On Sun, Jan 18, 2009 at 10:35 AM, Matt Burkhardt <mlb AT imparisystems DOT com> 
wrote:
> I have three jobs set up

Since nobody else has jumped in .. you have a *lot* of data you want
to get on S3, and S3 transfers just take a long time.  There's no good
way around that.  You can control the amount of data transferred in a
single night by adjusting volume size and runtapes, calculating back
from the average transfer speed you get - probably around 1G.  This
means getting all of your DLEs down to that size.  Since almost all of
your incrementals are tiny, this should work out OK.

For your music DLEs, Amanda will eventually distribute those full
dumps around the cycle.  It did them all at once because you gave it a
10G tape.

Your changes to tapecycle and dumpcycle look good.

If Amanda can't contact a host one night (because the PC is out of the
office), then that host becomes a higher priority for the next backup.
 Of course, if a host is missing for more than 'tapecycle' runs, then
you may end up overwriting all available backups of that host - which
would be bad.

Note that BackupPC has a scheduling algorithm designed for laptops and
other "transient" backup clients.  It might be worth taking a look.

Dustin

-- 
Storage Software Engineer
http://www.zmanda.com