On Fri, Dec 30, 2016 at 03:49:49PM +0100, Ole-Morten Duesund wrote:
On 30/12/16 15:38, Andy Smith wrote:
Hi Andy,
On Fri, Dec 30, 2016 at 02:07:56PM +0000, Andy Bennett wrote:
Optimise for speed and ease of restoration: back
up everything
I think the constantly-changing files like log files will interact
very badly with BitFolk's backup system and result in the need to
purchase very large amounts of storage. Much as I would enjoy the
increase in sales of storage, I expect this strategy will only be
feasible with the "unlimited backup space" type providers.
Crashplan, etc.
Which service do you use to back up your VPS(es) in this way? Maybe
you could add to the wiki article…
I back up everything, I use btrbk and btrfs-snapshots to a server at
home. It does incremental backups on a block-basis and is insanely
fast.
I does require one to run btrfs both on the VM and the
backup-destination though (actually, the destination could be
btrfs-images, but I've never tried that). And a certain organization
of btrfs-subvolumes at the source.
You can store the send streams as files. If you're using
incrementals with btrfs send -p, you need to keep the full chain of
incrementals from the last full. In that respect, it's much like a
traditional three-level tape schedule.
Then the restore process is just btrfs receive with each send
stream in turn, from an initial full send, following a chain of
incrementals to the one you want to restore. It does require a full
restore, though -- you can't just pick out individual files.
It's a bit hard to get going the first time, but
it is VERY nice
once it's up and running.
It's got its moments, certainly. :)
Hugo.
--
Hugo Mills | A clear conscience. Where did you get this taste for
hugo@... carfax.org.uk | luxuries, Bernard?
http://carfax.org.uk/ | Sir Humphrey
PGP: E2AB1DE4 | Yes, Prime Minister