I run the parity checks on a monthly basis. The errors jumped right around the same time I replaced one 24tb parity drive with a new 24tb parity drive. I have run the Parity check - Check will start Parity-Check. (Schedule) With this, --> Write corrections to parity. Checked then immediately unchecked. The result does not reduce the error count from 2900 or so back to zero as expected.

I went down a rabbit hole with ChatGPT and it seems to have ruled out bad RAM.

I am not sure if this will help or hurt the process of solving the problem, but here is my diagnostic journey so far...

https://chatgpt.com/share/e/6a2a1e6e-5fb4-832d-b4be-8ede9843fd11

unraid-diagnostics-20260605-1935.zip

Unraid diagnostics are also attached.

Happy to provide any additional information.

Thanks!

Edited June 17Jun 17 by Darkoptix

Diags only have a no-correcting check, reboot to clear the logs, run a correcting check, then another one (correct or non correct) right after and without rebooting.

Rebooted, and I am running one with the blue Checkmark toggled ON "Write corrections to parity"

14 hours ago, Darkoptix said:

I went down a rabbit hole with ChatGPT and it seems to have ruled out bad RAM.

Your link requires me to login to ChatGPT before I can see what transpired. Not gonna do it. Perhaps you could summarize it yourself.

44 minutes ago, trurl said:

Your link requires me to login to ChatGPT before I can see what transpired. Not gonna do it. Perhaps you could summarize it yourself.

So now I guess I just need to run a few more Parity checks as JorgeB recommended. Seems like I am doing the same thing over as I have already done but I have not specifically done the order.

1. Reboot

2. Run checked

3. Run unchecked

Although I have done 2, then 3 multiple times.

It is running right now with mdResyncCorr=1

And if the second check finds errors, post new diags.

Well, just like that. I ask for help, and it fixes itself lol. The RESTART then instantly do both of the checks immediately after reboot is the only thing I changed.

Thanks JorgeB

Edited June 17Jun 17 by Darkoptix

Ran another check just to be sure, but 5 parity errors snuck back in. You can see the history 2975 -->2980 -->2985-->0-->now 5

unraid-diagnostics-20260620-1028.zip

ChatGPT plan in case it helps anyone else. (Sounds reasonable to me).

There it is.
Jun 17 08:26:46 UNRAID emhttpd: unclean shutdown detected
And:
reboot system boot Wed Jun 17 08:25
That means the server experienced an unclean shutdown sometime before June 17.
Given your latest parity history:
2975 -> 2980 -> 2985 -> 0 -> 5
I think we've finally got the story:
You had a long-standing parity mismatch (~3000 sectors).
The recent correcting check successfully fixed it (proven by the subsequent 0).
Since then, only 5 parity errors have appeared.
You have a documented unclean shutdown on June 17.
An unclean shutdown during active writes is one of the most common reasons to get a handful of parity errors afterward.

My current conclusion
I no longer think you have:
a failing parity drive
a failing data drive
a broken Adaptec controller
a persistent parity corruption problem

The evidence now supports:
✅ The 2980-error issue was corrected.
✅ The array is currently healthy.
✅ The 5 new errors are very plausibly from the June 17 unclean shutdown.

What I'd do now
Do nothing drastic.
Let the current correcting parity check finish.
Then:
Run a non-correcting parity check next month.
Watch the error count.
Interpretation:
0 errors → issue resolved.
1–10 errors after another unclean shutdown → not surprising.
Hundreds/thousands of errors reappear → we reopen the investigation.
One thing I would still do

At some point, run Memtest overnight. Not because I think it's the primary suspect anymore, but because:
64GB non-ECC RAM
storage server
easy preventative validation
It's no longer urgent.
Bottom line
The most important datapoint of the entire investigation is that you got:
2985
↓
0
↓
5

That proves parity corrections are working and that the original large mismatch is gone. Right now, your array looks much healthier than it did when we started.

If there was an unclean shutdown, a few sync errors are normal.