[1106] in linux-scsi channel archive
Re: Oh yuck - now my Seagate disks are brain-dead?!
daemon@ATHENA.MIT.EDU (Andy Poling)
Mon Dec 16 03:07:02 1996
Date: Mon, 16 Dec 1996 03:03:28 -0500 (EST)
From: Andy Poling <andy@realbig.com>
To: linux-scsi@vger.rutgers.edu
In-Reply-To: <Pine.LNX.3.91.961214214930.17558A-100000@ws.realbig.com>
Well, I don't know if it's a good sign or a bad sign when you end up
following up your own message with the answer...
I wrote:
> A funny thing happened on the way to building a RAID-5 array on my DPT
> 3334UW. It apparently somehow scrambled the disks' manufacturers params.
>
> This SCSI system consists of a DPT PM3334UW (SCSI ID 0) and 5 Seagate
> ST-31055W disks (SCSI IDs 1-5).
[...]
> After a bunch of screwing around, I have determined that the drives spin up
> on command, and identify themselves (partially) correctly as ST-31055W's,
> but any attempt to read or write yields a check-condition (0x02) status with
> a "medium error" (0x03), with a sense code of 0x31 and a sense code
> qualifier of 0x00. This is according to my Adaptec 2940. The DPT just says
> "device not ready".
[...]
> 1) what happened? Did I do something I shouldn't-a done?
I still don't know for sure what happened here, but my best guess is that
using write-back caching on the controller is a bad thing. I've now set all
disks for write-through caching. What a shame. Maybe I'll be brave and
attempt the write-through cache again later.
> 2) what do I do with these disks? Given that I still have two sane disks
> and they're all identical, I still hold out hope that I can somehow
> either restore the default manufacturer params (perhaps via one of the
> "reserved" jumpers on the disks?) or copy them from a working disk to a
> non-working one.
The answer: format them. It took me a while to get up the nerve because
I've been told that it's a bad idea to low-level format a SCSI disk. but I
figiured I had little to lose, so I tried it and it worked.
All of my disks are once again working properly.
-Andy
(this sig left intentionally blank - too many hats for one sig)
