On 10/14/2016 01:38 PM, Austin S. Hemmelgarn wrote:
> On 2016-10-13 17:21, Alberto Bursi wrote:
>> Hi, I'm using OpenSUSE on a btrfs volume spanning 2 disks (set as raid1
>> for both metadata and data), no separate /home partition.
>> -----
>> I'd like to be able to clone verbatim the whole volume to another
>> volume, for backup purposes.
>> Now, I think I can do that with a brutal partition clone from my
>> "recovery" (a debian system installed in another drive) and then doing
>> the procedures to deal with a lost drive.
>> But I'd rather prefer a clone on a live system, and without doing a
>> brutal clone as that will keep the same UUIDs.
>> I can(will) script this so even if it is a tedious process or it
>> involves writing a huge line in the commandline it's not an issue.
> I'm not sure there is any way to do this on a live system.  You
> essentially need to either do a block level copy and change the UUID's
> (which recent versions of btrfstune can do), or use send with some
> manual setup to get the subvolumes across.  Both options require the
> filesystem to be effectively read-only, which is not something that any
> modern Linux distro can reliably handle for more than a few minutes.
> If I had to do this, I'd go with the block level copy, since it requires
> a lot less effort, just make sure to use btrfstune to change the UUID's
> when the copy is done (that may take a while itself though, since it has
> to rewrite a lot of metadata).

Heh, one of the reasons I migrated to btrfs was that I wanted to do 
things like these on a live system.

With my script I can already generate any necessary folder structure and 
send over all subvolumes with a bunch of btrfs send | btrfs receive with 
like 5 lines of script.

I was hoping there was some neat trick with btrfs send | btrfs receive
that allowed me to send a snapshot of / to the / of the new volume.

With rsync (from a ro snapshot of /) it should be possible to use the 
subvolume list as exclusion list, but I'd have rather wanted to use a 
btrfs feature instead.


Reply via email to