Can an RAID card destroy an SAS disk?

[Mattias Hedman]

I have an Dell Poweredge R720 with a Dell H710P mini (D1) RAID card.
I wanted to run zfs so I found a way to flash the RAID card, https://fohdeesha.com/docs/perc/, first time around I missed the part where I would write back the SAS address to the card. Everything else worked just fine with the flashing.
So I bought 10 900 GB Seagate SAS disks, put them into the server, tried to make a zpool of them, failed. Error 1.
Went to CLI ran fdisk /dev/sdX on every disk, 9 out of 10 gave me the same error: Input/output error while writing out and closing file system.
Bewildered I moved them to my older HP server, with an intact unflashed RAID card, tried them there, 6 out of 8, LED lights went orange. In the BIOS I was only able to see 2 disks.
At this time, I contacted the seller, told him the story. He says to me that the disks was destroyed when I tried to make the zpool with a, his words, faulty raid card.
After this a lot of time went into to narrow down the error. Reflashing the card, both with IT/HBA firmware and writing back the Dell firmware. I also tried to write zeros to all the disks. All in all in the end I had 4 out of 10 disks working (A tid bit, before sending me the disks he tried 4 disks and they were all working...).
At this time I asked the seller to give me the money back and I would send him the disks.
He refuses, blaming my handling of the disks to be the fault here.

I just need to get a reality check here.
Can a RAID card, flashed or not, destroy a disk? Rendering it semi-working?

 

[no_connection]

Did you ask Seagate if it's possible?

 

[Mattias Hedman]

Did you ask Seagate if it's possible?

No I havn't. It's not new disks bougth via ebay.

 

[no_connection]

If you ask Seagate and they say drives can't be bricked then you can tell that to the seller. They made the drives so they should know.

 

[jgreco]

At this time, I contacted the seller, told him the story. He says to me that the disks was destroyed when I tried to make the zpool with a, his words, faulty raid card.

I just need to get a reality check here.
Can a RAID card, flashed or not, destroy a disk? Rendering it semi-working?

In what way was the RAID card faulty? Failing to flash an address just means you're using all zeroes. The reason you don't want to do this is because SAS initiators and targets each need a unique address to communicate, and if two devices with the same address show up, that's a problem. So two initiators both with all-zero addresses would conflict with each other and confuse things. A single initiator with all zeroes isn't an issue.

I can tell you that for absolute certain there have been some instances here in the shop where a RAID card's SAS address didn't get restored when we were batch-flashing cards, and it has never destroyed a disk. I am not willing to call it impossible, but it seems incredibly unlikely. There isn't a mechanism that I can think of. Seagate didn't code a special check in for an SAS address of all-zeroes that would cause it to zorch the drive. Drives are designed to be extremely resilient against noise and comms errors on the SAS bus, including being able to robustly handle garbage sent by an initiator, because tomorrow's cool new feature codes won't be understood by today's drives, so it *has* to be able to cope.

Bought 'em on eBay? Not judging or anything, but just a word to the wise, never tell a seller anything they might be able to use against you.

 

[Newfoundland.Republic]

I wonder if the drives are have custom firmware (e.g., EMC)? I never used EMC drives but could this be an issue?

 

[Mattias Hedman]

I wonder if the drives are have custom firmware (e.g., EMC)? I never used EMC drives but could this be an issue?

How do I check that? If that is the case?

 

[Newfoundland.Republic]

Good question. I do not have any experience there as all my drives are "standard" new drives. What is the model number on the drives? Is there any reference to EMC, HP, etc. on the drives?

 

[HoneyBadger]

I'll make the wager that the disks are from a piece of enterprise storage hardware, and are using 520-byte sectors.

Any kind of model number or "part number" will help us tell that (or a photo of the top label, with the S/N blocked if you're concerned)

 

[Newfoundland.Republic]

I'll make the wager that the disks are from a piece of enterprise storage hardware, and are using 520-byte sectors.

Any kind of model number or "part number" will help us tell that (or a photo of the top label, with the S/N blocked if you're concerned)

If they are enterprise 520-byte sectors instead of 512-byte, Reformatting NetApp Disks from 520 sector size to 512 sector size using FreeNAS (FreeBSD) might work?

 

[Mattias Hedman]

The seller says they are 512. Model nr is ST900MM0006

 

[Newfoundland.Republic]

As @HoneyBadger noted - can you take a picture and post it?

 

[Mattias Hedman]

Here you go.

 

[HoneyBadger]

I don't see any vendor part number other than the Seagate, guess I'm off the mark on the 520bps.

But I'm not buying the seller's "explanation" of the drives being damaged by being attached to a bad controller. A bad power supply possibly, but not a controller.

Are you able to read SMART stats from the drives? smartctl -a /dev/daX

 

[Newfoundland.Republic]

Hmmmm... What does
camcontrol devlist
or
geom disk list
provide?

 

[c77dk]

According to the datecode the drive is from december 2014 - so good chance it was DOA when you got it :-/

 

[Mattias Hedman]

According to the datecode the drive is from december 2014 - so good chance it was DOA when you got it :-/

That is what I have tried to tell the seller, but now he says he will give me 10 new disks this time HGST. Just wait and see. If it was not for this COvid business I would have visited him.

 

[Mattias Hedman]

Thanks for all the help!
The seller has agreed to send me new disks, HGST this time, to replace these Seagates.
So now I packed them and will send them back as soon as the new ones arrive.