Now that the old router/NAT/fileserver is down and may be freely gutted, it was time to get the most critical disks out: the mirrored pair. I had to boot it up quickly to find out which two they were; I have hardware records for all my machines, but they're not so detailed as to include which drive letters map to what hardware.
The target for these drives was my main machine, which is an old P4 with a tube-audio motherboard. (Seriously. I bought it because I figured it would be a flash in the pan and I wanted to be able to prove that this monument to engineering overkill had existed. As it turns out, it's been fairly successful and AOpen has introduced additional models!) At the same time, I wanted to put in a spare 20G drive which will be the OS drive when I get around to installing XP on my main machine — my bacon has been saved more than once by having the OS on a separate spindle from my data, and I have been unhappy about having them sharing a disk, even though they're on separate partitions.
The case had plenty of room for all four disks (the mirrored pair, the existing disk, and the to-be-the-OS-disk), but I ran short on IDE space: in addition to hard drives, this machine has a CD-RW and an IDE-based PCMCIA reader. That's six devices versus four available positions, so I had to move over one of the old fileserver's outboard (on a PCI card) IDE controllers. Doing so allowed me to put each half of the mirror on a completely different controller: one attached to the motherboard, one attached to the outboard.
All these disks raised the specter of insufficient power and cooling, so I moved the fileserver's 650W power supply over, and I also took the three case fans from fileserver and put them in. The main machine's case has spaces for a LOT of fans: there are four fan spaces in front, two in back below the power supply, and there's a fan on the side as well. After moving those three, there were still three un-fanned spaces. While searching my closet for power-Y-cables, I found another case fan and stuck that in too.
The next little bit of fun was that I wanted to be able to install XP without having to open the case again, so I made the 20G drive the primary master, and the existing bootable drive the primary slave. Fortunately the BIOS is flexible enough that it can be set to boot from any IDE drive attached to the motherboard, but it wasn't clear how Windows 2000 would react. Well, in a rare display of good sense, Windows declared that the drive it booted from was still C:, regardless of its new position in the hardware scheme of things. Then it assigned letters to other partitions on the same drive, and it finished by assigning letters to other drives starting at the top. The other partitions and drives didn't worry me, since Windows allows you to set all of that manually, but if it thinks that the OS drive has changed letters, or even changed serial numbers, it freaks out. I've had to do remote registry hacking before in order to recover from that situation, and it wasn't fun. Thankfully, it wasn't necessary this time. In fact, the entire procedure went with great smoothness.
I attribute the smoothness to the blood. A sacrifice of blood always increases the odds of success, and I donated some this time. I've had dry patches on my forefingers for a while, for reasons unknown. Well, Monday night I managed to crack both of them badly enough to bleed. Icky, yet effective.
Other news: I have a dual FDDI ring. The machines which I expect to be powered down all have bypass switches, but the fileserver and my main machine don't fall into that category. With both of them down at the same time, the ring is distinctly partitioned. I spent a lot of time tracing the fibers which formerly lead to the fileserver so that I could patch around them. I really need to redo my ring so that it makes some kind of intuitive sense, in the same way that my machines' positions on the KVM switches map to their spatial positions on the shelves. Let me write that down so I don't forget...