Last night I was finishing off my unRAID server build.  As a part of this I fitted new SAS backplanes to my Norco 4020 case (Mine was a really early model where the SATA ports were spaced out right across the backplane which made for a complete cable mess.)  The problem is that the SAS ports are numbered 1-4 right to left, which is opposite of how I used to have it cabled up.

 

So when I started unRAID it said that the disks where in the wrong place.  Disk 1 was where disk 2 should be and visa versa.  Stupidly I clicked the override box and proceeded with the parity check.  Realising my error I stopped it soon after, but not after it had detected 300,000 parity errors.

 

What was happening at this stage?  Was it simply recalculating parity, or was it trying to correct the actual data?  If it is the parity then there is no harm done as I have now corrected the disk locations and fixed the parity.  But does unraid automatically fix the data based on the parity?  It's not the end of the world as I still have everything on a USB hardrive, but it will mean hours of copying again.

If you had the parity drive in the correct slot no harm done.  If you had it transposed with another data disk, you just overwrote the data disk.

 

If it was just data disks transposed, and the parity drive was correctly assigned, then it should not have mattered, AND there should not have been any parity errors.

 

Only way to get parity errors would be if one of your disks was not assigned or working. (At least that is what I theorize)

 

Good luck...

 

Joe L.

Yeah parity drive was where it should be (It's living up top with the cache drive.)  So hopefully no damage was done.  Not sure about the parity errors though.  So unless you assign a data drive to the parity drive slot, data shouldn't be changed from the parity information, right?

As long  as all the drives are present, parity should not ever change.  Even if you transposed the parity disk and a data disk, parity would not change.  (The parity disk will not mount but the parity calculations would be valid)

 

unRAID is looking for an even number of bits set to a "1" in a given position across all disks involved.  As far as "parity" is concerned, there should not be any difference between

 

1 0 1 1 1 1 0 1 0      =    6 bits set to "1", an even number

and

0 1 1 1 0 1 0 0 1      =    transposing first two and last two disks, still has 6 bits set to "1", an even number

 

If you have 300000 parity errors, something else is going on.  I'd do another parity check.  It better come out with zero parity errors.  if not, you have hardware issues to deal with.

 

Yeah, I placed the drives in the right slots and ran another parity check and it came out fine.  I have then added another 2 drives and the parity check is fine.  Under what circumstances does the parity data start to restore the data disks?

Yeah, I placed the drives in the right slots and ran another parity check and it came out fine.  I have then added another 2 drives and the parity check is fine.  Under what circumstances does the parity data start to restore the data disks?

Only if a data disk is currently disabled/missing and subsequently replaced.

I'm not sure what you did.

 

If 2 disks were wrong then unRAID should not have allowed you to even attempt to "upgrade" or "replace" one of the disks by starting the array using the "Start" button. unRAID should have just given a warning that too many drives are missing or too many drives have been replaced and it can not start.

 

The only command that should have worked would have been the Restore (now you type initconfig) and then unRAID would have just calculated a new parity disk meaning no errors would be given.

 

Personally, I would do some md5 checksums or something similar on data copied to the server just for peace of mind that the actual data is OK.

 

I just recently added a disk and then re-arranged my drives both physically and on the devices page. I use the commands

initconfig

mdcmd set invalidslot 99

 

and this allowed the array to start with a parity check instead of a new parity calculation. The parity check was fine with no errors even though the disks were all moved to different locations (both physical ports and device selection slots).

 

Peter