and suddenly, the microdrive suddenly died

[DiTBho]

HDDs don't usually die like SSDs, the miniaturized ones should be no exception, and this one was an Hitachi 4GB first gen.

So, nothing special.
Nothing too weird ...

... nevertheless ...

It was just performing a low disk-check pass
      badblock -swv -p1 /dev/sdc
it had only reported two bad sectors, when suddenly the micro-drive ... died.

I checked with my fingers, it was not hot, not warm, and it didn't bad smoke.

no-panic.

I shut down the computer, restarted, no dice. The Linux Kernel probes the SCSI-attached device, reports it as "sdc", but it's not accessible
      fdisk /dev/sdc
reports nothing but "cannot access /dev/sdc"
      hexedit /dev/sdc
reports nothing but "cannot access /dev/sdc"

no-panic, but w_t_f_? 


(RIP, it was a good disk)

[SiliconWizard]

How old was it?

[DiTBho]

How old was it?

used for 2 years, but that's not the point.
The Microdrive is a tiny 1" mechanical Hard Disk Drive and compatible with the IDE standard, and I've never seen a hard drive die like this way.

I mean, usually they manifested some IO errors, then notified by bad clusters, sometimes corrected by relocating the bad sectors, but what happened in this case? The electronics are not fried because the kernel correctly probes the drive, but no longer responds to any disk-seek request.

[AndyBeez]

I remember the Microdrive as being vaunted as the rich man's Compact Flash card - especially in the pro-photographer market.
But didn't these go extinct with the advent of the SD memory card? Try tapping it gently on its edge. The arm might be stuck shut 

Otherwise blame the god of failed capacitors and do an autopsy - post pics here  RIP

[magic]

dmesg, smartctl

[DiTBho]

But didn't these go extinct with the advent of the SD memory card?

some routers use it as rootfs, and don't accept CF.flash.

Try tapping it gently on its edge. The arm might be stuck shut 

thanks, good idea!

failed capacitors and do an autopsy

I am willing to build an ESR tester, and use it to test caps 

[DiTBho]

there are two big smd caps.


Interesting article, about smartools and storage devices check health, see  here 

[magic]

Try tapping it gently on its edge. The arm might be stuck shut 

thanks, good idea!

A dumb idea if you don't even know what's wrong with the disk.
Does it spin up? Can you hear it load the heads?

BTW, I once tried that tapping trick, quite ungently even, and it didn't do shit.
I still had to take the disk apart to unstick the heads manually.

And again, every risky procedure is dumb if you aren't sure that it's necessary.

[DiTBho]

When the microdrive died I shut down the computer and restarted, but I have a tool that saves the dmesg buffer when it's full, so I can still pull the last few error messages before the microdrive went haywire.

This is what the kernel said.

Code: [Select]

blk_update_request: I/O error, dev sdc, sector 2761152
sd 8:0:0:0: [sdc] tag#0 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
sd 8:0:0:0: [sdc] tag#0 Sense Key : Unit Attention [current]
sd 8:0:0:0: [sdc] tag#0 Add. Sense: Not ready to ready change, medium may have changed
sd 8:0:0:0: [sdc] tag#0 CDB: Write(10) 2a 00 00 29 b5 a0 00 00 f0 00

Buffer I/O error on dev sdc, logical block 341684, lost async page write
Buffer I/O error on dev sdc, logical block 341685, lost async page write
Buffer I/O error on dev sdc, logical block 341686, lost async page write
Buffer I/O error on dev sdc, logical block 341687, lost async page write
Buffer I/O error on dev sdc, logical block 341688, lost async page write
Buffer I/O error on dev sdc, logical block 341689, lost async page write
Buffer I/O error on dev sdc, logical block 341690, lost async page write
Buffer I/O error on dev sdc, logical block 341691, lost async page write
Buffer I/O error on dev sdc, logical block 341692, lost async page write
Buffer I/O error on dev sdc, logical block 341693, lost async page write


[AndyBeez]

I've probably read as much around that async error message as you have and other than people reporting, "it's a serious error", there seems to be nothing more enlightening than the message is generated when a SCSI device times out. Sounds like a fatal hardware error which may have been symptomatic for a while. Tapping or shaking gently might help. But first check connectors, ribbon cables, etc, etc. Otherwise, it's now a very small paperweight

[ejeffrey]

It was just performing a low disk-check pass
      badblock -swv -p1 /dev/sdc
it had only reported two bad sectors, when suddenly the micro-drive ... died.

If there were two dead sectors already, the death was not sudden.  But I'm not sure where you get the idea that mechanical hard drives don't "just die" -- they absolutely do.

