Is it possible to construct snapshots manually?

Is it possible to construct snapshots manually?

[Cerem Cem ASLAN]

By saying "manually", I mean copying files into a subvolume on a
different mountpoint manually, then mark the target as if it is
created by "btrfs send | btrfs receive".

Rationale:

When we delete all common snapshots from source, we have to send whole
snapshot for next time. This ability will prevent sending everything
from scratch unless it's necessary.
(https://unix.stackexchange.com/q/462451/65781)

Possible workaround:

I've got an idea at the moment: What I want can be achieved by
dropping usage of "btrfs send | btrfs receive" completely and using
only rsync for transfers. After transfer, a snapshot can be created
independently on the remote site. Only problem will be handling the
renamed/relocated files, but `rsync --fuzzy` option might help here:
https://serverfault.com/a/489293/261445

Anyway, it would be good to have a built in support for this.

Re: Is it possible to construct snapshots manually?

[Andrei Borzenkov]

19.10.2018 12:41, Cerem Cem ASLAN пишет:
> By saying "manually", I mean copying files into a subvolume on a
> different mountpoint manually, then mark the target as if it is
> created by "btrfs send | btrfs receive".
> 
> Rationale:
> 
> When we delete all common snapshots from source, we have to send whole
> snapshot for next time. This ability will prevent sending everything
> from scratch unless it's necessary.
> (https://unix.stackexchange.com/q/462451/65781)
> 

You can set received UUID as this link suggests. How it can be done was
posted on this list more than once.

> Possible workaround:
> 
> I've got an idea at the moment: What I want can be achieved by
> dropping usage of "btrfs send | btrfs receive" completely and using
> only rsync for transfers. After transfer, a snapshot can be created
> independently on the remote site. Only problem will be handling the
> renamed/relocated files, but `rsync --fuzzy` option might help here:
> https://serverfault.com/a/489293/261445
> 
> Anyway, it would be good to have a built in support for this.
> 

No. It is already possible to shoot oneself in the foot; there is no
need to make it easily done by accident or misunderstanding.