On Fri, Jun 30, 2006 at 03:40:50AM -0700, Joe Donner (sent by Nabble.com) wrote:
> 
> Dear all,
> 
> I think I may finally have cracked Amanda, but there are two things not
> quite clear to me:
> 
> 1.  When you do an amrecover, do you HAVE to rewind the tape first?  This
> isn't a problem as such, I'm just wondering whether or not it should work

I believe that amanda expects the tape used for recovery to be at the
start of the tape when you type "y" to continue.

> that way, i.e. in amanda.conf I've got specified tapedev "/dev/nst0", which
> I understand to mean that you're using your tape drive as a no-rewind device
> (which I believe is sort of required by Amanda).

Typically amanda rewinds the tape when it wants the tape at the beginning.
The above amrecover scenario may be an exception.

The use of the no-rewind device is particularly important during tape
writing.  Suppose amanda checks the tape label to see which tape is
in the drive, finds it is the correct tape, then closes the tape device.
If the tape auto-rewound on close, then when amanda started to write a
DLE, it would overwrite the tape label.

> 2.  I create a restore directory on the Amanda server, and then run
> amrecover as root from that directory to test doing restores from tape.  At
> the moment I'm backing up directories from two hosts - the backup server
> itself and a client linux machine.  I've noticed that when I restore stuff
> into the restore directory for one client, and subsequently do a restore to
> the same directory for a second client, then the first client's restored
> files get deleted just before the second client's files are restored.  In
> other words, the restore directory is cleared of all its contents, and then
> the second restore runs.  Is this intended behaviour?
> 
> I would think that in practice you'd set up your clients so that you could
> restore data straight back to the client into a temporary directory, as
> opposed to doing restores to the backup server and then copying the data
> back to the client.

What many mis-understand is that amrecover tries to get the directory
structure back to the state that existed at the date you specify for
the recovery (setdate).  This doesn't mean just get the files that
were there then, but eliminate those that were not there.  So when
you tried to recover a second, totally different client/DLE to the
same directory, it found things from the first client/DLE that did
not exist on the second client/DLE and it eliminated them.  Recovering
each to its own separate empty directory is the way to go.

When you recover directly to the original source tree, if you have
specified and set a date for a week ago, then you are going to be
eliminating those things created/modified since that date.


-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)