On Wed, Jul 23, 2008 at 3:49 PM, Dustin J. Mitchell <dustin AT zmanda DOT com> 
wrote:
> This is all very strange.  Did you compile from source?  If so, can I
> get you to put some dbprintf's in apply_config_overwrites and send me
> the results?  Better, is there any chance you could hop into #amanda
> on IRC so we can figure this out?

OK, we worked this out in IRC.  The deal with the -o options is that
there was a fix that wasn't properly ported to the 2.6.0 branch.  The
fix will be included in 2.6.0p2.

As for the block size and timing issues, Doyle's looking at overall
runtime of tar vs. overall runtime of Amanda, which isn't a fair
comparison.  Amanda is:
 - making estimates
 - planning
 - appying compression
 - sending the dump over a network connection (on the loopback
interface, but still)
 - creating an index
 - creating an incremental file
 - spooling to holding disk
 - writing a header file to the tape device

I expect that the actual taper time will be closer to the duration of
the tar invocation.

Dustin

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