On Mon, Jun 26, 2006 at 04:14:29AM -0700, Joe Donner (sent by Nabble.com) 
enlightened us:
> Hi,
> 
> can anyone please help me understand this:  Apparently, you have to
> calculate how many tapes you need for your backup schedule as the number of
> tapes needed for every day the backup will run in your dumpcycle + one extra
> tape to avoid the risk of overwriting a full backup performed at the
> beginning of the previous cycle (quoted very loosely).
>

The minimum amanda will run with is runspercycle * runtapes. I think most
people recommend 2 * runspercycle * runtapes + 1, so that there are 2
complete sets of backups, plus a spare tape.

> I want to:
> 
> 1.  Run a backup job once a day Monday to Friday (all my data will fit onto
> one tape),

All data Monday-Friday will fit on one tape, or each night all data will fit
on one tape?

> 2.  Have four weeks' worth of "historical" backups, i.e. be able to restore
> data to any given point within the last four weeks, and
> 3.  Permanently archive one monthly full backup.
> 
> So how would I calculate the appropriate number of tapes needed?
> 
> 5 days * 4 weeks + 1  = 21 tapes  ??
> 

Assuming 1. above means you need just 1 tape per night, then for your 4
weeks of historical backup, 21 tapes would be sufficient.

> If I don't force full backups, and therefore full backups are sort of
> scattered between different tapes, then I'll have to take out multiple tapes
> every month for archiving, right?  How should that be managed?
> 

You have 2 choices: Remove an entire "set" of tapes (1 complete dumpcycle,
in your case 5 tapes), or run a second configuration with its own set of
tapes that forces fulls. I do the latter.

> What would appropriate values be for dumpcycle, runspercycle, and tapecycle? 
> I'm thinking:
> 
> dumpcycle 7 days (to ensure a full backup at least once a week)
> runspercycle 5 days (Monday to Friday every week)
> tapecycle 21 (number of tapes in rotation assuming my calculation above is
> correct)
> 

That would work nicely.

> I'd really appreciate your advice on this type of backup scheme, because I'm
> struggling a bit to understand Amanda - they should have called her George
> or something :)

You're doing just fine :-)

Matt

-- 
Matt Hyclak
Department of Mathematics 
Department of Social Work
Ohio University
(740) 593-1263