On Fri, Nov 26, 2010 at 3:40 AM, david grant <dg@xxxxxxxxxxxxxxx> wrote:
> I am totally anal about having backups of a current operating systems
> and using those for testing I thought tat the best way to do this with
> btrfs was to rsync the file system to another partition but exclude all
> snapshots. This worked very well as long as I mounted only the root file
> system of the copy but what I did was add snapshots to the copy and at
> some point (probably at the start) the btree system was corrupted but I
> only saw this on backtracking and checking all messages. Also, I didn't
> want to boot from a snapshot of my working operating system for fear I
> could screw things up and have to re-install from scratch. In order to
> try again, I deleted all snapshots from the original system, did an
> rsync and checked the copy. I then made a snapshot of the copy via yum,
> used rootflags and it worked!!
>
> So, cwillu, after your scolding of me and your (perfectly reasonable)
> questioning of my understanding, I did get it together for booting.
>
> BUT I am still left with the problem that caused it for me: how do I
> backup (clone?) a btrfs file system with snapshots to another btrfs
> partition (apart from using dd).
I use rsync myself, and explicitly list the subvolumes and mirrors I
want copied, which sounds pretty much like what you were doing. The
corruption you saw definitely wasn't supposed to happen, but depending
on which kernel's you've used and (to a lessor extent) whether a few
particular kernel options are set, isn't too surprising. Things
_have_ been pretty stable for me for the last little while, basically
since 2.6.35+btrfs_git, and I use snapshots quite a bit.
What I use in a nutshell is:
mountpoint "${BACKUP_TO}/${TODAY}" || {
btrfs subvolume snapshot "${BACKUP_TO}/${YESTERDAY}"
"${BACKUP_TO}/${TODAY}" || exit 1
}
btrfs subvolume snapshot / /backup-snap && {
rsync -vaxR --inplace --delete --ignore-errors /backup-snap/./
${BACKUP_TO}/${TODAY}/
btrfs subvolume delete /backup-snap
}
This will give you incremental backups while avoiding the worst of the
duplication. I haven't verified that rsync actually does anything
useful COW-wise at the file level, but that's the idea behind the
--inplace option (without it, rsync writes to a copy, and replaces the
original, which COW can't help with).
This is a still a little ways from actually making new snapshots to
fully reproduce the existing filesystem, but I'm not certain that's
what you were after.
> I just hope I don't get scolded again and told I am not up to it.
In point of fact, I said that you _were_ up to it (you were), and that
you'd have it running in no time (you did) once you understood things
better (you do). I win? (I always win) =D
[Also, please post your replies under the quoted original, not on top.
Easier to follow the thread that way.]
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
