Well, I've had to give up :-( What I did format begin a format (knowing I'd
probably lose any hope of getting data off the slice I was hoping to get
dat aoff of...) And interrupted the format just as it began the
verification passes.
I used "fmthard" to re-label the disk and it took. I tried to mount the
stripe set but couldn't because the mount did not see the set as a UFS
fstype :-( I ran "fsck" in "-Y" mode after specifying an alternate/backup
boot block ("-b" option of "fsck"). It found LOTS of problems :-(
After it was done, I had about 300Mb of files restored (these were
undoubtedly the ones that did not stripe across the problem disk slice.
Thanks to the following for trying to help me with my problem.
Charlie Mengler <charliem@anchorchips.com>
Matthew Stier <Matthew.Stier@tddny.fujitsu.com>
Mike Ghicas <mghicas@debit.stern.nyu.edu>
Ramindur Singh <ramindurs@yahoo.com>
Their response and the original post are below.
At 6:48 PM -0500 3/28/99, I wrote:
>...I've got a bad disk drive :-( which was part of a DiskSuite stripe set. I
>feel that the only problem with it is that it is now unlabelled. I've tried
>fmthard but it fails with a message that the disk is unformatted.
>
>The man pages for the "fmthard" say that I should use "format" for writing
>a disk label on an unlabelled disk. "Prtvoc" on the drive gives an error of
>"unable to read disk geometry".
>
>Any ideas? TIA.
-------------- Responses -------------
From: Charlie Mengler <charliem@anchorchips.com>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>What OS & version?
>Has the disk physically been replaced?
>How certain are you that the disk is really OK from a hardware reliability
>standpoint?
>
>What has been done between when it was last working & now?
From: "Matthew Stier" <Matthew.Stier@tddny.fujitsu.com>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>Sounds like you used cylinder 0 in a raw partition used by a database
>application.
>
>The disk geometry is written into the first block on the device. Applications
>like mkfs (newfs is a wrapper around mkfs) understands this and does not use
>the first few blocks in partition 0 in any filesystems it creates. DiskSuite
>also understands this, and indexes its metadatabase for the same reason.
From: "Mike Ghicas" <mghicas@stern.nyu.edu>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>If it is a standard drive (seagate, etc) and the drive is good, you should be
>able to read the geometry. Are you sure that the drive is not completely
>fried? Have you tried to replace it with a new disk and see if the problem
>still occurs? Format (is the right command to label the disk).
>
>One way to check to see if the drive is really dead: from the ok prompt, do a
>probe-scsi (if you have an ultra1 or better, you must first do a "setenv
>autoboot? false" and reset the system - then dont forget to set autoboot back
>to truewhen you are done). If the drive does not show up properly after doing
>a probe-scsi the disk is dead.
>
>If it doesshow up okay, then i'm not sure why you can't label it.
From: Ramindur Singh <ramindurs@yahoo.com>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>Long shot but have you tried using format?
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:13:17 CDT