zfs discuss - Mar 2007 - recover user error

ZFS claims that it can recover user error such as accidentally deleting of
files. How does it work? Does it only work for mirrored or RAID-Z pool? What is
the command to perform the task?

Also for COW, I understand that during the transaction (while data is been
undated), ZFS keeps a copy of the previous data. However, once the transaction
is successfully completed, isn''t it the case that the previous data
blocks are unreferenced and can be used by future writes? So there should be no
guarantee that the previous data can be recovered.

Thanks
 
 
This message posted from opensolaris.org

Hi Brian,

Brian Gao wrote:
> ZFS claims that it can recover user error such as accidentally
> deleting of files.
Can you show us where you read that ? At the moment, the only way this
is possible, is by taking regular snapshots of your ZFS filesystems,
allowing users to go back to a previous snapshot in order to recover
their deleted files.

Have a look at the section about snapshots in the zfs(1M) man page at:
http://www.opensolaris.org/os/community/zfs/docs
to learn more about snapshots.
> How does it work?
If a user takes a snapshot, a data on that filesystem is frozen
at that point in time, allowing a user to read that copy at any time.
Subsequent changes filesystem are done to copies of any changed blocks.
You might be interested in reading Matt''s blog post at:
http://blogs.sun.com/ahrens/entry/is_it_magic
> Does it only work for mirrored or RAID-Z pool?
Nope snapshots happen at the filesystem level - not the pool level, so
all ZFS filesystems can use snapshots. Think of ZFS pools as ways to
configure the physical storage on your system in terms of
performance/redundancy, and ZFS filesystems as ways of managing or
organising your data.
> What is the command to perform the task? 
"zfs snapshot mypool/myfilesystem at snapshotname"

Some folks (me included) have written scripts to take snapshots at
regular intervals, so that older user data is always available.
> Also for COW, I understand that during the transaction (while data is been
undated)
>, ZFS keeps a copy of the previous data. However, once the transaction is
> successfully completed, isn''t it the case that the previous data
blocks are
 > unreferenced and can be used by future writes? So there should be no 
guarantee
 > that the previous data can be recovered.

Right - but only in the case where those blocks aren''t referenced by a
previous snapshot, in which case the data isn''t unreferenced.

hope this helps,

	cheers,
			tim


-- 
Tim Foster, Sun Microsystems Inc, Solaris Engineering Ops
				  http://blogs.sun.com/timf

On March 6, 2007 11:23:26 AM -0800 Brian Gao <brianchaogao at gmail.com>
wrote:
> ZFS claims that it can recover user error such as accidentally deleting
> of files. How does it work? Does it only work for mirrored or RAID-Z
> pool? What is the command to perform the task?
zfs snapshot
> Also for COW, I understand that during the transaction (while data is
> been undated), ZFS keeps a copy of the previous data. However, once the
> transaction is successfully completed, isn''t it the case that the
> previous data blocks are unreferenced and can be used by future writes?
Yes, except of course when you''ve taken a snapshot.
> So there should be no guarantee that the previous data can be recovered.
That''s right.  If you haven''t taken a snapshot the data is
lost.

-frank