On Tue, Feb 06, 2007 at 09:21:52PM +0100, Joachim Schipper wrote:
> On Tue, Feb 06, 2007 at 10:33:31AM -0500, Josh Grosse wrote:
> > ...It's not in the ports tree, but I will look into what it can do, if it
> > works for me I will report back, and then port it. :)
>
> Always good!
Yes, testing of shunt and flyisofs show that it'll work. However ...
1) If I still want dump to handle volumes, then I must at least write to
FFS file, using 4.4GB or so as temporary transition space. In that
case, I can at use flyisofs to create a functioning ISO from that one
dump "tape volume" -- it saves having to create multiple <4GB files to
meet ISO9660 requirements.
2) If I don't care about volume management, I can use "dump -0af -" and
let shunt/flyisofs do it all ... however, to restore without lots of
intermediate hard disk means that I need multiple levels of named pipe:
example:
mkfifo allvols v1 v2 v3...
restore -rf allvols &
cat v* > allvolls &
mount /dev/cd0c /mnt
cat /mnt/* > v1 &
umount /mnt
mount /dev/cd0c /mnt
cat /mnt/* > v2 &
etc......
With dump managing logical volumes, it's a little bit less effort to
restore.
> If you are trying to get fancy, add checksums and/or redundancy. Note
> that gzip-compressed data is very vulnerable to read errors, which is
> one reason why I've seen people recommend bzip2 (the increased
> interoperability of gzip is nice, though, and bzip2 is supposedly very
> slow - though my PC has always been able to keep up with its burner,
> which is sufficiently fast as far as I'm concerned).
Good point. Checksums may not help except to say, "this offsite backup is
worthless") In this particular case, these backups should only be needed
from optical if the hard drive fails.
I have completed tests with small files, and am now running a test w/
8GB or so of data.
I am going to look into amanda, it may solve all my operational issues. It
could make disaster recovery a little more complex, as it is 3rd party
software.
Even if I end up going with Amanda, I may still port shunt so that is
available for others.
-J-
