Well, it was complaining about being unable to find, yes. It couldn't find the source file. In fact, the entire drive had disappeared. The system event log was full of ATA errors, and there was an ominous clicking. I shut down the machine, figured I'd deal with it in the morning. I hoped that power-cycling it might reset the drive and get it working enough that I could copy files off. As a lay in bed thinking about what was on that drive — which was actually a striped array of three physical disks — I realized that I could replace virtually all of it... except for a couple things which I had no copy of anywhere else nor source media for anymore.
When I powered the machine up this morning, click click click click... The BIOS found the drive and declared it failed. Windows just called it "missing". Tried again, and this time the BIOS said there was no secondary master disk. Third time, not lucky. Still no secondary master disk. I left the machine off while I brooded over my options, then tried a fourth time, which got it recognized but it was still failed and "missing".
So I went up to Fry's and bought a disk large enough to replace the entire array. Since the motherboard has SATA on it, I bought a SATA drive, my first one ever. Because it was my first one ever, I did not know until I got it home that SATA drives do not have a standard molex power connector.
Allow me to digress a moment: why, oh why, have these drives abandoned the standard molex power connector which we have known and loved since 1981 and probably even earlier? The pins on the new connector are different lengths, which is a technique used for hot-pluggability, but that makes no sense. If you're going to be hot-swapping disks, they're going to be in caddies of some sort, and the connectors on the caddy will handle the specialized requirements for hot-pluggability. The only other reason I can think of to redesign the connector is to allow the disk to communicate with the power supply, so the power supply can warn the disk that the power is about to go down and it should flush its buffers. Of course, the operating system should have done that already, but if you're running Windows the OS won't always be able to do it and maybe having it in hardware makes some sense. Is hardware design now being dictated by Windows' shortcomings?
Anyway, I needed an adapter. Instead of going to Fry's, I tried the little local computer store in the strip mall a few blocks away. They were closed. I consulted the little Yellow Pages book that I keep in my car and located another store which was also closer than Fry's. They were closed too. I can understand that, since they're in a light-industrial park and probably wouldn't get many customers on a weekend. The first one is in a strip mall, though. There were lots of people around, but still, they were closed.
Oh well. I went up to Fry's. There was an obvious place where such an adapter should have been, in the cable display next to the hard drive shelves. Since I couldn't find them myself, I asked, and no fewer than three people led me to that same cable display while saying, "Oh yeah, we have those. They're right over... huh. Well, this is where they should be." So I ended up buying a new power supply, one that came with SATA power connectors.
Everything worked fine once I installed all the new gear. Now I just have to restore or recreate the files.
Postscript the first: the new power supply is black and had that wonderful curing-enamel smell. Mmmm.
Postscript the second: one of the things that has been bugging me about the new fileserver is that it doesn't have as many PCI slots as I'd like. I really wanted three PCI cards (FDDI, USB2, and UDMA IDE) but it only had two slots. Since the BIOS is perfectly happy with a 200-gigabyte drive connected to the onboard IDE (a bit surprising for a ten-year-old P133 — I've had much newer machines which would lock up hard during the POST upon seeing anything larger than 32 gigs), I was able to do without the IDE card, but that means these very fast drives are running at PIO4. Theoretically even PIO4 is faster than the achievable speeds over the network, but that's very theoretical. Actual disk performance is usually much lower than the interface's maximum transfer speed.
Do you remember the old multi-IO cards from the early days before the standard controllers moved onto the motherboard? They had IDE, and serial, and parallel, and game ports, and maybe even a floppy controller. I began to wonder if such a thing still existed. In particular, I wondered if there were such a thing as a card with IDE and USB2 on it, killing both my birds with one slot.
Well, yes, there are. SIIG, makers of weird and wonderful interface cards such as the four-port serial card which I have, actually makes a couple different models! You can get the one with two IDE chains, USB2, and firewire, or you can get the model with one IDE chain, SATA, USB2, and firewire. Score! I ordered the simpler one since I don't expect to need SATA in the fileserver — I can't replace it's special-formfactor power supply! And the cool thing is that this multi-IO nouveau card actually has more USB2 ports than the USB2-only card which I'm using now. Beyond that, it also has the firewire ports, which should work OK for attaching storage as well.