultaneously on my 3/260 with SunOS-4.1.1u1.  I didn't fiddle too much
> with the drivers though and I was mostly holding off until I could
> upgrade to NetBSD and help with the effort to complete the sun2 (sc)
> driver but of course that hasn't happened yet.  You might have better
> luck with your 4/260 if there's still an 'sc' driver in the sparc
> SunOS-4 version.

I will see if I can find those notes.  Maybe some juicy info is there.

> I've also as yet been unsuccessful at getting two sun3 (si) controllers
> working under NetBSD, but then the last time I tried was a couple of
> years ago and I didn't spend very much time on it either.

Mee too.....  always dma scsi timeouts, ad nauseum.  Yet, the 4/110 with
built-in esp-wierd runs fine..... VME dies.... oh, well.

Thanks  for the info....

Bob
 






More information about the rescue mailing list