Drive errors during parity check which yet finished with 0 errors

[Opawesome]

Hi all,

 

I am probably too concerned but I hope you guys can give me advice.

 

After a "non-clean" shutdown, Unraid started a parity check. During the process, one of the data drives returned 92 errors. Yet, the parity check finished with 0 errors (see image attached).

 

I am thinking of replacing the incriminated drive (which is rather old) but should I do something else? I am correct to assume that none of my data is corrupted because of the 92 errors ?

 

Many thanks

 

OP

Edited November 28, 2020 by Opawesome

[JonathanM]

Read errors that are successfully calculated from parity and rewritten to the drive in question do not result in failing the drive. However... it's kind of misleading, and at first glance I'm hesitant to trust the result 100%, since the parity was forced to be correct at those addresses that failed the read.

 

That's a whole can of worms that's a little above my pay grade, if @limetech isn't too busy maybe he can pop by and provide some clarity on what happens when a read error occurs during a parity check. My gut says parity is assumed to be correct for those specific sectors, with no way of verifying that unless you have 2 parity drives. That means it's possible for the data to be wrong at those addresses. Not particularly likely, just possible.

 

As far as the drive itself, I'd run the extended smart test and see what results you get.

[Opawesome]

1 hour ago, jonathanm said:

That's a whole can of worms that's a little above my pay grade, if @limetech isn't too busy maybe he can pop by and provide some clarity on what happens when a read error occurs during a parity check. My gut says parity is assumed to be correct for those specific sectors, with no way of verifying that unless you have 2 parity drives.

Hi @jonathanm,

Thank you very much for taking the time to reply and share your thoughts.

I did not mention that I indeed have 2 parity drives. 

I would indeed be glad if @limetech could indeed what happens when a read error occurs during a parity check.

Meanwhile, I am running an extended S.M.A.R.T. test as you suggested. I will report back with the result when available.

Best,

OP.

Edited November 28, 2020 by Opawesome
typo

[JorgeB]

If a read error occurs during a parity check that(those) block(s) will be recalculated from the other disks plus parity and written back to the offending disk, if there's a write error the disk will be disable, if there isn't data can be considered OK (assuming of course that parity was 100% valid before).

[Opawesome]

On 11/29/2020 at 10:06 AM, JorgeB said:

If a read error occurs during a parity check that(those) block(s) will be recalculated from the other disks plus parity and written back to the offending disk, if there's a write error the disk will be disable, if there isn't data can be considered OK (assuming of course that parity was 100% valid before).

Hi @JorgeB,

Many thanks for your reply. So given the context and outcome, I understand that my disk encountered a read error and not a write error.

Attached is the S.M.A.R.T. report after an extended test of the drive. It says the drive passed the test (although the Raw_Read_Error_Rate is 792 !?!) so I may not have to hurry too much replacing it.

Best,

OP 

 

Edited February 24 by Opawesome

[JorgeB]

Disk is OK for now, since it passed the SMART test, but the previous errors were almost certainly a disk problem, and you should keep an eye on these attributes:

ID# ATTRIBUTE_NAME          FLAGS    VALUE WORST THRESH FAIL RAW_VALUE
  1 Raw_Read_Error_Rate     POSR-K   200   200   051    -    792
200 Multi_Zone_Error_Rate   ---R--   200   200   000    -    60

 

If they continue to increase it will likely fail again soon.

[JonathanM]

4 hours ago, JorgeB said:

data can be considered OK (assuming of course that parity was 100% valid before).

This is the issue, as the parity check was due to unclean shutdown.

[JorgeB]

19 hours ago, jonathanm said:

This is the issue, as the parity check was due to unclean shutdown.

I missed that part, but it would be really bad luck that any out sync sectors caused by the unclean shutdown (usually mostly in the beginning of the disk) coincided with the read errors, and if that was the case I would expect they would be detected by parity2.

[Opawesome]

Dear @JorgeB and @jonathanm,

 

Many thanks for your insights.

 

The data on the disk ,o which the errors occurred is a backup from another location. I guess I could run an rsync command with the "checksum option" to double check integrity of that backed-up data.

 

I understand however that because: (i) I had a parity2 drive and (ii) the parity check finished with 0 error, it is likely that my data is not corrupted. 

 

FYI, I also run another parity check today, which resulted in 0 read error and 0 parity check error.

 

Thank you again.

 

Best regards,

OP