[Whales]

Is it spinning?  Can you hear/feel it?  Sudden death and timing out on reads might be because the drive isn't spinning.

[DiTBho]

If there were two dead sectors already, the death was not sudden. 

makes no sense.

With HDDs you have some dead-sectors and the disk still in working state, correctly probed by the kernel, and still accessible.

Or do you assume that died sectors means dead motor? dead I/O heads?

But I'm not sure where you get the idea that mechanical hard drives don't "just die" -- they absolutely do.

I have seen something like a thousand HDDs, from 1992 to 2023, they just die, but never seen that way.

Usually, I have seen the bad-block list increasing to a point where it was no longer convenient to use the disk, which was still accessible anyway.

To prepare a software I wrote to test disks (like badblocks), I opened a Conner HDD, and voluntary damaged the upper plate with a pin. And you guess what? Still working, still accessible, just with hundred bad blocks.

I mean
            hexedit /dev/sdc
doesn't return any error!

Whereas with this microdrive, as soon as the badblocks tool hit the console with the bad news of two dead sectors, the drive immediately crashed and has been like ZERO blocks ever since.

I mean
            hexedit /dev/sdc
returns an error as it cannot access the device.

[AndyBeez]

Maybe the controller chip went futt?

[DiTBho]

Maybe the controller chip went futt?

yup, it seems so 

[DiTBho]

Is it spinning?  Can you hear/feel it?  Sudden death and timing out on reads might be because the drive isn't spinning.

While answering your question for { 2.5", 3.5", 5.25" } is an easy task as you hear the motor spinning and feel the vibrations it produces - especially with my SCSI 12000-RPM Raptors that vibrate like the washing machine going into the spin cycle - my tiny 3600-RPM micro-drive makes no sound and does not vibrate.

it doesn't make any sound and doesn't make any vibration because ... well, because it's dead?

Well not exactly ...

... I have in total 21 working micro-drives by { Hitachi, Seagate }, although with a lot of difficulty you can hear at least the Hitachi' spinning, especially the first generation, with the Seagate's ones is an hardly task as well as with the Hitachi's 4GB, especially with the second (4GB, 6GB) and third generation (6GB, 8GB).

So it was so hardly to say ... that to answer your question I bought a clip-microphone, which I am sure I will make good use.

The sound of a working Hitachi 4GB (first gen) micro drive:
- for the first second it seems that the drive motor is trying to spin
- then hear nothing but noise, so you assume it's the sound of a successful attempt
- just to be sure, run something that continuously reads disk sectors, and hear the heads clicking

The sound of the dead Hitachi 4GB (first gen) micro drive
- it sounds like the disk motor is trying to spin
- but after six seconds it sounds like five failed attempts
- and since then it sounds like the silence of death


so ... it is the motor? the ESR of a bad capacitor on the controller? plates? heads? dunno 

[DiTBho]

meanwhile, I just bought 20 Hitachi 4GB first gen NOS micro-drives.

NOS = new old stock, practically it sounds like abandoned in an abandoned warehouse, still sealed in their anti-static bags, but probably they have been mistreated when the company went bell up.

Then recently found by some computer-drug dealer of those you can find on eBay, the ones who must have found a stack of micro-drives because he was paid to clean an industrial area and thought it best to list them as "NOS" to make more money.

Those who know what they sell, also ask 200 USD for a pair of micro-drives, whereas I paid 50 USD for the whole stock.

How many will work? Let's check it out 

[magic]

A normally spinning HDD produces something similar to white noise and a faint 120Hz (7200), 90Hz (5400) or 60Hz (3600) tone. The latter can often only be heard by pressing your ear against the disk.

During startup, the motor may make a buzzing or beeping noise before the normal spinning noise appears. After reaching operational RPM the disk makes a few clicks as it unparks the heads and reads some internal data off the platter.


If you can't hear normal spinning noise and clicks, while beeping and buzzing repeats several times for longer than usual, the spindle and platters are probably stuck mechanically. This could be due to seized bearing or the heads being crashed onto platter surfaces and stuck to them.

Head crash usually damages the affected platter area and generates debris which may affect reliability. Seized bearing is a seized bearing. Either way, the disk is probably a writeoff if it can't spin up.

If you have data to recover, try the tapping trick or look up videos how to disassemble a disk and fix stuck heads. This means forcibly rotating the spindle with a screwdriver (usually torx) and manually shoving the head arm to the side. I have done it once and recovered everything except a few thousand sectors (less than 1MB total). YMMV.

