Hey there

 

I was checking my parity (without writing corrections) and got tons of errors (about 800 million). The parity check was canceled at ~75%.

The server is on UPS and there was no unclean shutdown. The hard drives don't have any uncommon SMART data.

So I guess the problem must be somewhere else. 

 

I did a parity swap some months ago to use a larger party HDD but I'm quite sure that I checked the parity afterwards with 0 errors (last check was 3 months ago with 0 errors).

 

I checked the log and got 100 of these messages:

 

Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045064
Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045072
Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045080
...
Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045840
Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045848
Jun 26 01:53:48 unRAID kernel: md: recovery thread: P incorrect, sector=11721045856

 

I guess it's quite strange that all these errors occurred at 01:53:48. I canceled the parity check (at ~ 75%) and I'm currently running SMART extended self-tests on the disks.

 

Hopefully somebody has an idea

 

Edited June 29, 20206 yr by heyper
diagnostics deleted

52 minutes ago, heyper said:

Hopefully somebody has an idea

Possibly, if you post the diagnostics.

7 minutes ago, johnnie.black said:

Possibly, if you post the diagnostics.

Sorry, forgot the attachment. Added

Nothing jumps out, most likely it's related to the earlier parity swap, for some yet unknown reason some users then have sync errors in the extra parity section, run a correcting check, then a non correcting one, as long as no more errors are detected you should be fine.

Thanks for your answer!

Good guess, the errors started after the time the old party check would have been finished (before the parity swap)

 

I'm running a correcting check and afterwards a non-correcting one and let you know the result

Edited June 26, 20206 yr by heyper

The correcting run was done and corrected many errors.

The non-correction check afterwards ended with zero errors, so everything should be fine  

Thanks!

Edited June 29, 20206 yr by heyper

Had this issue recently only to find out that replacing a parity drive with a larger drive, sometimes causes errors within the unused prtion of the drive. This is where it writes all zeros I believe. I had 800 million errors but it was corrected after a single sync.