Samsung PM983 dropped out of ESXi 7

[986box]

Picked up Samsung PM983 recently and used it for a few months. It dropped out of ESXi 7.0 yesterday without warning.
CrystalDisk reported good health as before but space reported is off. It should be 960GB.

What can caused it to drop off?

 

[oneplane]

It's health is not Good, but CrystialDisk is not smart enough to figure it out. Your SSD is dead and running of the fallback bootrom, or using an unconfigured bootrom because it can't load the device-specific configuration.

As for if it can be repaired or revived, I don't know. What your'e seeing on your screen is basically an SSD that says: "my CPU is working, but that's about it".

 

[986box]

@oneplane Is there a tool to check what is wrong?

Does Magician works? Magician

 

[oneplane]

None of the consumer tools will help you here. Either drive firmware recovery tools or manufacturer tools, neither of which are publicly available.

 

[111alan]

It has entered ROM mode(1GB capacity was shown). Basically it means that the controller can't detect the firmware inside the NAND.

 

[986box]

What would cause the problem? overheating? Trash the drive? I have another on the way.

 

[oneplane]

Can be anything. Bad NAND, bad DRAM, bad controller, bad PCB, bad firmware. It's impossible to tell since you can no longer communicate with anything in the drive except the bootrom and that won't tell you anything. This requires physical inspection.

 

[gb00s]

None of the consumer tools will help you here. Either drive firmware recovery tools or manufacturer tools, neither of which are publicly available.

That's not quite accurate >> PC 3000 with all its variants, as one of them, is available from here

EDIT: Added price list if someone is interested

 

[oneplane]

That's not quite accurate >> PC 3000 with all its variants, as one of them, is available from here

I suppose the commercial data and drive recovery tools might have some options, but even Flash Extractor, Arvika's database, Atoll, MRTLab, Rusolut, DFL etc (and probably a bunch more that used to be mechanical only, like SalvationData?) all need a flash map, firmware modules and either chip off board adapter or special interface card to do anything.

I did wonder if maybe the drive has a simple serial interface (most do) where some more details as to what is exactly broken can be gathered (i.e. unable to even talk to the NAND vs. talking to the NAND but some data being corrupted). Maybe it's something dumb like a memory overflow in a SMART table. Or maybe the bootrom can't find the main firmware and stops loading and it can be fixed by simply re-flashing the common firmware, leaving the device specific configuration and runtime data as-is. But it's all guessing at this stage And I don't think 986box has a serial interface and logic probe or logic analyser on hand to find communication pins would be a fun exercise tho.

 

[986box]

And I don't think 986box has a serial interface and logic probe or logic analyser on hand to find communication pinswould be a fun exercise tho.

Yeah, I do not have want to use whatever resources needed to figure this out. Unless it’s a quick 10 mins and free s/w.

in addition, I don’t have important VMs on this storage. So the loss was not a big deal.

luckily,the rest are on the consumer Samsung 970 I bought off here a few years ago.

 

[oneplane]

Yeah, I do not have want to use whatever resources needed to figure this out. Unless it’s a quick 10 mins and free s/w.

in addition, I don’t have important VMs on this storage. So the loss was not a big deal.

luckily,the rest are on the consumer Samsung 970 I bought off here a few years ago.

In that case: trash it.

 

[Stephan]

I commend Samsung for teaching alot of people all over the world lately about the value of having a working backup solution. /s

 

[111alan]

I commend Samsung for teaching alot of people all over the world lately about the value of having a working backup solution. /s

But Samsung is still one of the brands with the lowest fail rates.

It's just the whole nand industry is getting worse due to cell shrink and disturbance. No one is actually free of issues.

 

[111alan]

Yeah, I do not have want to use whatever resources needed to figure this out. Unless it’s a quick 10 mins and free s/w.

in addition, I don’t have important VMs on this storage. So the loss was not a big deal.

luckily,the rest are on the consumer Samsung 970 I bought off here a few years ago.

If you still want the drive, try security erasing the drive as is using nvmecli. There is a chance that the software fault may be cleared and the drive may come back to life.

Then you can doe some full drive reads and writes to see if you can trust it with data.

 

[Stephan]

@111alan Always interesting for me to read your insights into the industry! Glad you are contributing.

 

[986box]

I do have backup running. The VMs on the drive were inactive anyway.
Had pulled the drive off the server to check on a different machine. Will give the nvmecli when I have some time.

Had been looking at picking up a few SSDs to add. May pickup extra one for datastore.

Original idea for the post is understand what causes the failure. So not to repeat again.