Running a correcting parity check and the "sync errors corrected" numbers is increasing a lot, but the parity check is in the space *after* the data. My biggest data drive is 18TB and the parity drives are both 20TB. The check is currently at 18.3 TB. My understanding of parity is that if the parity is bigger than the array, all the *extra* parity should just be 0? No drives are reporting any SMART errors, syslog seems clean other than an odd cron error that has to be unrelated.

deepervisor-diagnostics-20230903-1437.zip

Edited September 3, 20232 yr by hasown

There have been previous reports of this space sometimes not being correctly zeroed when the parity is initially built, but I do not think anyone knows why.   I would let it complete this time and not get worried unless it continues to happen in future checks.

Did you do a parity swap in the recent past?

1 hour ago, JorgeB said:

Did you do a parity swap in the recent past?

Yes, I've done a couple of parity swaps to get the 20TB drives in. My previous parity drives are data drives now.

There's known issue where sometimes the new parity capacity after a parity swap does not get correctly zeroed, so the first check will find those errors, it should be fine once you correct them.

I precleared the parity drives to stress test them before doing the swaps, so I thought they would've been zeroed anyway. But I'll run a second parity check and see how it goes.

14 minutes ago, hasown said:

so I thought they would've been zeroed anyway

It won't matter, bug is that Unraid writes something else instead of zeros to the extra parity capacity during the swap procedure, instead of actually zeroing it.

Just following up to finish this thread: the second parity check just finished and 0 errors corrected. Hopefully helps someone in the future who searches for this issue.

@JorgeB

 

Just wanted to come here and say thanks so much for this thread. I recently did a parity swap from a 6TB to a 12TB parity array. The first 6TB ran perfectly smoothly with 0 errors and I was excited for the process to speed up once it was over 6TB since all the data was done checking. Started getting a ton of `Sync errors corrected` a little ways into that and after all the other disks had shut down. I ended up freaking out and canceling the parity check, but after reading this I'm going to just let it run to completion now.

 

It might be worth considering if this should be documented in the FAQ here? https://docs.unraid.net/legacy/FAQ/parity-swap-procedure/

 

For record I had a disk 1 failure so I did the following:

• My old parity was a 6TB drive, so all the disks were 6TB or less
• Bought a new 12TB disk and ran it through the full pre-clear + zeroing with no issues.
• Did the parity swap procedure, the old parity copied fine over to the new parity drive.
• The new disk 1 (the old parity) rebuilt fine and all the CRCs for the files checked on the rebuilt disk correctly.
• I decided to run a parity check to just be confident everything was ok (it mentions its a good confidence builder at the end)
• The first 6TB of the rebuild went smoothly with 0 `Sync Errors Corrected`
• Around 6.1/6.2 TB into the parity (at this point all the array drives had spun down since all the data drives were < 6TB) I started accumulating a huge number of `Sync Errors Corrected`. I would have assumed this part would have been quick with no errors after the pre-clear/swap procedure?

 

EDIT: Just wanted to add I checked and none of the drives had any SMART errors or any issues during the parity check. I was only seeing this on the unused (supposedly zeroed) portion of the parity drive.

 

Just wanted to post this if anyone else has a similar experience. I'm on Unraid `6.12.11` when I did all this. I'm going to assume these corrections are basically just that the upper unused 6TB of parity wasn't zeroed out properly?

Edited August 14, 20241 yr by phrozen087

27 minutes ago, phrozen087 said:

I'm going to assume these corrections are basically just that the upper unused 6TB of parity wasn't zeroed out properly?

Almost certainly.

Just wanted to post a quick followup for anyone who stumbles on this after me with some observations. My drive had 4kb physical sectors with 8 logical sectors in it. I initially did a full pre-clear and then a parity swap procedure. It looks like during the swap procedure the unused portion of the parity array had 1/8 logical sectors that wasn't zeroed. So I'm guessing the bug is something related to a certain bit not being properly zeroed during a rebuild if that portion of the parity disk is unused. I did some math on it and also was able to confirm from where the parity corrections were made that it was only the unused portions.

 

On following parity checks everything has been smoothly since as well, so the advice in this thread is good also A bit scary for me since I'd had some disk issues, but it seems to just be a quirk/bug when doing a larger parity swap procedure.