[SOLVED] Array expanded - parity errors

[Kir]

Hi guys,

This is what I did:

1. Ran parity check - OK

2. Pulled 1Tb HDD out

3. Replaced it with the precleared 2Tb HDD

4. Started the array, allowing it to expand

5. Started parity check again (nocorrect) - 3 errors already (?!?)

 

What should I do?

 

PS: Running UnRaid 4.7

[elkay14]

You got bitten by the md driver bug that still exists in 4.7.

[Kir]

You got bitten by the md driver bug that still exists in 4.7.

 

Could you elaborate, please?

What do I do with these errors now?

[Joe L.]

You got bitten by the md driver bug that still exists in 4.7.

 

Could you elaborate, please?

What do I do with these errors now?

all you can do is re-run the parity check in "CORRECT" mode.

With any luck, the parity errors were at low addresses.

[Kir]

all you can do is re-run the parity check in "CORRECT" mode.

With any luck, the parity errors were at low addresses.

 

Can I check which file(s) these addresses belong to?

 

Also, are these "real" errors (the files on the new disk are corrupt)?

[Joe L.]

all you can do is re-run the parity check in "CORRECT" mode.

With any luck, the parity errors were at low addresses.

 

Can I check which file(s) these addresses belong to?

I've never discovered a method.  All you 'll know is a block address on the disk.  It might be housekeeping area, empty space, somewhere in the middle of a file or files, file-system structure.    If an early address, odds are it is in the housekeeping area used in the journaling of file system transactions.

 

Also, are these "real" errors (the files on the new disk are corrupt)?

They are real. 

 

If they are a result of the bug described (caused by writing to a disk while it is being re-constructed) then the data was not written to the data disk and parity is correct.  The suspect files are those written to the disk while it was being re-constructed.

 

If they are a result of an un-writable (or subsequently un-readable sector on the parity disk), then the data disk is fine and the parity disk in error.  Re-run in CORRECT mode and it will fix them.

 

If they are at random block addresses each time you re-run a parity check with the "NOCORRECT" option, then it is a hardware fault on any one of the disks, memory, the disk controller, or the motherboard. (or a overloaded/defective power supply)

 

 

[Kir]

The errors seemed to be in the houskeeping area - they showed up immediately after starting the parity check.

I ran reiserfsck on all data drives, just to be on a safe side - no errors.

 

So it does seem to be the described bug...