I'm running a small Unraid server using 1 drive + 1 parity, plus a ssd cache drive. Automatic monthly parity checks have been completely error free since setting up this server 2 years ago.

Recently, suffered a power outage in the middle of a parity check. Since then, have been seeing a couple parity errors on subsequent parity checks. Just the other day, a non-correcting scan showed 2 errors, I then re-ran a correcting scan which then indicated 3 errors corrected.

I've run extended SMART testing, which is coming back error-free on primary, parity and cache drives. Also ran 4 passes of Memtest86, no errors found. I've checked cables and everything is seated properly.

Other than these 2-3 parity error flags, I'm not having any issues with the server. That being said - should I be looking at something else, or just live with it and monitor for increasing error counts? Thanks.

1 hour ago, AC3 said:

Just the other day, a non-correcting scan showed 2 errors, I then re-ran a correcting scan which then indicated 3 errors corrected.

Were there any more errors besides these two checks?

1 hour ago, AC3 said:

then re-ran a correcting scan which then indicated 3 errors corrected

After correcting parity errors, if you have some doubts, run another parity check after the errors have been corrected. If it still has errors after those were corrected, you still have a problem.

I ran another non correcting check last night, it was looking good with 0 errors till I hit the end of the check this morning and 1 error was flagged. Same sector that was apparently corrected during the previous scan. No other errors aside from these.

Prior to Aug 2 (when the power failure occurred), a non-correcting showed 2 errors.

On Aug 2, I ran a correcting check, but it ended the test by finding a new error ( sector=22465777176 being the new one):

Aug  2 08:29:16 Tower kernel: mdcmd (36): check correct

Aug  2 08:29:16 Tower kernel: md: recovery thread: check P ...

Aug  2 08:33:03 Tower kernel: md: recovery thread: P corrected, sector=106653568

Aug  2 15:12:54 Tower kernel: md: recovery thread: P corrected, sector=10752092000

Aug  3 00:54:06 Tower kernel: md: recovery thread: P corrected, sector=22465777176

Aug  3 02:00:29 Tower kernel: md: sync done. time=63073sec

Aug  3 02:00:29 Tower kernel: md: recovery thread: exit status: 0

The non correcting one last from last night cleared the first 2 errors, but the 3rd error remains:

Aug  5 14:19:16 Tower kernel: mdcmd (36): check nocorrect

Aug  5 14:19:16 Tower kernel: md: recovery thread: check P ...

Aug  6 06:41:59 Tower kernel: md: recovery thread: P incorrect, sector=22465777176

Aug  6 07:48:36 Tower kernel: md: sync done. time=62960sec

Aug  6 07:48:36 Tower kernel: md: recovery thread: exit status: 0

So not sure why this particular sector shows as corrected on Aug 3rd, but still shows as error flag today - and since no errors came up during SMART or memtest, not sure what else to look at. 🤔

Run one more to see if that sector is still flagged incorrectly; also, it is good to run memtest, but since it was a single sector,it can be difficult to detect the error, even if it's bad RAM.

6 hours ago, JorgeB said:

Run one more to see if that sector is still flagged incorrectly; also, it is good to run memtest, but since it was a single sector,it can be difficult to detect the error, even if it's bad RAM.

with correction, or none?

Either way would do, but may as well be correcting.

On 8/6/2026 at 9:36 AM, JorgeB said:

Run one more to see if that sector is still flagged incorrectly; also, it is good to run memtest, but since it was a single sector,it can be difficult to detect the error, even if it's bad RAM.

So i did run one more with correction: it now flagged a new error (3364554496) that wasn't picked up before, but what I'm not getting is that it is still flagging and correcting 22465777176 - which was previously flagged & corrected.

Aug  7 06:38:05 Tower kernel: mdcmd (37): check correct

Aug  7 06:38:05 Tower kernel: md: recovery thread: check P ...

Aug  7 08:38:53 Tower kernel: md: recovery thread: P corrected, sector=3364554496

Aug  7 23:01:50 Tower kernel: md: recovery thread: P corrected, sector=22465777176

Aug  8 00:08:16 Tower kernel: md: sync done. time=63011sec

Aug  8 00:08:16 Tower kernel: md: recovery thread: exit status: 0

Not many errors, but still a bit concerned about data integrity as they are still happening, and same sector showing up after previously being corrected. I'd understand if the drives were throwing SMART errors and reallocating sectors, but that isn't the case. Not sure what to do at this point - further troubleshooting, hardware swap, or just live with it?

The way the errors are occurring suggests a hardware problem; start by running memtest.

4 hours ago, JorgeB said:

The way the errors are occurring suggests a hardware problem; start by running memtest.

I ran 4 passes of memtest, no errors. I ran another extended SMART last night, no errors. I'm on the fence about replacing hardware if there are no hardware errors and nothing to suspect imminent failure.

The timing of when this issue started (power failure during a parity check) makes me wonder if it's not just some odd data corruption, and not hardware related? Just strange how it is continuing to flag a previously corrected sector on subsequent checks (and adding new ones that weren't there on the previous scan). Is it worth a try to unassign the current parity drive, wipe it, and rebuild the array - might that be a suitable option?

No reason to think rebuilding parity ( which I assume is what you mean by "rebuild the array" ) would result in anything other than what a correcting parity check does.

18 hours ago, AC3 said:

if it's not just some odd data corruption, and not hardware related?

Don't think that's an option; by the symptoms, I suspect RAM, CPU, or a disk, but memtest doesn't always find errors, especially when they are small.