Let me try to explain my problem. I have a system with three identical hard drives. Two drives are used in a RAID 1 array, the third drive is used as a backup drive. The two RAID drives are connected to SATA 1 and SATA 2 on my computer. The third is connected to SATA 3.

Let me call the two RAID drives HDD-R1 and HDD-R2, and the backup drive HDD-BU. At boot, grub numbers them as follows:

HDD-R1: hd0
HDD-R2: hd1
HDD-BU: hd2

In my grub menu.lst, I have setup grub to first try to boot from hd0 (the first RAID drive), and if that fails try to boot from hd1 (the second RAID drive).

For testing the RAID, I simulated a fatal disk failure by disconnecting HDD-R1. The system booted fine. I then reconnected HDD-R1 and removed HDD-R2. Again in booted fine. I reconnected both HDD-R1 and HDD-R2 and tested the fallback again, but this time by simply choosing from which harddrive to boot from the grub menu. When I chose to boot from hd0, all was fine. But when I chose to boot from hd1, it gave me an "Error 15: File not found".

It took me a while to figure out what had happened.

When I first disconnected HDD-R1, grub renumbered the drives as follows:

HDD-R2: hd0
HDD-BU: hd1

It booted fine, not because grub chose the fallback boot drive, but because it simply succeeded booting from hd0, the default boot option.

When I removed HDD-R2 and reconnected HDD-R1, grub numbered the drives as follows:

HDD-R2: hd0
HDD-BU: hd1

Again, it succeeded to boot using the default boot option.

In the last test, when I reconnected HDD-R1 again, grub numbered the drives as follows:

HDD-R2: hd0
HDD-BU: hd1
HDD-R1: hd2

You can see the obvious problem here. Grub will not be able to boot from the fallback drive since hd1 no longer is the second RAID drive, but the backup drive. This of course defeats the purpose of a RAID 1 setup.

So how can I get grub to number the drives the same each and every time? Perhaps label the drives and have grub check the volume labels instead of the physical boot order?

I think you need to change your point of focus. Grub doesn't do the hardware detection. The BIOS has hardware information, but it isn't passed on the the OS. So, the hardware detection is done somewhere else.

I think you should start looking into udev rules and persistent device names.

I'm not sure I understand how that's going to help me. Grub expects something like "hd0", not device names (like "/dev/..."). Or am I wrong?

You have a "dev" entrance : root=
1)example1, pclos:
title linux
kernel (hd0,0)/boot/vmlinuz BOOT_IMAGE=linux root=/dev/hda1 acpi=on splash=silent vga=788
initrd (hd0,0)/boot/initrd.img
2)example2, ubuntu7.10 :
title Ubuntu 7.10, kernel 2.6.22-14-generic
root (hd1,0)
kernel /boot/vmlinuz-2.6.22-14-generic root=UUID=3334ac95-c30d-4156-b611-90f10b3d086c ro quiet splash
initrd /boot/initrd.img-2.6.22-14-generic
quiet
..........both examples from /boot/grub/menu.lst

so....good luck with finding your device-names.

In:

"/dev/hda1" is a kernel command-line parameter. Not a grub parameter. So changing it to a persistent device name with udev (in, for instance, "/dev/disk/by-name/raid1") won't help grub to find the right drive.

grub at mbr or at boot partition does not understand devices under /dev or udev or sysfs so forget it.

but zippy...its not clear to me...that you have realised you have 3 mbrs...each mbr should have a bootloader if what you are trying to do...is test the fail of a drive.

its easy to put a bootloader into each drive's mbr.....then change the bios boot order or remove cables to test that bios jumps to drive x mbr and then to grub and then to its booting files.

The third drive is a backup drive. I can put a bootloader on it if I want, but it doesn't hold anything to boot from.