zfs discuss - Dec 2005 - Why does reading volume snapshots require cloning them?

On page 77 of of
http://www.opensolaris.org/os/community/zfs/docs/zfsadmin_1205.pdf is
"Snapshots of volumes cannot be accessed directly, but they can be
cloned..."
Just as snapshots of filesystems are available as read-only filesystems under
.zfs/snapshot, why aren''t snapshots of volumes available as read-only
volumes under something like .zfs/volumesnapshot? Sure, you can clone the volume
snapshot and then read the clone, but if that''s necessary, why not also
make it necessary for filesystems too, and remove the .zfs/snapshot feature? It
seems inconsistent to allow direct readability of filesystem snapshots but not
volume snapshots.
This message posted from opensolaris.org

On Tue, Dec 13, 2005 at 11:12:01AM -0800, Andrew wrote:
> On page 77 of of
> http://www.opensolaris.org/os/community/zfs/docs/zfsadmin_1205.pdf is
> "Snapshots of volumes cannot be accessed directly, but they can be
> cloned..." Just as snapshots of filesystems are available as read-only
> filesystems under .zfs/snapshot, why aren''t snapshots of volumes
> available as read-only volumes under something like
> .zfs/volumesnapshot? Sure, you can clone the volume snapshot and then
> read the clone, but if that''s necessary, why not also make it
> necessary for filesystems too, and remove the .zfs/snapshot feature?
> It seems inconsistent to allow direct readability of filesystem
> snapshots but not volume snapshots.
This seems like a perfectly reasonable RFE, there just hasn''t been a
convincing reason to implement it.  Please file an RFE through
opensolaris.org with some use cases so we can prioritize it
appropriately.  Hopefully you weren''t serious about removing
.zfs/snapshot; I don''t think that''s an option.

Incidentally, would you want the devices to show up as
''/dev/zvol/dsk/pool/vol at snap''?

- Eric

--
Eric Schrock, Solaris Kernel Development       http://blogs.sun.com/eschrock

> Hopefully you weren''t serious about removing
> .zfs/snapshot; I don''t think that''s an option.
Actually I was serious about removing .zfs/snapshot. But I''m not
advocating removing filesystem snapshots; I''m just advocating accessing
them as filesystem at snap rather than as filesystem/.zfs/snapshot/snap.
See my message "Simplifying ZFS via consistent use of COW and snapshot
namespaces" in this forum (at
http://www.opensolaris.org/jive/thread.jspa?threadID=4275&tstart=0).
> Incidentally, would you want the devices to show up as
> ''/dev/zvol/dsk/pool/vol at snap''?
Yes.
This message posted from opensolaris.org

On Tue, Dec 13, 2005 at 05:40:33PM -0800, Andrew wrote:
>
> Actually I was serious about removing .zfs/snapshot. But I''m not
> advocating removing filesystem snapshots; I''m just advocating
> accessing them as filesystem at snap rather than as
>
Accessing them how?  We have an open RFE:

6338043 need a method to access snapshots in alternate locations

That we have been mulling over for a while.  This would expand the
ability of the ''snapdir'' property to allow them to be mounted
in
alternate locations.  For example, you could do:

	# zfs set snapdir=legacy tank/home/eschrock
	# mount -F zfs tank/home/eschrock at yesterday /mnt

However, this still requires an explicit mount to access the filesystem.
Is this what you''re looking for?

- Eric

--
Eric Schrock, Solaris Kernel Development       http://blogs.sun.com/eschrock