Xref: utzoo comp.unix.ultrix:4706 comp.periphs.scsi:1142
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!zaphod.mps.ohio-state.edu!math.lsa.umich.edu!caen!paul
From: paul@caen.engin.umich.edu (Paul Killey)
Newsgroups: comp.unix.ultrix,comp.periphs.scsi
Subject: Re: ANy joy with 3rd party SCSI disks on DS3100?
Message-ID: <1990Sep28.175454.18103@caen.engin.umich.edu>
Date: 28 Sep 90 17:54:54 GMT
References: <1990Sep25.083654.9716@brolga.cc.uq.oz.au> <14630@cbmvax.commodore.com> <1990Sep25.233255.1551@brolga.cc.uq.oz.au>
Sender: news@caen.engin.umich.edu (CAEN Netnews)
Organization: University of Michigan Engineering, Ann Arbor
Lines: 40

Other scsi problems include not having the switches set correctly
on the drive that determine powering the scsi termination.  If that
is goofed up, so is the scsi bus, and you get the problem where
you can't see the disk (or much of anything) on the bus.  decs are
more susceptible to that than are other brands.

We run maxtor drives here (8380, 4380, 8760, lxt-200) on decs.
Also have had luck with micropolis and cdc or seagate or whatever
they are now.

I've noticed that the bits controlling different read ahead/caching
parameters seem to be different on different drives.  I hacked up
my own formatter.

When you format your drives, you might as well change the
zone (for bad block forwarding) to be a track, and not a cylinder.
I think the sun formatter does this for you automagically.  However,
I noticed that formatting disks on a sun 3 (emulex whatever scsi
controller) was kind of a loser, at least for running disks on the
decs, because of other parameters it (silently) changes.

rzdisk on 3.1 is a no-op for this, because you change the params
with the -c thing, but then it goes ahead and ignores the modified
params and formats the disk with the old (unchanged) saved params.

I am going to do an exhaustive search of all scsi option combos
and plot some results.  I am still mystified as to why fast bus and
fast disk do not add up to fast throughput.

While I'm at it, changing disk from synch to asynch will fix
some problems, but the real fix is usually getting newer firmware
that fixes a bug involved in some size data transfers.  E.g. level
9 dumps would hang, because of this, because of how dump would be
seeking around and reading bits and pieces to see what should be
saved.  level 0 dumps did not hang because dump knew it had to save
everything anyway, and skipped that part.

Sometimes turning off cacheing is also a fix for that.

--paul