I was trying to get monitoring up with HD temperatures with influx/grafana/etc... and in following a guide installed a docker that allowed access to /dev.  I accidentally gave it read/write access.  Moments after turning it on I got a message about errors on 5 of my 10 disks.  I stopped the container right away and started a parity scrub, with write corrections to parity checked.  It only ran for a few mins, but basically found(and corrected) errors on what looks like every sector it scanned.  I forget the total # but it was probably a million or more.  I stopped the scrub, did a reboot and did another scrub(without writing corrections) and it only found one or two errors.

 

How worried should I be and what are my options?  I have a 10 disk setup with double parity.

 

thanks,

 

-dev

It seems reboot just clear the system problem, before that the parity error should be false.

For me I would ensure all disk filesystem health first and then force parity to re-create,I won't concern parity correct or not.

Parity just for drive fault case, it really no help for other issue. I will focus on each disk filesystem ( may have checksum ) and individual file hash stamp.

Edited December 1, 20178 yr by Benson

6 hours ago, Benson said:

It seems reboot just clear the system problem, before that the parity error should be false.

For me I would ensure all disk filesystem health first and then force parity to re-create,I won't concern parity correct or not.

Parity just for drive fault case, it really no help for other issue. I will focus on each disk filesystem ( may have checksum ) and individual file hash stamp.

Yes, correcting parity is a last resort operation that should only be needed after an improper shutdown where not all writes may have reached all drives.

 

Beginning recovery with correcting parity just makes sure that the parity will be destroyed in case one of the drives has issues. Rule #1 is to verify if individual data disks contains valid data and are in operating order. Another thing is that parity can never recover from incorrect data having been written to a disk - the parity will have been already recomputed to consider the bad data as being the expected data.