Thanks to:
Pat Wilson <paw_at_northstar.dartmouth.edu>
alan_at_nabeth.cxo.dec.com
Barry Treahy <treahy_at_allianceelec.com>
Martin Mokrejs <mmokrejs_at_mail.natur.cuni.cz>
for more input on the subject.
Pat Wilson wondered if the problem might be related to the disk spindown
powersvaing feature. This had indeed given me trouble in the past with
DU4.0B, and I had disabled it. I did check to make sure it was still
disabled after the upgrade.
alan_at_nabeth.cxo.dec.com suggested comparing the old and new ddr.dbase
files to see if the default features for unrecognised devices had
changed (we didn't have any custom entry for the Quantum). I should do
this to satisfy curiosity, but having been bitten twice now by this
drive model think I would rather avoid using it completely.
In fact, I did just look, and tagged queuing is enabled on both 4.0B and
4.0D - so the actual cause must lie elsewhere. Bad block recovery and
command reordering are enabled on 4.0B, and not in 4.0D. I don't see any
other differences.
Barry Treahy had much the same experience with Quantum drives in
general.
Martin Mokrejs informed me I should change the settings in
/etc/ddr.dbase to disable tagged command queuing. Again, I'll be looking
at this out of interest, but having already switched my system to use
some apparently trouble-free Seagate drives, I hopefully don't actually
have to solve the problem this way.
Thanks to everyone for all the help.
Graham Allan
Physics Department, university of Minnesota
My original message plus responses follow:
-----
This is more of a warning message than a question, perhaps.
I recently did an update of one of my Alphastation 255 systems from
Digital UNIX 4.0B to 4.0D. The disks in this system were Quantum
Fireball TM3200S (both system and user data).
The upgrade went without incident, but ever since then, the machine
would run normally for between 10 and 60 minutes, then hang completely.
Messages about i/o errors (deferring i/o on all available scsi devices)
appeared on the console. Shutdown was impossible, the only recourse was
the halt button.
It seems likely to me that the Quantum disks are to blame (especially
having transferred the system to a couple of Seagate drives, and it now
operating normally). This same model of drive behaved in almost exactly
the same way to me about a year ago, when fitted in a VAXstation 4000/90
(OpenVMS 6.2). That is, it would work just well enough to act as a
write-only medium - long enough to put data on it before hanging the
system!
The final answer is in dejanews somewhere under comp.os.vms, but if I
remember rightly, the Quantum drives don't perform command tag queuing
properly. I assume that the SCSI drivers in Unix 4.0D have been enhanced
to use this feature, with the consequence that the Quantum drives are no
longer safe to use. If, indeed, they ever were. I doubt I will ever use
a Quantum drive in anything fancier than a Mac in future.
Graham
Date: Wed, 25 Mar 1998 12:57:57 -0500
From: paw_at_northstar.dartmouth.edu
Message-Id: <199803251757.MAA13778_at_phibes.dartmouth.edu>
To: allan_at_mnhep1.hep.umn.edu
Subject: Re: Quantum disks, DU 4.0D
Hmm. I'm just wondering whether it might be something with the egregious
power management stuff (which spins down disks after some period of
inactivity) that keeps creeping into the kernel. Worth a check, at least,
before you completely give up on the disks...
Pat Wilson
paw_at_dartmouth.edu
From: alan_at_nabeth.cxo.dec.com
Message-Id: <9803251800.AA12503_at_nabeth.cxo.dec.com>
To: allan_at_mnhep1.hep.umn.edu
Subject: Re: Quantum disks, DU 4.0D
In-Reply-To: Your message of "Wed, 25 Mar 98 11:51:03 CST."
<fbfc772c48%gta_at_mnhep1.hep.umn.edu>
Date: Wed, 25 Mar 98 11:00:30 -0700
X-Mts: smtp
What may have changed is that the default features allowed
for unrecognized devices had tagged command queuing allowed.
You'd probably have to compare the new ddr.dbase and the
old one. Or, you may have had a custom entry for the
Quantum drives that didn't try to use TCQ.
Message-ID: <351947CA.E5A6520A_at_allianceelec.com>
Date: Wed, 25 Mar 1998 11:07:06 -0700
From: Barry Treahy <treahy_at_allianceelec.com>
Organization: Alliance Electronics
X-Mailer: Mozilla 4.04 [en] (Win95; U)
MIME-Version: 1.0
To: allan_at_mnhep1.hep.umn.edu
Subject: Re: Quantum disks, DU 4.0D
References: <fbfc772c48%gta_at_mnhep1.hep.umn.edu>
Content-Type: multipart/mixed; boundary="------------1957BF450A672E91FF09516D"
Stung once... Quantum is one of those manufacturers that I stay clear of,
period... I figured if I couldn't get them to work reliably on desktop PC's
(Compaq shipped a lot of systems with Quantum drives) years ago, I'd be damned
if I could trust them to my VMS and Unix boxes... Too many other more well
known players in the drive market that I've never had problems with, doesn't
make sense to revisit it, does it? Seagate has become my first choice too...
Barry
Posted-Date: Wed, 25 Mar 1998 19:49:45 +0100 (MET)
Date: Wed, 25 Mar 1998 19:49:45 +0100 (MET)
From: Martin Mokrejs <mmokrejs_at_mail.natur.cuni.cz>
Reply-To: Martin Mokrejs <mmokrejs_at_mail.natur.cuni.cz>
To: Graham Allan <allan_at_mnhep1.hep.umn.edu>
Subject: Re: Quantum disks, DU 4.0D
In-Reply-To: <fbfc772c48%gta_at_mnhep1.hep.umn.edu>
Message-ID: <Pine.OSF.3.96.980325194612.20882C-100000_at_prfdec.natur.cuni.cz>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Hello,
I sent your message to one the best folks in Linux world. I hope he will
mail you details about Fireballs. Not all Quantum drives are so bad.
Try to disable tagged queing in /etc/ddr.dbase file. You will probably
have to re-do ddr.db file. I don't how what's the proper way. Probably
using makemap which comes with sendmail-8.8.
'makemap dbm /etc/ddr.db </etc/ddr.dbase'
Mail me back how did you manage.
Martin
Received on Wed Mar 25 1998 - 21:52:11 NZST