Right now, I am using fairly expensive tape media to perform a once-only
archive before deleting a lot of historical and unused data.  After having
used 5 tapes up it occured to me that I might be able to fill tapes to
capacity with compression if I perform multiple amdumps with no tape in the
drive and then use amflush multiple times once I filled the staging area up
sufficiently (more data than could fit onto two tapes even with hardware
compression on).  I checked the source, and indeed it seems amflush will
happily do this for me (I am testing this right now as I write this), based
on the way it handles errors.

I am also now finding that I want to go back and restore the data from the
other 5 tapes and rewrite them, removing some level 1 dumps that got in
there by accident from configuration errors.

Based on what I have seen so far of amanda's behaviour, it would be possible
to do this by:

- amrestore the tape(s) to the staging area
- delete any unwanted restored files
- amflush as many times as necessary
- clean up the logfiles and index directories so that the only references to
those dumpfiles is in the newly created logfiles (that would have been
created by amflush)

Also, in theory, using this same approach, I could create a copy of the
indexes in a new configuration and restore/re-dump a second copy of the data
to smaller, less expensive media (assuming I never encounter a single backup
image bigger than a single tape).  The reason I mention a second copy of the
media is I'm not so sure amanda would be happy to have two tapes for each of
the backup volumes.

If someone could let me know if my ideas are valid despite being unorthodox,
I'd appreciate it.

Roy

---
Roy Hooper
Project Manager & Senior UNIX Architect
Decisive Technologies, Inc.
[EMAIL PROTECTED]



Reply via email to