<SOLVED> Encrypted drive partition table disappears?

[Glycerine]

The environment:

UNRAID 6.8.3

Data Drives: 12TB shucked WD EMAZs EDIT: All purchased from various stores 4 & 5 months ago.

Supermicro X10 board

m1015 LSI-flashed HBA card connecting data drives, parity/cache on mobo SATA

Array drives BTRFS encrypted

 

The issue:

Powered down to do some swapping around of gear above the Unraid machine in the rack. After booting the box, two drives simultaneously went missing (single parity drive  ). Poked around dmesg and the drive logs and see the below:

 

Quote

Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 UNKNOWN(0x2003) Result: hostbyte=0x00 driverbyte=0x08
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 Sense Key : 0x5 [current] 
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 ASC=0x74 ASCQ=0x79 
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 CDB: opcode=0x88 88 00 00 00 00 00 00 00 00 00 00 00 00 08 00 00
Nov 18 10:03:42 <hostname> kernel: print_req_error: I/O error, dev sdk, sector 0
Nov 18 10:03:42 <hostname> kernel: Buffer I/O error on dev sdk, logical block 0, async page read
Nov 18 10:03:42 <hostname> emhttpd: error: ckmbr, 2030: Input/output error (5): read: /dev/sdk
Nov 18 10:03:42 <hostname> emhttpd: ckmbr error: -1

 

Drives appear in UD wanting to be formatted.

Attempting to mount via command line while array isn't started (because it can't  ), cryptsetup returns:

Quote

WARNING: Locking directory /run/cryptsetup is missing!

Device /dev/sdd is not a valid LUKS device.

 

Fdisk -l return Input/Output error.   Browsing /mnt/disk/by-id/ shows all the "working" drives and their partition e.g. sda, sda1, etc but just the drive and not partition for the two affected drives.

 

When looking at disk info in the GUI, the partition format shows "error" for the two drives, instead of GPT-4k.

 

I have tried placing the affected drives on different spots in the backplane and on the mobo SATA ports to the same results. All cables are secure. EDIT: No SMART errors shown.

 

So with allow the above, it sounds like my partition table is toast on both drives. I've got a couple of drives arriving today so I can dd the drives before attempting to fix, but I'm not sure where to start.  Is there anyway I can recreate the partition data on the drives without blowing away the information? Is something else going on that I'm not seeing?

 

Greatly appreciate any help because I'm getting out of my depth quickly!  

 

 

 

 

wolfwood-diagnostics-20201118-1029.zip

Edited November 18, 2020 by Glycerine
Solved

[JorgeB]

55 minutes ago, Glycerine said:

Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 UNKNOWN(0x2003) Result: hostbyte=0x00 driverbyte=0x08
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 Sense Key : 0x5 [current] 
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 ASC=0x74 ASCQ=0x79 
Nov 18 10:03:42 <hostname> kernel: sd 7:0:7:0: [sdk] tag#1537 CDB: opcode=0x88 88 00 00 00 00 00 00 00 00 00 00 00 00 08 00 00

These errors on both drives suggest it's not just a partition problem, comparing SMART from both affected drives to the others of the same model there is this difference:

 

ATA Security is:  Disabled, NOT FROZEN [SEC1]
Wt Cache Reorder: Enabled

 

ATA Security is:  ENABLED, PW level HIGH, **LOCKED** [SEC4]
Wt Cache Reorder: Unknown (SCT not supported if ATA Security is LOCKED)

 

So first thing would be to unlock them.

 

 

[Glycerine]

Aha! That rings some bells. Thank you! Was playing with the secure-erase option when setting up these drives before I gave up doing so. These two must have been the first in that list.

 

Little googling shows I need to run through some hdparm options and pray I remember the passphrase from 6 months ago.  I'll play around and see if that gets me to a workable drive and update topic if solved.

[Glycerine]

Marking this solved as it makes the most sense. Still running through potential passwords to confirm but I believe the locked drive from the secure erase previously is the issue.