On Wed, Jul 1, 2009 at 9:21 AM, Andreas
Kuntzagk<andreas.kuntzagk AT mdc-berlin DOT de> wrote:
> we have some genome sequencers which produce lots of data. ATM these are
> stored on 2 big Solaris Fileservers (~34TB each) but in the long term we
> want to move these to some kind of archive. One idea is to bye a second tape
> library (we already have one SL24) and write these datasets to tapes. Since
> these datasets can span more then one tape (upto 3.5 TB) just using tar or
> dump is not the solution (at least I'm not aware how to do this).
> So can I use amanda for this? I found something about long-term archiving in
> the faq but it is not to elaborate.

I'm surprised nobody wanted to answer this.

Amanda is really designed around the idea of repeatedly backing up the
same data, rather than doing one-off dumps.  Folks certainly have
cajoled Amanda into doing what you describe, through manual
configuration (editing the disklist all the time).

If you're feeling more motivated, we're to the point where Amanda has
the tools to do this sort of thing for you -- with a bit of Perl and
some studying of the Amanda::* POD, you could write an application
that takes a stream as input (say, a tarball) and write it to tape:

tar -czf - /my/sequences | write-to-tapes SequenceBackup

and still use the usual Amanda recovery tools to read that data
(amfetchdump, probably -- amrecover would be useless since there are
no indices, but I doubt you need indices).

If you're interested in working on a script like this, I'll be happy to help.

Dustin

-- 
Open Source Storage Engineer
http://www.zmanda.com