On Thursday 08 August 2002 07:44, Martinez, Michael - CSREES/ISTM 
wrote:
>Why does hardware compression confuse Amanda?

To state the not so obvious, amanda counts bytes *sent to the tape*.
It does this after having applied any compression you may or may not 
have specified in your dumptype entry for that disklist entry.

When the tape drives compressor is on, then amanda and the drive are 
speaking a different language as far as how much data can fit on a 
tape.  Since some data will expand when fed to that dumb RLL 
encoder in the drive (like a .bz2 file may double in size!) you'll 
be forced to use very conservative estimates for the tapes capacity 
in your tapetype specification, thereby wasting tape capacity.

With drive compression turned off, and tapetype run against your 
drive and that data used as your tapetype, then amanda knows to the 
byte how nuch she can put on a tape.  By using a dumptype that 
varies from disklist entry to disklist entry, you can turn the 
software compression on and off to suit the data in the individual 
disklist entry.  Start out by useing 'compress server(client?) 
best' for everything and read the email you'll get when the run is 
done.  Any disklist entry that reports a compression of more than 
100% is an entry containing data that even software compression 
can't compress so it got larger.  OTOH, I have 3 or 4 entries that 
regularly report a compression ratio of less than 20%, so they get 
really small on the tape.

Hopefully this clarifies it a bit for you.

-- 
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz  512M
99.10% setiathome rank, not too shabby for a WV hillbilly