Hi,

 

I am running version 5.0 rc12a and need some advice on the following.

 

I decided to replace a 750GB disk with a new precleared 3TB disk. During the upgrade, one of the other disks (2TB) in the array has started to show errors. At the moment I have 12287 errors in the Error column of the GUI. The upgrade has for some periods run at normal speed, but sometimes the speed has gone down dramatically (at the moment speed is 52.14 KB/sec and estimated finish 390377 minutes!). The read/write light is on steady on this disk, while at short moment all the other disks light up. If I understand correctly the process is trying to relocate data on the 2TB disk by using parity from the other disks.

 

Now, obviously I will have to replace the faulty drive. I suppose the text book way would be to stop the process, insert the previous 750GB disk, choose "trust parity" and then replace the 2TB disk with a new one. Thereafter I would again have to go through the upgrade procedure for the 750GB disk

 

My question is if I can cut some corners in this process. The upgrade at present is at position 1.77 TB on a 3TB array. Thus the upgraded disk is well past the end of data of the old disk that has been upgraded. The new disk is precleared and should for the remaining space just contain zeros. Is it not correct that the remaining upgrade process thus would not alter parity?  If so, I assume it would be safe to stop the upgrade process, use the "trust parity" alternative, and then just replace the faulty disk and reconstruct the data.

 

Any comments would be much appreciated.

 

//Jan

My question is if I can cut some corners in this process. The upgrade at present is at position 1.77 TB on a 3TB array. Thus the upgraded disk is well past the end of data of the old disk that has been upgraded. The new disk is precleared and should for the remaining space just contain zeros. Is it not correct that the remaining upgrade process thus would not alter parity? 

no

If so, I assume it would be safe to stop the upgrade process, use the "trust parity" alternative, and then just replace the faulty disk and reconstruct the data.

Not so.

Thanks for the comments.

 

An update on my issue. Shortly after I made my above post, the upgrade process decided that it was "done" with the faulty 2TB disk and spun it down. Also for some reason another 2TB disk was spun down. This was at position approx 1.79TB. The upgrade process then continued up until expected finish, 3TB. The two spun down disks accumulated slightly more than 50,000,000 errors in the GUI during this process. No additional writes on the parity disk were reported in the GUI.

 

Anyway, I shut the system down and replaced the faulty 2TB disk with a new one, started up the system and initiated a reconstruction of the faulty disk. It fininshed without further problems and the data seems to be quite ok.

 

So to me my theory of zeros in areas past the end of smaller disks still seems valid. How elso would the parity calculation treat the lack of data past the end of a smaller disk - but assuming zero in every position?

 

//Jan

unRAID does not zero replacement disks and does not assume zeros. Only added disks that increase the number of data drives are zeroed.