On Wed, Nov 16, 2005 at 07:05:15PM +0000, Rodrigo Ventura wrote:
> 
> But are all tapes operations non-blocking? The most efficient way
> seems to be to perform all operations in blocking mode, meaning that
> the next operation is performed after finnishing the current one.

The problem is correct action from the drivers.  An example of what
was happening to me, from the command line, not from within amanda.

I'd do two commands like:

        $ mt rewind
        $ tar x

And I would get an error from tar that the device was not ready,
or did not exist, or ???  I forget the exact message.  The problem
is that the device driver was telling mt the tape was rewound but
I could see it still running.  Even after it stop moving there were
several seconds before the hardware was ready for the next command.

So, "knowing the rewind was complete", mt exited and released the
drive so that tar could begin extracting.  But the drive was not
really ready, so when tar opened it and tried to issue the next
tape command it received an error.

This is the type of delay the sleep is trying to deal with.  A time
when everything says the drive should be ready for the next operation
but it isn't really.

> What happens if one sets sleep to zero?

Set it and find out what happens on your system; then let us know.

I use chg-mtx and had to add some sleeps or increase the sleeps in
that script.  Otherwise it tried to read the next tape header too soon,
thought it was an empty slot or "non-amanda" tape and skipped that slot.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)