On Mon, Jul 18, 2005 at 04:01:25PM -0400, Jon LaBadie wrote:

> > > The amrestore command is getting the file in uncompressed format
> > > even if it was compressed.  If you want to transfer the compressed
> > > form I think you could add the -r option.
> > I think it's -c actually and the manpage is a bit ambiguous on the
> > actual workings it. From the phrasing I'd expect the files to be
> > uncompressed if necessary and then recompressed if requested using -c.
> > That would mean processing the whole archive twice - first uncompressing
> > then recompressing.
> It also says the -r options causes the output to be exactly as on "tape".
> Thus it 'should' avoid the double processing.

Right. But ;) it goes on to say that this also includes amdump headers.
What good would that do? I wouldn't be able to extract it using the
original archive/dump program, would I? Or could I feed it to amrecover
as implied below?

> > > Using amrecover might be possible.  I think when you specify to
> > > extract a directory it extracts recursively anything under that
> > > directory.  So adding "/" or "." to the extract list (I forget the
> > > semantics) might make recovery of the entire dump possible using
> > > amrecover.
> > Problem is: amrecover always restores to a directory on the backup
> > server, doesn't it? So I'd have to NFS-mount some directory from the
> > actual target machine to have the data go where I need it. If I want to
> > restore the whole box anyway I find the archives more flexible.
> Someone will correct me if I'm wrong.  The backup program may not be
> available on the tapehost.  The client may have been some other
> architecture or OS.  I'm pretty certain that amrecover was made to
> run on a remote client, transfer the dump using amrestore (and its
> "-r" option) on the tapehost, and do the extraction on the client.

Ah, but that closes the circle: I wanted to know how to extract stuff
from virtual tapes using amrestore in the first place. Once I know that
I'm quite happy to do amrestore -p | ssh box cat/tar/whatever.

But you peeked my interest: How do I feed the raw amdump file to
amrecover to extract individual files on the client? Wouldn't it need
the index from the server? I'm confused. ;)

Ah, reading the man page and RTFMing the documentation helps a bit - I
think I get it now. So I just run amrecover on the box to be restored
and it requests the index from the index server, lets me select the
files I want and then requests those *over the network* as well, but
from the tape host.

That, of course, would be a viable alternative to extracting the
archives using amrestore. But how efficient/fast is that for whole disks?
The documentation on www.amanda.org seems to focus on amrestore for
whole disks and advertises amrecover for individual files.
-- 
Michael
Sauced it has to be, sauced!