[DiTBho]

A normally spinning HDD produces something similar to white noise and a faint 120Hz (7200), 90Hz (5400) or 60Hz (3600) tone. The latter can often only be heard by pressing your ear against the disk.

During startup, the motor may make a buzzing or beeping noise before the normal spinning noise appears. After reaching operational RPM the disk makes a few clicks as it unparks the heads and reads some internal data off the platter.


If you can't hear normal spinning noise and clicks, while beeping and buzzing repeats several times for longer than usual, the spindle and platters are probably stuck mechanically. This could be due to seized bearing or the heads being crashed onto platter surfaces and stuck to them.

Head crash usually damages the affected platter area and generates debris which may affect reliability. Seized bearing is a seized bearing. Either way, the disk is probably a writeoff if it can't spin up.

If you have data to recover, try the tapping trick or look up videos how to disassemble a disk and fix stuck heads. This means forcibly rotating the spindle with a screwdriver (usually torx) and manually shoving the head arm to the side. I have done it once and recovered everything except a few thousand sectors (less than 1MB total). YMMV.

with the clip microphone pressing against the disk I can hear five beeping and buzzing repeats for 5 times before the kernel driver issuing some garbled messages on serial console.

[ejeffrey]

If there were two dead sectors already, the death was not sudden. 

makes no sense.

With HDDs you have some dead-sectors and the disk still in working state, correctly probed by the kernel, and still accessible.

Or do you assume that died sectors means dead motor? dead I/O heads?

I mean  by the time you see dead sectors the drive is already heavily damaged.  All modern drives (and I mean since 1990) have over provisioning and relocation tables for bad sectors.  If you see a persistently bad sector: not just a read error, but an error that can't be cleared by writing clean data, that means either there are enough bad sectors to exceed the over provisioning or that something else is wrong such as a mechanical failure or a problem with the drive controller.  In any case such a drive is already on its last legs

The correct way to determine health status of a (S)ATA hard drive is the SMART status.  This will tell you things like the relocated sector count before exceeding the drives over provisioning limit.  A drive that has a gradually increasing relocated sector count should be removed from service as soon as practical.

Badblocks was only ever really an important tool for MFM drives and other early hard drives that didn't have an integrated controller.

[DiTBho]

Code: [Select]

[1407966.883667] blk_update_request: I/O error, dev sdc, sector 17048
[1407966.883774] Buffer I/O error on dev sdc, logical block 2131, async page read
[1407966.888240] sd 7:0:0:0: [sdc] tag#0 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
[1407966.888361] sd 7:0:0:0: [sdc] tag#0 Sense Key : Not Ready [current]
[1407966.888470] sd 7:0:0:0: [sdc] tag#0 Add. Sense: Medium not present
[1407966.888578] sd 7:0:0:0: [sdc] tag#0 CDB: Read(10) 28 00 00 00 42 98 00 00 08 00
[1407966.888687] blk_update_request: I/O error, dev sdc, sector 17048
[1407966.888794] Buffer I/O error on dev sdc, logical block 2131, async page read
[1407966.892280] sdc: detected capacity change from 4095737856 to 0


Code: [Select]

[1407966.892280] sdc: detected capacity change from 4095737856 to 0
macmini-intel # badblocks -swv -p1 /dev/sdc
badblocks: No medium found while trying to determine device size

W_T_F_?_!_?_!_? 

[DiTBho]

I bought another lot of qty=8 disks, 4 are fully functional, 4 are...defective.
but 2 of the ones in the defective-group are weird

• they successfully pass the first badblock 0xaa write and comparison test
• they get hot, not too hot, but hotter than others in the "perfectly-working"-group
• they fail the second badblock 0x55 write and compare test
• they go nut to the point the kernel says "detected capacity change from 4095737856 to 0" and since then "No medium found while trying to determine device size"

perplexed, what's happening here?

[SiliconWizard]

What makes me perplex is what would make you think that the drives would be functional to begin with?
This is pretty old tech and probably pulled out of gear that has seen better days, I'm perplexed that you are perplexed.

[DiTBho]

What makes me perplex is what would make you think that the drives would be functional to begin with?
This is pretty old tech and probably pulled out of gear that has seen better days, I'm perplexed that you are perplexed.

from a stock of 24 total disks, I have now 17 dead bodies on the vodka case, which the seller declared NOS parts.
(probably lies  )

worse still routers like rb532 accept nothing but the old Hitachi 4GB 1st gen and Mikrotik refuses to tell me why their buggy firmware behaves like this

anyway, s.m.a.r.t. is not supported, I cannot say how many hours they worked
what I can do is stress them and see how they behave