Also sprach Dave Cottlehuber at 10/12/12 2:22 AM:
> My old iMac finally died which gave me the convenient excuse to
> upgrade my MBP to 16GiB RAM + an additional 512GiB SSD -- your envy is
> welcomed.
There's enough to share with everyone.
> Along the way dropbox did its thing and wiped all the data.
Dropbox on ZFS? I take it there's an easier way than tricking Dropbox
into running on a mountpoint than the way I did.
> However it became apparent I don't understand how snaps and clones
> work when it comes to promotion. Why does cloning and promoting a
> *snapshot* move the preceding snaps across to the clone, and not leave
> a CoW copy or something under the original tub/test* tree?
It does, actually. History is still tied to the old snapshots (which
are tied to the original filesystem), but exists in a different
location in the tree. A little confusing, but it works.
ZFS gurus: Is there a way to see that dependency chain at once other
than trying to non-recursively delete a filesystem that has snapshots
and clones?
I've only promoted once (a VirtualBox image for WinXP had gone
horribly wrong and everything was broken), and it's ~3:30 AM, so I'll
leave a more in-depth explanation to someone that isn't half asleep,
but that's the gist of it.
> As a side note if you know how I could merge the resulting snaps back
> into a single timeline (kinda like a git merge rebase combo) that
> would be cosmetically nice. Rsync did the job but it wastes space as
> it treats renamed files as new.
Not a clue... have never thought of doing that. Maybe take a hint from
how Solaris handles boot environments?
.... ....
On second thought, look at how Solaris handles boot environments
(maybe try one yourself). That should clear up some confusion.