On 04/23/2016 09:17 PM, Scott Wheeler wrote: > On Apr 23, 2016, at 9:03 PM, John <john@commonpeople.uk> wrote: > > You don't have to copy the whole data set -- that's the whole notion > behind "copy-on-write". LLVM (or any copy-on-write system) only has > to note the inode tree in the snapshot and then store the live data > as a set of changes against that (i.e. update its inode tree to point > to the changed blocks). This makes snapshots fast. > > (Note: I've never actually used LLVM to do this, but that's the basic > logic behind copy-on-write systems.) > > -Scott Yeah, I agree with John. I don't use LLVM, I use BTRFS, but the way to go is to snapshot your system and backup the snapshot. You'll back up a slightly outdated version, but that doesn't matter. A copy on write filesystem will basically just create a second pointer to your home and only create copies for the snapshot when a block actually changes. Cheers, Bennett -- GPG fingerprint: 871F 1047 7DB3 DDED 5FC4 47B2 26C7 E577 EF96 7808
Attachment:
signature.asc
Description: OpenPGP digital signature