Pre-Read fails on SAS drive. No SMART errors. Read errors when in array

[Shinobu]

Hi,

 

So I have an issue with a 12TB SAS drive I have. I've previously had this in the array, after a pre-clear, but it started giving occassional read errors (about 2 at a time) and eventually the disk was marked as disabled by unRAID. SMART tests all come back showing no errors, though SMART does seem to not quite function correctly with SAS drives in unRAID.

 

The drive is close to brand new and a pre-clear will succeed, but a pre/post read will fail, so if you skip the pre/post reads it will complete.

 

Looking at logs, there are repeated logs for this disk alone (sdg):

kernel: Buffer I/O error on dev sdg, logical block 1387229646, async page read

kernel: blk_update_request: I/O error, dev sdg, sector 11098420608 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 0

 

The drive is connected via a Broadcom/LSI 9207-8i SAS controller. I have tried it on both ports, which obviously use different cables, and get the same results each time. I have 7 other HDDs connected to the same controller and none have issues, though all of them are SATA, though I'd be surprised for the controller to have an issue specifically with SAS drives and not SATA drives. I only have this one SAS drive so can't test if others have the same issue.

 

There are PCIe errors too, but I imagine that's just due to the errors coming from the disk rather than the controller itself. If the disk is not doing anything, the PCIe errors do not occur, even though all the other drives are active.

 

I'm mostly looking to sell this drive as I've already replaced it with a SATA version, but don't want to sell a faulty drive, if that is the case. Any input is appreciated.

srv-unraid01-diagnostics-20220417-0340.zip

Edited April 17, 2022 by Shinobu

[JorgeB]

Long test never completed, but drive looks OK, disable spin down for that drive, it should get rid of most of the log errors, then post new diags if there are still more.

[Shinobu]

23 minutes ago, JorgeB said:

Long test never completed, but drive looks OK, disable spin down for that drive, it should get rid of most of the log errors, then post new diags if there are still more.

It's completed a few long tests, none of them have reported any SMART errors. It hasn't been running for the 17 hours or so it takes to complete the current long test, so is showing as still running.

 

I've tested before with disabling spin down on the drive while it was in the array, still got read errors from it. I've just put it back in the system and disabled spin down again, though the drive's not really doing anything as it's unassigned. During a pre-read I'd expect it not to spin down, so I'd be surprised if spin-down were the cause for a pre-read failure.

 

Started a pre-read again to see if it reports the same errors with spindown disabled. I'll allow it to run for a bit and post new diags later

Edited April 19, 2022 by Shinobu

[Shinobu]

10 hours ago, JorgeB said:

Long test never completed, but drive looks OK, disable spin down for that drive, it should get rid of most of the log errors, then post new diags if there are still more.

Right, first set of errors came in at about 65% complete on the pre-read. Before then, everything looks good as it normally does.

 

Spin down is disabled globally, the only spindowns were done manually on drives in the array as I knew I wasn't going to be using them for 8+ hours.

srv-unraid01-diagnostics-20220419-1924.zip

[JorgeB]

Not seeing any disk errors as before.

[Shinobu]

On 4/20/2022 at 7:45 AM, JorgeB said:

Not seeing any disk errors as before.

Right, so the pre-read, clear and post-read have all completed successfully now for some reason. I did disable spin down on this drive before, but not globally, so wonder if that helped somehow.

 

Fairly happy to say the drive appears healthy then and I'm good to sell it on. Thanks for your help.