Win 2003 Srvr doesn't boot after expanding smartarray 641 RAID5

[Pascal Roman]

This isn't dramatic as it is happening on a test server but I can't put the finger on the problem.

The setup: ML350 w/ 2GB RAM, Smartarray 641 RAID5 3x72,8GB, Windows 2003 server Std Ed.

The array is configured as a single logical drive divided in 1 4GB boot partition, 2 primary partitions and 1 extended partition holding 1 logical partition.

I want to expand the array with 2 additional disk, increase the logical drive size and finally all the partitions sizes (with the exception of the 4GB boot partition holding the OS).

I'm using the following method (which works fine when I'm replacing IDE HDDs with bigger ones in workstations):

-start the server with a Knoppix Live-On Linux CD
-create the cciss device
-copy the whole logical drive as an image file on another HDD (bit-to-bit with dd, over the network with netcat as the IDE controller of the ML350 doesn't support DMA)
-mount the partitions from the image file and open a few files at random to test the copy -> OK
-shut down the server, install the 2 extra HDDs, delete the logical drive and recreate it from the ROM.

From now on, I've tried those two variants with the same (lack of) results:

A)
-start with Knoppix again and create the cciss device
-using fdisk, create empty partitions at the new (increased) size

B)
-make a blank install of Windows 2003 server
-using the disk manager, create empty partitions at the new (increased) size

Then the last steps:

-using dd and netcat, copy each partition one by one from the image file into the new partitions on the server
-mount each partition and open files at random to test the copy -> ok
-remove the Knoppix CD and restart the server

During boot, the server hangs at "Attempting to boot from C:"

If I use the recovery console from the Windows 2003 server CD, I can access the files on the boot partition. Using FIXMBR and FIXBOOT hasn't helped.

I've tried a lot of ways to try and fix this. Instead of the "Attempting..." message, I've sometimes received a "A read error has occurred, Press Ctl-Alt-Del now" but I can't remember what I had done at this time.

However, if I remove 2 HDDs from the array, the server will start all right after restoring the partitions one by one using the same method as described previously.

I've also tried to install a GRUB to 'jumpstart' Win 2003. I need to read the GRUB manual pages further because I think that my setup was faulty. However, I've noticed this: when linux is loaded and the cciss device is created, the geometry of the logical drive is 255 heads, 63 sectors and 35419 cylinders. The geometry stays the same with a 3 or 5 HDDs RAID array, only the size of the cylinders change. During boot, invoking the geometry command of grub shows 255H, 63S, 1024C.

I suspect a geometry problem in the first stage of the boot. Could it come from the boot sector from the boot partition?

I could do with any pointer. Thanks in advance.

[Pascal Roman]

Oops, I forgot to add that I use ntfsresize to increase the size of the file system in the partitions freshly copied from the image file.

However, the boot partitions stays at 4 GB due to the smart array 641 limit.

[Pascal Roman]

Bump

[Matthijs Wijers_1]

Does the "default" entry in boot.ini reflects the right disk and partition?

Regards,
Matthijs

[Pascal Roman]

Yes, boot.ini looks ok.

Diskpart looks fine too.

[Pascal Roman]

I've found the cause, and the solution.

I hope that there is a good reason behind this mess because otherwise it would be pretty retarded.

On HP side: disk geometrys change when you change the array: 3 HHDs giving a capacity of 135 GB show a geometry of 255H, 32 S, 35419C. This is surprising because I thought that on modern drives, geometry was always 255H 63S when the capacity was > 2GB.
5HDDs giving a capacity of 270 GB show a geometry of 255H, 63S, 35419C.

On M$ side: apparently, the geometry is encoded in the boot sector of the system partition. If the geometry changes, the OS hangs or crashes at boot. Isn't that effing great?

To fix the problem, check http://support.microsoft.com/?kbid=167045 and http://support.microsoft.com/kb/119467/.

It works with the floppy. As I don't fancy having to boot from a floppy after the array upsize, I'll try the goofy trick of the small FAT first partition.

I've also tried with Linux (Debian), and reinstalling GRUB was enough to fix the problem.