Hello TSMers, or more precisely, the intersection between the TSMers and
WINDOWSers
I spent the last 24 hours struggling with the ASR bare metal recovery
process with W2K3 and TSM. Got some good things, and some strange things which
I would like to share/comment, and also seek for some answers. Here's the story:
First of all, I used TSM client 5.3.2.2 to backup all disks on the servers,
including System State/System Services. Open Files backup also working fine.
First machine to be recovered was a test machine with one disk split in two
partitions (C: and T:). Restore went file, and ASR recreated my two partitions
and restored C:. The T: partition was left empty, but this is a minor problem.
After this degree of success, we went for the restoration of a MS Cluster
node running SQL Server, on a brand-new IBM xSeries 366. Here begins the real
story....
Problem #1: This machine uses SAS disks and matching RAID controller, for
which I had to provide the drivers. When I pressed F6 during text-mode setup,
it prompted for the driver, an went on, but then it wouldn't accept the ASR
diskette anymore. I finally solved it by integrating the RAID controller driver
into the W2K3 CD.
Problem #2: The network cards present on the machine (6 in total) are not
recognized by the standard Windows setup. After a call to MS's support, I got
the answer that "only local restores are supported". After many hours and
iteractions of "burn CD-RW (15 min)/boot the server (20 min)" I could get
Windows to accept a set of network drivers integrated as OEM drivers pointed by
the "Unattended Install" mode. I had some dark thoughts because the ASR process
uses its own kind of response file, and soon enough I was feced with the
effects of this. After the device detection, Windows setup would simply freeze.
After some 10 minutes, I gave up I went to reboot the machine. By chance I
pressed Enter and... oops!! Setup resumed.... Not too good to document, I
think. "If setup freezes, get a cup of coffee an try pressing Enter several
times"....
After that, Windows finally came back, but it started complaining on every
boot that the cluster service could not be started because the quorum disk
could not be accessed (Quite right, since the other node has it). But after 60
sec, cluster service restarts and the node joins the cluster normally. That is
problem #3.
The last problem# is by far the silliest.... The (standard) MS Sql server
install placed in the registry the SHORT file names (intended for 16-bit
programs that must use the old 8.3 naming convention). So, the directory named
"Microsoft SQL Server" had a short name of MICROS~1 before the ASR process, and
MICROS~2 after (because there was a "Microsoft MOM" directory which was
restored earlier ad got MICROS~1). Obviously SQL Server wouldn't start, and I
had to rename both directories to get it to work.
I am currently trying (after some sleep) to solve these so I can get a
decent automated process. If someone gets any light on these questions please
let me know.
Thanks to all,
Paul van Dongen
|