From mboxrd@z Thu Jan 1 00:00:00 1970 From: Max Waterman Subject: Re: RAID5 -> RAID6 Date: Sat, 28 Mar 2009 22:54:56 +0200 Message-ID: <49CE8EA0.9080704@fastmail.co.uk> References: <49CE20B3.9010101@fastmail.co.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org List-Id: linux-raid.ids NeilBrown wrote: > Wait 3 months :-) > Sounds good. I'm in no particular hurry. Increasing capacity would be nice, but I'm not sure I want to do that since I only have a 1TB drive for backup...as such, the slow version of 1a/ sounds reasonable - I have a spare 80GB drive in the same machine that I could make use of to make it not-so-dangerous. I guess I might consider a grow too - perhaps I'll have another drive by then so my backup can be bigger. Thanks for the advice...I'll keep an eye out for the new support. Max. > 2.6.30 should contains support for this sort of conversion. It is > already written (mostly) but still needs some testing. > > > Your options would then include: > 1/ convert that raid5 to a raid6 of the same size but with one > extra device. This device would store all the 'Q' blocks so > it could become a write bottle neck > 1a/ as above, but then restripe the array so that the Q block is > rotated among the drives. This process is either dangerous - in > that a crash would kill your data, or slow - in that all the data > would need to be copied elsewhere in chunks while the corresponding > chunk of the array was restriped. > 2/ convert to raid6 and grow at the same time. i.e. add both spares > using one of them to support the conversion to raid6 and the > other to increase the space. You could then arrange to restripe > an grow at the same time which is faster/safer than striping in-place. > 3/ Possibly you could restripe-and-grow, then restripe-and-shrink > so you end up with a 7 device RAID6 with properly rotating parity, > but don't go through the slow/dangerous restripe-in-place. > I'll need to do some experiments to see if that would actually > be faster