NON-SUMMARY: Mismatch between physical geometry and disk label contents

From: Johan Hartzenberg <jhartzen_at_csc.com>
Date: Wed Apr 23 2003 - 02:02:21 EDT
Hi Gurus,
Unfortunately, I did not get any solution.  I had suggestions to just
"Label" the disk, but my problem is not fixing a disk with an incorrect
label, but rather finding such disks (LUNs) across many systems because
these disks are currently "in use" and as a result, some portions of the
disk is not available.

Richard Hobbs mentioned having seen the same thing in the past when moving
disks from one system to another.

Using "format" to label a disk with an existing label does not re-read the
disk geometry from the physical disk, it reads it from the existing label.
The only way to really re-label a disk is by first corrupting the label, eg
using the following procedure:

First, corrupt the first sector on the disk c#t#d#
      dd if=/path/to/any/file of=/dev/rdsk/c#t#d#s2 count=1 bs=512

Next, use format to write a new label to the disk
      format
      select disk
      label? YES
      quit

Then finally, encapsulate the disk, or partition and start using it as per
normal.


My Original question:

> > Hi gurus,
> >
> > I was happily scripting a couple of weeks ago and came across something
> > that produced a message similar to this (off the top of my head)
> >
> > Warning: The label contents does not match the geometry reported by the
> > physical device.
> >
> > Now I can not remember the command I ran, but I want to use it in
another
> > script.  I was busy writing a script to report on raw disk sizes as
seen
> by
> > Solaris (not Veritas), and I was playing with commands like
> >
> > format, prtvtoc, fmthard
> >
> > I've been searching the web and sunsolve, but I can't find out what it
is
> I
> > did.  Makes me want to "script" to record everything I do... !
> >
> > The situation I have is that sometimes when the storage administrators
> > allocate a whole bunch of LUNs I do not check each individually, and
> > sometimes I don't realise I am not seeing the full LUN (eg if the first
> > sector of an expanded LUN was previously the first sector of a smaller
> > LUN).  I am trying to write a script which will tell me where I may be
> > using LUNs without getting access to the whole full size of the LUN.
> >
> >
> >
> > moreover, I have tried to write something to compare the output of
fdisk
> -g
> > with that of fdisk -G, but I always get the following message:
> >
> > # fdisk -g /dev/rdsk/c3t11d4s2
> > * Label geometry for device /dev/rdsk/c3t11d4s2
> > * PCYL     NCYL     ACYL     BCYL     NHEAD NSECT SECSIZ
> >   39516    39516    2        0        15    96    512
> > # fdisk -G /dev/rdsk/c3t11d4s2
> > /dev/rdsk/c3t11d4s2: Cannot get physical disk geometry.
> >
> >
> >
> > devinfo appears totally useless, and on Solaris 2.6 I get for -g and -G
> > both:
> >
> >       Disk does not support fixed disk partition table
> >
> >
> > Any ideas/suggestions ?
> >
> > Thanx,
> >  _Johan








|---------+--------------------------->
|         |           "Hobbs, Richard"|
|         |           <hobbs          |
|         |           @mongeese.co.uk>|
|         |                           |
|         |           17/04/2003 04:14|
|         |           PM              |
|         |                           |
|---------+--------------------------->
  >-----------------------------------------------------------------------------------------------------------------|
  |                                                                                                                 |
  |        To:      Johan Hartzenberg/GIS/CSC@CSC                                                                   |
  |        cc:                                                                                                      |
  |        Subject: Re: Mismatch between physical geometry and disk label contents                                  |
  >-----------------------------------------------------------------------------------------------------------------|




Hello,

How do you know the disk label is valid though?? What told you that it was
valid?

To produce the error, i took the 2GB disk out of my Ultra 1, and then took
two
9GB disks out of an old veritas-managed disk array. I then put the two 9GB
disks
into my Ultra 1, and tried to install Solaris 9 onto them. These two disks
were
going to stay in my Ultra 1 permanently from now on.

At some point in the install (fairly early on), i got the error you had...
about
the wrong physical geometry. I checked several things:

 - Whether that type of disk would actually work in my machine at all.

 - Whether the disk type was compatible with the bus i had plugged them
onto.

 - Whether the disks were actually dead (by putting them in a working
array).

 - Whether there were any master/slave jumpers on the disks.

 - Whether a "probe-scsi-all" command at the ok prompt could pick up the
disks.

 - Whether a boot into single user-mode, followed by a 'format' would help.

 - ...and finally, i tried labelling them, and that worked!

I don't know too much about it technically, i'm afraid, but there's my
story.
Hope it helps!!

Try posting on my forum!! http://unixforum.co.uk

Thanks,
Richard.


Quoting Johan Hartzenberg <jhartzen@csc.com>:

>
>
> Hi,
>
> Thanx for your reply.  The veritas stuff is not the problem, the problem
is
> that the disk label was valid, but the geometry was "wrong".
>
> What did you have to do to actually get the system to produce this
message?
>
> So far I only got your reply, plus one from someone who didn't properly
> read my question!
>
> Thanx,
>   _Johan
>
>
>
>
>
> |---------+--------------------------->
> |         |           "Hobbs, Richard"|
> |         |           <hobbs          |
> |         |           @mongeese.co.uk>|
> |         |                           |
> |         |           17/04/2003 03:45|
> |         |           PM              |
> |         |                           |
> |---------+--------------------------->
>
>
>
-----------------------------------------------------------------------------------------------------------------|

>   |
>                                       |
>   |        To:      Johan Hartzenberg <jhartzen@csc.com>
>                                       |
>   |        cc:
>                                       |
>   |        Subject: Re: Mismatch between physical geometry and disk label
> contents                                  |
>
>
>
-----------------------------------------------------------------------------------------------------------------|

>
>
>
>
> Hello,
>
> I had that problem when i put two disks from a veritas-managed array into
> an
> Ultra 1. It said the disk was reporting geometry that didn't match the
> physical
> disk.
>
> I simply had to label it i think. Veritas doesn't label disks the same
way
> Solaris does, so this could be the problem.
>
> I'd be very interested to see the summary when yo figure it out though!
>
> Thanks,
> Richard.
>
>
> > _______________________________________________
> > sunmanagers mailing list
> > sunmanagers@sunmanagers.org
> > http://www.sunmanagers.org/mailman/listinfo/sunmanagers
> >
>
>
> --
> Richard Hobbs
> hobbs@mongeese.co.uk
> http://mongeese.co.uk | http://unixforum.co.uk
>
> "There's only one way of life, and that's your own" - The Levellers
>
> _____________________________________________________
> Send all your jokes to jokes@fishsponge.co.uk !!
> To subscribe, email: jokes-subscribe@fishsponge.co.uk
>
>
>
>
>


--
Richard Hobbs
hobbs@mongeese.co.uk
http://mongeese.co.uk | http://unixforum.co.uk

"There's only one way of life, and that's your own" - The Levellers

_____________________________________________________
Send all your jokes to jokes@fishsponge.co.uk !!
To subscribe, email: jokes-subscribe@fishsponge.co.uk
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Wed Apr 23 02:12:24 2003

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:09 EST