Every time a submission like this pops up I expect someone who worked at Seagate during that era to pipe up and explain what went wrong and why ... but I haven't seen it yet. Surprising.

I think it's sad that the hacker/maker community doesn't have enough knowledge/skill/inclination to disassemble this kind of thing and figure it out for themselves.

It's a bad sign when tech complexity gets out of the expertise of a hobbyist - I believe that's the start of a stagnation of that technology as new young engineers never get excited about disassembling and modding hard drives in their free time, and in turn it becomes paid engineers only who, I suspect, are less innovative.

When was the last time you heard someone who had modded their hard drive to, for example, seek faster?

The technical answer of what part failed and why is only a small part of the overall answer, though.

For example, with the Challenger explosion, it's O-rings vs. the culture and organizational deficits that led to: the O-rings, failure to revise the seal structure despite evidence of problems on prior launches, and the final decision to launch Challenger.

All of that is arguably the most interesting part.

Equipment cost is a huge problem there. You need a proper clean room if you're going to tear apart HDDs and expect them to work again when you reassemble them. Not to mention, the extremely precise tolerances involved would make mechanical mods a tall order indeed.

Firmware mods would be more doable, assuming it isn't locked down.

WDs tend to be a bit more durable... especially in smaller capacities:

http://www.brendangregg.com/blog/2021-05-09/poor-disk-perfor...

If you want to give it a go, I've still got a bunch of these drives lying around. We also have some still running in our zpool which I would gladly replace with something else, we're running out of space yet again... ;)

haha, it's even better if you replace in your comment hard drives with CPUs

But people do mod CPU's... By overclocking, by hacking into SMM mode, by liquid nitrogen cooling...

People also make their own toy CPU's for fun with quite some regularity - using FPGA's and other tools, the same as the pros.

And people decap and reverse engineer new CPU's like Apples M1.

And people do hack hard drives as well:

* install Linux to your hard drive firmware

https://spritesmods.com/?art=hddhack

The difference is, these hacks we've mentioned are software and solid-state electronics hack. On the other hand, doing a hard drive failure analysis in the mechanical realm is totally a different challenge, a hard drive is hard.

I think I missed something in my comment, because I focused more on processor fab

All I have seen is not having rotational vibration sensors.

https://www.tomshardware.com/news/seagate-hdd-failure-lawsui...

> The Seagate 3 TB models failed at a higher rate than other drives during the Backblaze deployment, but in fairness, the Seagate drives were the only models that did not feature RV (Rotational Vibration) sensors that counteract excessive vibration in heavy usage models -- specifically because Seagate did not design the drives for that use case.

Do drives in this type of stationary assembly (i.e. a server rack or a stationary PC) ever trigger their vibration protection? Barring of course during a bit of bumping in the server rack. I mean, these drives are not sitting in laptops. I have a hard time thinking the lack of this feature bears any weight on the huge failure rates Backblaze and discrete customers saw, though I'd accept the argument if there was a 2.5" laptop model of that drive seeing the same failure rate.

As I understand it another driver in the same rack doing work is enough to effect the performance of the other drivers in the rack. This is why you use fancy expensive "enterprise grade" drives as they can cope better with it.

https://www.dtc.umn.edu/publications/reports/2005_08.pdf

> In the presence of an external vibration caused by adjacent disk drive seek operations the bandwidth performance of a consumer-grade disk drive “feeling” these vibrations will decrease about 10%-15% when reading data and about 25%-40% when writing data

Don't really know if this is enough to actually cause damage if the disk does not have the hardware to detect this vibration.

There's also this great example of Brendan Gregg screaming at hard disks to demonstrate exactly that sort of latency spike: https://www.youtube.com/watch?v=tDacjrSCeq4

I don't know if the latency spike is caused by sensors detecting the extra vibrations and delaying a read or seek operation or if its just vibrations causing minor read-head errors which either require a 2nd pass (that isn't reported through to the OS via SMART) or some sort of extra processing to pull the correct signal out of the raised noisefloor.

Edit: person shouting at the disks is Brendan Gregg, not Bryan Cantrill.

Apparently, yeah. Here's the classic video on the topic: https://www.youtube.com/watch?v=tDacjrSCeq4

Is there any way to tap this hdd sensors data on linux? I remember there were some apps on windows that read these hdd sensors data and use them as accelerometer / gyro, which was pretty neat when run on a laptop (this was when iBeer and virtual lighter apps were still a hot thing).

A search through data recovery forums like hddguru.com (they tend to have a secretive "trade union" attitude, but some useful stuff leaks through...) and the Internet reveals some more info like this:

https://www.recoveryforce.ca/forums/viewtopic.php?f=12&t=157

Most of them seem to be head problems, and there are some outright head crashes too. Seems like bad media (leading to head failure) is mainly to blame:

https://www.cheadledatarecovery.co.uk/2016/02/seagate-dm001-...

It sounds vaguely reminiscent of the IBM DeathStars, which failed in a similar fashion.

Huh. So many fixes in there are listed as 'replace head(s)'. I didn't realize that was a common fix for hard drives, thinking it would require near cleanroom conditions to ensure no dust got into the drive as its being worked on, and that it would be much cheaper/faster to maintain backups and swap in a new drive.

It's a "fix" in the context of data recovery business. The aim is to get the drive in a working condition for a few hours, just enough to get an image of it. After that the drive is trashed (or more precisely, put in a big box so it can be used as a donor for other drivers in future)

I've read that the replacement heads also often get damaged while reading the media during the recovery process.