Brian Cuttler wrote:
Does it also have the ability to write different chunks to different
spool areas if more than a single spool is defined ?
Yes.
Would it be possible to write the different chunks to tape and
assemble them on recovery if needed ? That is, only in the case
where the completely reassembled DLE dump will not fit on the
remaining space of the tape volume.
That implies that you can calculate the remaining tapecapacity exactly.
Until recently (until the introduction of taperalgo) only planner did
calculation
with tape capacity. Now that I can see how good taperalgo does it's job,
I think it's certainly a valid strategy.
There used to be the argument that extracting parts and reassembling
would be more difficult, especially without amanda tools around.
I think it is easier to write an extractor in standard tools (sh, dd,
gzip, grep)
than e.g. reading rait tapes. And of course, you should make the splitting
a configurable option so people can avoid this added complexity.
3) Recovery occurs a lot less often than dumping, at least at
my site. If I have to periodically restore/reassemble multiple
chunks from more than one tape to spool before extracting the
lost files from the DLE the additional step will likely be
worth the added flexability.
I used to say that a good backup program is optimised to make restores
easy, more
then doing the backups. (It is usually at restore time that you are
under pressure...)
Of course I say all that without specing out the modified amtape,
amrestore, or new amreassembly.
I strongly believe that if someone adapts Amanda for this, it would be
included
immediatly. But doing it right is not so easy (I looked at it two years
ago, but got
burried under another workload).
| 