On Fri, Jul 21, 2006 at 04:05:42PM -0400, Joshua Baker-LePain wrote:
> On Fri, 21 Jul 2006 at 3:57pm, Ronald Vincent Vazquez wrote
> 
> >Running 2.4.5_p1 on Gentoo Linux (no tape spanning, right?).  Ok, let's
> >say that all images but the last one have been written to tape.  The last
> >image to be dumped to tape is too large to fit on the tape remaining.
> >
> >My assumption is that since amanda knows how long the tape is, it won't
> >attempt to write the last image to the first tape but use the second tape
> >to write the last image.  Am I correct in my assumption?  I am still
> >looking for evidence in the documentation but haven't yet found anything.
> 
> The thing is, amanda *doesn't* know how long the tape is.  It knows how 
> long you told it the tape is.  But what if you're using HW compression and 
> that's just a guess?  What if you were being conservative?
> 
> Amanda will try to write the image, and only stop trying when it hits EOT.
> 

Just to add a bit to Joshua's comments.

This behavior is typically not a problem.  The dump is on the
holding disk and you merely start over writing it to the next tape.

But you will have to pay special attention to situations where that
last dump might be writing directly to tape, either you specifically
said "holdingdisk no", the dump was too large for the holding disk,
or whatever other reasons.  Now the dump must be aborted in the middle
and the entire dump restarted.

There could be hell-to-pay if the backup was of a service that was
taken off-line during the dump or needed to fit into a window of
inactivity.

-- 
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)