Subject: Re: Dual SCSI, single chain?
To: None <>
From: der Mouse <mouse@Rodents.Montreal.QC.CA>
List: port-alpha
Date: 03/29/1999 22:10:38
>> As Wilko says, there are a multitude of factors that may prevent
>> these things from working.
> Right.  And as he also said, he has arrangements like this that do
> work that he uses himself.

This is not directly relevant to NetBSD/alpha, but I do have a couple
of contrasting SCSI war stories.

One is on an Auspex, a higher-end machine designed to be an NFS
fileserver.  (Much the sort of target market as the Network Appliance
boxes, but done very differently, done as a full general-purpose
machine that happens to have add-ons to improve NFS server performance,
rather than a machine so stripped-down that it can't do anything *but*
NFS service.  If anyone is interested I can go into the reasons we
picked Auspex over NetApp, but that is *definitely* off-topic, so ask
me privately if you're interested.)

The other is on a SPARCstation 1+ running NetBSD/sparc.

On the Auspex, we had disks flaking out and we eventually called for
hardware support to swap the board with the SCSI interfaces on it; we
eventually discovered that the problem was actually termination - once
we replaced the passive terminator (fine for most situations - it was
not just a busted terminator) with an active one all the problems went
away.  It was very touchy about termination.

On the other hand, my SS1+ at home has a Seagate disk on it.  I had
been adding things to and pulling things off of that SCSI bus, dumping
new source trees onto disks and the like.  (In case anyone is curious:
esp0 at sbus0 slot 0 offset 0x800000 level 3: ESP100, 25MHz, SCSI ID 7
scsibus0 at esp0: 8 targets
sd0 at scsibus0 targ 0 lun 0: <SEAGATE, ST12400N, 8650> SCSI2 0/direct fixed
sd0: 2048MB, 2621 cyl, 19 head, 84 sec, 512 bytes/sect x 4194685 sectors
.)  At one point, after rebooting after the latest changes and working
for a while, I realized I was idly toying with the (external)
terminator, did a double-take, and checked - and found the SCSI chain
was completely unterminated at the away-from-host end.  And neither the
host nor the disk seemed to care.  (I've also seen SCSI busses work fine
when terminated at both ends *and* at some intermediate device.  I even
saw one mostly-work when "terminated" at *four* points.)

It can go either way.  Some things are touchy, some are tolerant, and
there even seem to be a few that are just randomly cantankerous.

					der Mouse

		     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B