On Tue, Apr 7, 2015 at 4:20 PM, Hugo Mills <h...@carfax.org.uk> wrote:
> On Tue, Apr 07, 2015 at 01:08:59PM +0200, arnaud gaboury wrote:
>> On Tue, Apr 7, 2015 at 1:03 PM, Hugo Mills <h...@carfax.org.uk> wrote:
>> > On Tue, Apr 07, 2015 at 12:58:28PM +0200, arnaud gaboury wrote:
> [snip]
>> >> After more reading, it seems to me creating a top root subvolume is
>> >> the right thing to do:
>> >> # btrfs subvolume create root
>> >> # btrfs subvolume create root/var
>> >> # btrfs subvolume create root/home
>> >>
>> >> Am I right?
>> >
>> > Not how I'd recommend doing it. I'd say:
>> >
>> > # btrfs sub crea root
>> > # btrfs sub crea var
>> > # btrfs sub crea home
>> >
>> > Then fstab:
>> >
>> > LABEL=myfs / btrfs subvol=root 0 0
>> > LABEL=myfs /var btrfs subvol=var 0 0
>> > LABEL=myfs /home btrfs subvol=home 0 0
>> >
>> > And finally, add this to your kernel command line in the bootloader
>> > to specify how to mount root:
>> >
>> > rootflags=subvol=root
>> >
>> > Hugo.
>>
>> Thank you Hugo. May you please tell why your recommendation is better
>> than my settings, if I thereafter make root as the default sub-volume?
>> I changed the root name to active and avoid any confusion:
>
> Well, keeping the subvolumes separate from each other, rather than
> nested, makes it a bit easier to do things like replace one subvolume
> with an earlier snapshot. If you nest them, then your recovery process
> from (say) a screwed-up / looks a bit like this:
>
> # mount -o subvolid=0 /dev/sda1 /media/btrfs
> # mv /media/btrfs/root /media/btrfs/old-root
> # btrfs sub snap /media/btrfs/old-root/snapshots/last-root /media/btrfs/root
> # btrfs sub snap /media/btrfs/old-root/var /media/btrfs/root/var
> # btrfs sub snap /media/btrfs/old-root/home /media/btrfs/root/home
> # btrfs sub set-default $n /media/btrfs
> # umount /media/btrfs
>
> If you don't nest them, and don't use set-default, then the same
> recovery process would look like this:
>
> # mount -o subvolid=0 /dev/sda1 /media/btrfs
> # mv /media/btrfs/root /media/btrfs/old-root
> # mv /media/btrfs/snapshots/last-root /media/btrfs/root
> # umount /media/btrfs
>
> (This also explains why I wouldn't use set-default, but instead
> mount / using a subvol=<path> option).
>
> Now, for making the snapshots in the first place, with nested
> subvolumes, your snapshot script looks a bit like this:
>
> # btrfs sub snap / /snapshots/last-root
>
> and with the approach I'd recommend, it looks like this:
>
> # mount -o subvolid=0 /dev/sda1 /media/btrfs
> # btrfs sub snap /media/btrfs/root /media/btrfs/snapshots/last-root
> # umount /media/btrfs
>
> which is a little more complex, but the approach is precisely similar
> to the recovery process: mount the top level of the FS on
> /media/btrfs, manipulate the subvolumes, and then unmount. This keeps
> the management of subvolumes a completely separate concern from the
> structure of the filesystem tree.
>
> Hugo.
Thank you for all these info.
Here is my setup now. I hope I fully understand all the hints.
$ cd /mnt/btrfs
# btrfs subvolume create active ---> Just in case I want to add
another subvolumes
# btrfs subvolume create active/rootvol
# btrfs subvolume create active/home
# btrfs subvolume create active/var
# btrfs subvolume create active/etc
$ cd /
# umount /mnt/mountpoint
# mkdir /mnt/active_btrfs
# mount -o subvol=active/rootvol /dev/partition /mnt/active_btrfs
# mkdir -p /mnt/active_btrfs/{home,var,etc}
# mount -o subvol=active/home /dev/partition /mnt/active_btrfs/home
# mount -o subvol=active/var /dev/partition /mnt/active_btrfs/var
# mount -o subvol=active/etc /dev/partition /mnt/active_btrfs/etc
Then install OS (not yet done, so everything can be changed).
The OS will be on a SSD, and is aimed to be a server with regular
snapshots. Snapshots will be in a Btrfs partition on a HD
>
>> # btrfs subvolume list .
>> ID 260 gen 17 top level 5 path active
>> ID 261 gen 15 top level 260 path active/home
>> ID 262 gen 16 top level 260 path active/var
>> ID 263 gen 17 top level 260 path active/etc
>>
>> # btrfs subvolume set-default 260 .
>>
>> In this way, I will not need any kernel command line.
>>
>> >
>
> --
> Hugo Mills | Vote early, vote often
> hugo@... carfax.org.uk |
> http://carfax.org.uk/ |
> PGP: E2AB1DE4 |
--
google.com/+arnaudgabourygabx
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
