[2628] in linux-scsi channel archive
Re: SCSI medium error
daemon@ATHENA.MIT.EDU (Michael Weller)
Tue Oct 14 06:32:41 1997
Date: Tue, 14 Oct 1997 12:29:04 +0200 (MESZ)
From: Michael Weller <eowmob@exp-math.uni-essen.de>
Reply-To: Michael Weller <eowmob@exp-math.uni-essen.de>
To: Marc SCHAEFER <schaefer@alphanet.ch>
Cc: David J Brenneman <brennemd@emu.edu>, linux-scsi@vger.rutgers.edu
In-Reply-To: <Pine.LNX.3.96.971012193448.3564A-100000@vulcan.alphanet.ch>
David J Brenneman wrote:
> On Sun, 12 Oct 1997, David J Brenneman wrote:
[Some Scsi error; Question about it's meaning]
Well, it mentions 'medium error' and 'unrecoverable read error' in there.
Is it possible for a message to be any clearer?
[Some suggestions of mine for other messages including a few four letter
words about the state of the disk removed from this mail]
> > Thanks for the help, what I want to know is, is the disk going to
> > crash completely and unrecoverably, soon? some opinions I have heard
Well, if it is only one or two bad sectors showing up on occasion it's
probably ok. If it's more, get a new disk NOW!
> > seem to indicate this... I have heard that the Quantum Fireball has
> > a relatively short life expectancy.
Can't comment on this.
On Sun, 12 Oct 1997, Marc SCHAEFER wrote:
>
> Normally not. It is possible that bad blocks create over time.
> Usually they should be replaced. Maybe the scsiinfo command
> can do replacements, I do not know, sorry.. Or maybe the
No, instead the disk must do these replacements itself. However, it must
a) be configured to do replacements.
b) Have large enough replacement areas for the data available.
I strongly encourage you to get scsiinfo and check if block reassignment
is enabled, if there is large enough replacement space reserved, how many
new (not factory) bad blocks already developed.
Scsiinfo can do both of this, version 1.7 (which seems to be around now)
contains a short readme of mine how to do all this. Also, if b) is not the
case and you change it, the disk MUST BE lowlevel formatted to make your
change take effect. (scsiinfo 1.7 has a lowlevel formatter too).
I found that usually there are no replacement areas allocated on the disks
as factory default as they reduce the capacity of the disk and noone cares
for stability anyway (at least 95% of their customers, that is), BTW. Esp.
on disks which are not primarily intended to be used in servers (which
is probably valid for the Quantum)
> kernel can be triggered to do automatic reassign blocks ?
No, it's the job of the disk to do that.
> I personnally, when I encounter a bad block, use the badblock
> command or the host adapter facilities to scan the disk and
> replace all bad blocks.
Well, ext2fs (and other fs's) contains support for a badblock list. I
don't know though if it is possible to add any badblocks to a filesystem
ALREADY created.
If so, either e2fsck or debugfs will be the tool to do it. badblock simply
reads the disks and produce a list of the faulty blocks. NOW: If your scsi
disk does properly reallocate bad blocks it will do so for every badblock
which 'badblock' hits.
Similar for the host adapter facilities. I doubt your host adapter tools
have any knowledge about the structure of a linux fs. Instead they'll do
A BUNCH of (hopefully non destructive) reads and writes s.t. the disk will
notice (and replace) and bad blocks. Well, ok. There might be a command to
force a reallocation (I'm unsure right now). But in any case you'll have
to have some spare space on your disk for it.
BTW, I'm a purist with regards to disks. I replace them, or at least have
a new disk in the machine and only use the old disk for less important,
well backuped or easily reproducable data when it starts to evolve ANY bad
blocks. I take these reallocations as an early crash warning.
Michael.
(eowmob@exp-math.uni-essen.de or eowmob@pollux.exp-math.uni-essen.de
Please do not use my vm or de0hrz1a accounts anymore. In case of real
problems reaching me try mat42b@spi.power.uni-essen.de instead.)
