>I still have problem to understand what is the strategy of amanda on how
>it will dump the machines.
It's magic :-).
>I do understand that each machine will be dumped at each run. The level
>of this dump will be set through 2 paramters : bumpsize , bumpmult. But
>I don't really know how.
Actually, it's three parameters (bumpsize, bumpmult and bumpdays), and
then a whole lot of other things come into play (length of time since
the last full dump, priority, total estimated size, and so on).
Have you read the two Amanda papers available at the ftp site:
ftp://ftp.amanda.org/pub/amanda/lisa-vii.ps.gz
ftp://ftp.amanda.org/pub/amanda/usenix92.ps.gz
They cover a lot of the scheduling ideas, most of which are still valid
after almost a decade.
>The amandad will assure us a minimum full dump in a DumpCycle. But it
>will spread this work during the TapeCycle. ...
No. The guarantee is that within a dump cycle you will have at least
one full dump of a given file system. The tape cycle has nothing to do
with that, except that it must be larger than dump cycle (otherwise you
could clobber the only full dump you have).
I have configurations with a dump cycle of two weeks but over a hundred
tapes. The higher tape cycle lets me go back further in time to restore
a file, but it has no effect on how often full dumps happen.
>So could I interpret this as
>"Amanda do spread the dump level 0 to keep the same amount of data
>dumped at each run the more constant".
That's the goal, and is called "balancing" in the documentation.
>Briner
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
