OK.  I changed the blocksize to fixed at 32768 and sure enough, I was
able to flush.  I will now go back and make things more generic as you
suggest to a blocksize of 0 for variable.

I also looked at all the info you referenced on this problem before and
using the stinit.def file to define the tape parameters on bootup.  This
would also be good.

Now the question:  I will have to go back and relabel all those tapes,
no ?

And to Jon, yes I was using /dev/nst0 for Amanda, and /dev/st0 for the
mt test.  Incidently, there is a read/write tester (binary executable)
delivered with this tape drive (Seagate Scorpion 40) specifically
written for Linux.  It worked pretty well to establish that the tape
drive was working as expected.  This is why I suspected an AMANDA setup
problem.

Amazing...all these years of using AMANDA I never ran up on this problem
before.

On Fri, 2004-04-30 at 11:39, Paul Bijnens wrote:

> This has fixed blocks of 512 bytes.
> I never understood why this is the default value on some configs.
> It works better if you set variable blocksize, indicated by 0 bytes.
> 
> I never exactly understood the boundary cases, but it seems
> that each OS has it's own semantics when reading blocks from tape.

-- 
James D. Freels, Ph.D.
Oak Ridge National Laboratory
freelsjd AT ornl DOT gov