On Tuesday 08 August 2006 11:01, Jon LaBadie wrote:
> One thing still up in the air (to me anyway) is final tape selection
> from within the tapelist and physical tape changer.  Your description
> gets to which tapes are "eligible" to be selected, but not which tape
> (or runtape number of tapes) among that set is ultimately chosen.

Correct. That is because in the presence of multiple eligible tapes (many in 
P, or none in P and many in I), the exact selected tape is undefined. In 
practice, it is the first such tape encountered by Amanda, but that in turn 
depends on the position of the changer before the scan algorithm begins.

> [snip]

What you describe is more-or-less accurate in terms of final effects, but 
that's not how the algorithm is actually implemented. Instead, taper will 
repeatedly load a tape. If it's active (no-reuse or one of the tapecycle most 
recent tapes), it will move on without further checks. If it's the 
least-recently-used tape or a new tape (that is, if it's in P), then it will 
immediately stop scanning and use that tape. Otherwise, it will note this 
tape for later and move on to the next tape. If at the end of the scan, no P 
tape has been found, then amtape will go back to the first I tape it found 
earlier.

The rule is slightly different if the changer is a gravity stacker; in that 
case, the first tape in I (without regard to P-membership) will be accepted.

Cheers,

--Ian
-- 
Forums for Amanda discussion: http://forums.zmanda.com/