Up until 2.4.2, a negative value for chunksize (other than -1) caused
images estimated to be larger than the absolute value to go direct to
tape. For instance, "-1024 Mb" would cause anything larger than 1 GByte
to go direct to tape and anything smaller to go through the holding disk
(if there is enough space, etc).
Before I conned the other administrators here into providing obscene
amounts of holding disk space :-), I used to set this to slightly less
than half the holding disk space to prevent Amanda from going into what
I call "ping pong" mode with lots of large dumps. Since there was not
enough space for the two largest images, but was enough for one of each
of the several largest, it would spend a long time dumping into the
holding disk with no tape activity, then a long time writing to tape
with no other dump activity, then go back to dumping. Very un-parallel.
And there might be other uses for forcing direct to tape.
With 2.4.2, backup images may be split across multiple holding disks
(yeah!) and that makes using chunksize for this direct to tape feature
seem like not such a good idea since it's not really related to each
specific holding disk. So we're looking for input:
* Does anyone use negative chunksize (other than -1) to force direct
to tape?
* The suggested change would be to add a new "maxholding" parameter
in the general (non-holdingdisk) area to do the same thing. Would
that be acceptable?
FYI, 2.4.2 will complain if it sees a negative chunksize (other than -1)
to help catch anyone who forgets.
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
