Splitting tar output (and hairs):

[Jim Kuzdrall]

The reasoning-out-loud below leads to a solution (discovery).

> This syntax did not work?
>
> tar cf - ./ $bkoptions | tee monthly.tar - >/dev/st0

    Correct, it did not work.  Both tar and dd give "/dev/st0 Cannot 
read: Cannot allocate memory.  At beginning of tape, quitting now.  
Error is not recoverable: exiting now."

    I used "dd if=/dev/st0 of=dd-extract.tar ibs=4096 obs=512" to match 
how the tape file was made.

    However, if I write directly to the tape using tar, everything works 
fine.

> I don't know why you are chasing down the directory tree twice...

    Ignorance.  Since it is writing at 4AM (it isn't that late yet is 
it? ... guess not), I had other more pressing priorities for the past 
year or so.  But, as a dedicated, sensitive, empathetic design 
engineer, it brings a tear to the eye to think of that poor hard disk 
arm flailing about on that expensive SCSI platter.
 
> I would have done this instead:
>
> cd /
> tar -cf $bkdir/monthly.tar ./ $bkoptions
> dd if=$bkdir/monthly.tar of=/dev/st0
>
> at worst in restoring you would:
>
> dd if=/dev/st0|(cd $bkdir ; tar xf -)

    That approach works as "dd if=/dev/st0 ibs=4096 | (tar -tf -)".  dd 
needs the block size but tar by itself doesn't.  Or maybe...

    Eureka!  That is it.  I wasn't putting "-b 8" with tar.  With that, 
tar reads the dd tape directly.  Like a dummy, I have been using "-b 
4096".  "b" is the number of 512 blocks.

    I think I can get the tee working now.

    I am guessing the HP DDS-3 should really nip along with that 
blocking factor change.

    Thanks for the wake up call.

Jim Kuzdrall