Parity Read Errors - What are my next steps?

[codearoni]

I run an array with 2 parity drives and 4 data drives. They're all the same model.

Recently, Parity #2 begin having errors. After running a short self-test, I get the result "read error".

 

I am attaching both my diagnostics, and the SMART report for the parity drive in question.
 

What I'm wondering is: what should I do in this situation? If it's a read error, can I simply do a Parity check and write corrections and move on? Or is this something most drastic that needs to be addressed, such as replacing the parity drive.

alexandria-diagnostics-20240213-1915.zip alexandria-smart-20240213-1913.zip

[itimpi]

If a drive cannot pass a SMART test (Short or long) then it we normally recommend it should be replaced.

 

There is also this in the SMART information:

197 Current_Pending_Sector  -O---K   100   100   000    -    16

which is concerning as any recovery after a data drive failure would require it to be read error free to avoid data loss.

 

It is always possible that rewriting the Pending sectors would cause them to be cleared so you could try doing a preclear on the drive to rewrite it to see if that helps, but even if that worked I would still keep a close watch on the drive.

[codearoni]

Thank you @itimpi

 

"which is concerning as any recovery after a data drive failure would require it to be read error free to avoid data loss."

 

Regarding this statement, is it still a concern even though I have two parity drives?

[codearoni]

To further clarify: my expectation is that Parity #2 could burst into flames, and I could still recover the array.

[trurl]

1 minute ago, codearoni said:

Parity #2 could burst into flames

If a data disk needs to be rebuilt, then if every bit of parity and all other data disks in the array can be reliably read, you can reliably rebuild every bit of that data disk.

 

Parity by itself can recover nothing.

[codearoni]

Ty @trurl, Sorry if I'm being obtuse, I just want to make sure I'm understanding everything.

 

In summary: I should get Parity #2 replaced ASAP while I don't have data drive issues. Assuming Parity #2 is replaced and both parity's are OK - I will be ok in the event of a future data drive failure.

In it's current state, if a data drive falls over right now, I might be in trouble during recovery, as Parity #2 is also borked.

 

?

[trurl]

Parity should be enough to recover a single disk, provided

16 minutes ago, trurl said:

every bit of parity and all other data disks in the array can be reliably read

 

[trurl]

https://docs.unraid.net/unraid-os/manual/what-is-unraid/#parity-protected-array

[codearoni]

Thank you!

What is missing from the unraid docs above is the Two Parity use-case.

 

If was my understanding that two parity's mean two drives can fail without a loss of data, is this incorrect?

 

I think I'm getting tripped up on your statement: "every bit of parity"

 

When you say "every bit of parity" are you referring to both parity drives being fully operational, or a single one?

 

Thanks again I appreciate all your explanations!

[trurl]

1 minute ago, codearoni said:

two parity's mean two drives can fail

Correct. By every bit, I mean all bits on all disks involved in the rebuild. If there is only a single disk to be rebuilt, only one parity needs to be involved with all the other data disks.

 

Dual parity allows 2 disks to be rebuilt at the same time, or allows one disk to be rebuilt while another disk is disabled.

[trurl]

Parity2 calculation is different from parity calculation, so they are independent. Parity2 is not a copy of parity. It wouldn't allow 2 disks to fail if it were the same.

[codearoni]

Many thanks appreciate all the info!