On Fri Apr 15, 2016 at 09:26:48PM +1000, Eyal Lebedinsky wrote: > When I request to replace sda with sdj (RAID6 in my case), does it > create sdj from the other members (NOT using sda) or does it make a > copy of (by reading) sda? > > I hope it is the former, assuming sda is having issues and the action > is precautionary. > It's the latter. The data is copied from the existing disk, unless a read-error occurs, in which case it'll be reconstructed from the other members. This will be faster and have a far lower impact on performance than using the other members for everything. If you cannot trust the disk you're replacing then it should be failed and a normal rebuild carried out instead. > And while I am here, can I replace two members at the same time? This > should save time as it takes about 10 hours to reconstruct a 4TB member. > I've no idea on this one I'm afraid. I can't think of any immediate reason why this wouldn't be possible, but whether the current implementation supports it or not is another matter. You could test with loopback devices. Cheers, Robin -- ___ ( ' } | Robin Hill <robin@xxxxxxxxxxxxxxx> | / / ) | Little Jim says .... | // !! | "He fallen in de water !!" |
Attachment:
signature.asc
Description: Digital signature