Re: [Scsirastools-developers] New user

[Cress, A. R <and...@in...>]

Daniel,

The sgsafte log was revealing. =20
Apparently most of your Fujitsu disks seem to have exactly the same
serial number.
The serial number is used to judge when Linux is confused and reports
the same physical device on another /dev/sg* device (an impostor), so
they get marked failed.  It must be unique by the SCSI-2 standard.=20

Hmmm.   Fujitsu often has longer 12-byte serial numbers and is in the
vend12 list in sgsafte.c to handle that.  There is also a -m option to
tell it to use longer 12-byte serial numbers, can you  try sgsafte -m ?


Andy

-----Original Message-----
From: Daniel Smolik [mailto:ma...@my...]=20
Sent: Friday, February 23, 2007 4:07 PM
To: Cress, Andrew R; scs...@li...
Subject: Re: [Scsirastools-developers] New user

Cress, Andrew R napsal(a):
> Ah, now we are getting somewhere.  There is a later version of mdadm
> that might compile better than the one I have copied in this tree. =20
> But, the main issue:
> It looks like we have functioning sg* tools, but there is an issue
with
> the devices on bus 5:
> 4 /dev/sg4 /dev/sdd 5:0:0:0 Disk SEAGATE  ST39103LCSUN9.0G 034A
> 9949862528 ready
> 5 /dev/sg5 /dev/sde 5:0:1:0 Disk
> failed
>=20
> The next step would be some debug.
> 1) "cat /proc/scsi/scsi" to see if these devices show up there.
> 2) "sgsafte -x" and attach /var/log/sgsafte.log
> 3) "sgdiag -x" and attach /var/log/sgdiag.log
> 4) "tail /var/log/messages" to see if there is any detail around the
bus
> error.
>=20
> This should tell me why a bus error is occurring. =20
> It could be something physical, like termination, or it could be a
> software bug of some sort. =20
Hi,
there are some logs that you want.

			Regards
					Dan