Disk drive fired 'Frisbees of death' across data centre after storage admin crossed his wires

Re: "Not an uncommon story", you mean.

Unless you saw it yourself, and can give me the make & model of the drive in question. But you heard the story from someone else, who heard it from ... It's "a friend of a friend", all the way down, with no details. I was one of the guys who used to go out and trouble-shoot these stories (for DEC in my case). It always turned out that somebody got pissed off, ripped the drive out, and threw it at the wall/door/whatever.

Quite. Looking at those packs you'll see that they won't come apart and fling their platters around without a serious amount of energy being required. Way more than a linear drive actuator can deliver. Orders of magnitude more. One thing, the actual mass of the head carriage is pretty low. For obvious reasons. There's a hefty magnet as part of the positioner, but that one's stationary.

Yes, you'll ruin the pack, and quite likely the positioner as it rams into the end stop. But separating the platters from the spindle? Not bloody likely. And that would have to happen before the drive would start flinging the platters, frisbee-like, across the room.

One of the early DEC RK-series drives had a positioner that used a photocell and a striped glass plate as feedback. Unfortunately, the glue used to affix the glass to the positioner wasn't quite up to the task, and under heavy load (such as the drive exerciser) the glass had a tendency to become loose. Ergo, no position feedback, resulting in the positioner trying to join the heads with the spindle with great gusto.

Re: "Not an uncommon story", you mean.

Have been there, did not do that.

As a graduate student in EE, I managed to get a little independent work experience by hiring on as a replacement assembly line tech at the local DEC plant in Westfield MA (during the summers of 1976 and 1977). It was there that LA36 DECWriters, VT7x video terminals and RK06 disk drives were manufactured and tested.

I was hired as a final test tech on the RK06 (The RK06 was short lived, as the double density RK07 came out as we were ramping up the production line.) disk line. I was given a test procedure, schematics and a theory of operation manual, and told to test the drives and debug as required. I cannot stress enough how valuable this experience was, as I look back on it 40 years later. I got to work on a real design, see how real engineers did things, learn about manufacturability and testability, and get paid for doing it.

In any case, I was warned about the linear motor, the heads and the alignment pack. The RK06 had three or four "quad" modules, one of which was a linear board. It drove the linear motor, getting velocity and direction feedback from an optical slide and a couple of infrared sensors shining through it.

There was an emergency retract system, which applied battery power to the linear motor voice coil when feedback was lost. I think you can see where this is going.

While testing, there was a switch which could be used to disengage the voice coil drive. This was used to test the optical feedback sensors before closing the loop. There were other checks required of the wiring between the linear motor and the drive circuitry. Only after the feedback circuitry and the linear motor connections were verified good, was the switch to be closed.

I have personally seen, and heard, the results of miswired voice coil connectors. IN becomes OUT and vice versa. With speed too fast for human reflexes to react. It was common to rest one's hand on the top of the magnet housing, while using fingers to move the head carriage, while testing the feedback circuitry. We were warned multiple times to keep our fingers clear of the gap between the carriage and the housing. MOST of us managed to avoid getting our finger tips nipped by an unexpected emergency retract (feedback and wiring good, problem in the linear board). Some of us enabled the feedback circuitry without having adequately checked the connector. The result was usually heads driven into the alignment pack spindle, or nipped fingers.

As the platters were aluminium, and the spindle speed was a constant 3600 RPM, determined by power line frequency, there wasn't any possibility of the platters shattering (except, perhaps, by mechanical defect). The whole pack (2 platters, three data and one servo surface, 30MB?) was pretty solidly build, as I recall. But heads were routinely sacrificed.

Good times. I found my documentation last year, sitting in the bottom of an old file cabinet.

FAIL

the linear motor should have had a speed limiter and a cut out that operated as soon as the linear motor had rotated enough times (if it was geared) or beyond a certain angle (if it was direct drive) that the head would be off the disk or squashed up against the spindle.

A linear positioner in those disk drives is not something that rotates, and it's not geared either. It runs on rails or sliders, driven by a voice coil enclosed in a magnet. Not unlike a loudspeaker. The limits are stops on the rails of some sort, usually its mounting points.

And if the feedback is knackered, there's no way the controller can tell the positioner is moving too fast or past its range.