Locating Parity errors

[pengrus]

So I searched the wiki and the forum but can't seem to find a straight answer to this.  How do I locate an error in parity found during a parity check?  I have dual parity, and my last check unexpectedly showed 7 errors.  I have what seem to be sectors identified in the log, but isn't that on the parity drives?

 

Thanks!

 

-P

tower-diagnostics-20200804-1548.zip

[itimpi]

The simple answer is that you cannot    All Unraid knows is that the parity calculation for a particular sector does not agree with what is stored on the parity drive.    Unraid does not know which drive caused this to happen.

[pengrus]

Thanks @itimpi.

 

Perhaps I was mistaken, I thought there was some way with P+Q parity to be able to find out where the corruption (or not) was.  Is this not the case?

 

-P

[testdasi]

1 hour ago, pengrus said:

Thanks @itimpi.

 

Perhaps I was mistaken, I thought there was some way with P+Q parity to be able to find out where the corruption (or not) was.  Is this not the case?

 

-P

No. That would be mathematically impossible with just P+Q parity alone. Parity correction only applies to a (or two) self-identifying failure(s) i.e. you have to know in advance which block (in Unraid case, disk) fails.

So a parity error can only tells you there is a failure and you have to go identify it yourself.

 

You might be confusing Unraid parity with RAID + checksum filesystems (e.g. ZFS / BTRFS). So for ZFS / BTRFS, a scrub not only verifies parity but also the checksum of the blocks.

So if there's a parity error, it would rely on the checksum to identify the wrong block (actually I think it works the other way round i.e. it identify checksum error on a block and rebuild parity that depends on that block).

 

A way to accomplish what ZFS / BTRFS RAID do is to have array disks use BTRFS file systems (instead of XFS). That way when you have parity error, you can just run a scrub and find out which disk has the error and if it's serious you can then rebuild the disk.

[JorgeB]

8 hours ago, pengrus said:

I thought there was some way with P+Q parity to be able to find out where the corruption (or not) was.  Is this not the case?

There could be a way but it's not implemented since if there is corruption on multiple devices you would end up corrupting more disks, more info here:

https://forums.unraid.net/topic/46170-unraid-server-release-620-beta20-available/page/10/?tab=comments#comment-456693

 

[pengrus]

Thank you all for your responses, I really appreciate it.  Not trying to resurrect a dead thread here, but what would cause 7 parity errors to be found on every subsequent parity check?  Ever since this happened, every monthly check has returned 7 errors.  I am set to correct parity errors during a check, btw, and have been up for 65 days, the reboot was intentional and orderly.  Diagnostics attached, thanks!

 

-P

tower-diagnostics-20201102-1843.zip

[JorgeB]

8 hours ago, pengrus said:

but what would cause 7 parity errors to be found on every subsequent parity check?

All the checks on the diags were non correct, so it will keep finding the same errors.