Quoting John Peacock (jpeac...@messagesystems.com): > Hardcoding the snapshot directory to be lxcpath+snaps is somewhat > limiting. I have mounted a btrfs subvolume on /var/lib/lxc, which works > fine; I can freely clone using btrfs snapshots in less than a second. > However, lxc-snapshot fell over completely because /var/lib/lxcsnaps was > a different filesystem, so the snapshot failed.
Hm. Yeah, that's unfortunate. > I wound up creating a "container" directory under /var/lib/lxc and > overriding lxcpath, so that the "containersnaps" was in the same > filesystem. > > Would it be reasonable to create a config value for the snapshot > directory (lxcsnappath) and allow that to point to a hidden directory > under the top level lxcpath folder (which would require making lxc-ls Yes I think I'd be good with that. > smart enough to ignore hidden directories)? Alternatively, the snapshot > folder could alternatively be rootfs.snap under the container folder That might be ok, but not named 'rootfs.snap' since rootfs implies it's not a collection of full container snapshots. But that might be more problematic with zfs. > itself (which is virtually guaranteed to be the same filesystem). > > I would be happy to attempt a patch to make either option available. A patch for the first option would be great, thanks. -serge _______________________________________________ lxc-devel mailing list lxc-devel@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-devel
