[lxc-users] LXD snapshots on separate btrfs volume

Fajar A. Nugraha list at fajar.net
Mon May 30 23:13:21 UTC 2016


On Tue, May 31, 2016 at 3:37 AM, <david.andel at bli.uzh.ch> wrote:

> Hi
>
> How can I best put snapshots of LXD on a separate (slower) btrfs volume?
> Probably I cannot simply symlink /var/lib/lxd/snapshots/ to the other
> btrfs volume, can I?
> I guess that could bring problems since /var/lib/lxd is on the fast btrfs
> volume so the snapshots will probably be generated as subvolumes of that
> volume.
>
>

I'm pretty sure that for btrfs and zfs, lxd snapshot utilize fs snapshot.
So it's not "new volume that uses full space of container", but "a snapshot
of the current container fs, which refer mostly the same blocks, and uses
little extra space"



> Of course the most elegant solution would be to have this done by the fs
> itself as suggested here:
> https://btrfs.wiki.kernel.org/index.php/Project_ideas#Hybrid_Storage
> But that may take a while...
>
> So for the time being I am looking into ways to store my snapshots away
> from the fast disks on slower and cheaper disks.
>
>

That would be really useful if the snapshots were full copy, but no so much
if they were filesystem snapshots (which they are)

If you really want to do it anyway, AFAIK an alternative would be to do it
manually:
- send the fs to new, slower storage
- send the incremental snapshots
- delete OLD snapshot from the original storage

Works with zfs (zfs send), probably works with btrfs as well (haven't
tested it)

-- 
Fajar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxcontainers.org/pipermail/lxc-users/attachments/20160531/7f37b5c0/attachment.html>


More information about the lxc-users mailing list