ADSM-L

Bare metal recovery of Windows NT

1999-04-21 12:17:28
Subject: Bare metal recovery of Windows NT
From: Thomas Denier <Thomas.Denier AT MAIL.TJU DOT EDU>
Date: Wed, 21 Apr 1999 12:17:28 -0400
My site just went through a disaster recovery test. In the process we
discovered a critical (and as far as I can tell, unstated) assumption in IBM's
documentation of bare metal recovery processes for Windows NT.

The Windows system restored during the test did not have a separate recovery
partition, so we used a variant of the method IBM suggests. We installed
Windows NT and the ADSM client from CDROM into directories other than the
normal ones. We then restored all of the files originally present in the
partition and attempted to reboot. I think we would have had the same problems
if we had used a separate recovery partition.

We were successful in recreating the original population of files.
Unfortunately, some of the files in that population were incompatible with the
hardware environment at the hot site. Our production system ran on single
processor HP system with the C partition on an IDE drive. Our hot site vendor
provided us with a dual processor Compaq with only SCSI drives. Our Windows NT
specialists are still working on documenting the machinations they went
through to get the system running at the hot site. As nearly as I can figure
out what happened, they had to edit the boot.ini file in the root directory to
get it to show the right disk type, hardware address, and partition number for
the partition containing the Windows system, and then copy a number of files
from the \wintemp directory tree (the one containing the copy of Windows
installed from CDROM at the hot site) to the \winnt directory tree (the one
containing the copy of Windows restored from the ADSM server). These included
the hardware-oriented registry hives, hal.dll, and a file whose contents
depended on the number of processors.
<Prev in Thread] Current Thread [Next in Thread>