Multi-volume multi-filesystem restore problems...

Question

I'll try to make this short. I wrote a script to automate backing up Solaris 2.6 and 8 systems but am having problems with the restores when more than one tape is involved -- ufsrestore seems to become confused.
There are two slices being backed up, root and /var/spool on slices 0 and 3 respectively (1 is swap, 2 is backup, 4 thru 6 unused). I dump them using the command "ufsdump 0uf /dev/rmt/0bn" for each filesystem and restore them with "ufsrestore rvf /dev..."; here is an example of what happens on the restore when I needed 3 tapes to backup.
- ufsrestore starts with 1st tape (Vol 1)
- ufsrestore requests Vol 2
- ufsrestore requests Vol 3
- restore of root slice ends, /var/spool begins
- ufsrestore requests Vol 2
- ufsrestore requests Vol 3
- ufsrestore requests Vol 4 (huh? only 3 tapes!)
- now stuck in endless loop asking for non-existent volumes increasing in number. It searches each tape I insert for a few seconds then asks for the next in this "sequence". If I break out of the script I find that the directory structure underneath /var/spool is not recreated.
I have tried inserting Vol 1 when it requests Vol 2 at the beginning of the /var/spool slice, no go; ejecting and re-inserting Vol 3 didn't work either. As a test I created a system with one large root slice and I was able to successfully restore it even from multiple tapes with no problem.
Is there a bug in ufsdump and / or ufsrestore that I'm not aware of? Why does it work for multiple tapes one filesystem or multiple filesystems one tape, but choke on multiple filesystems with multiple tapes? HELP!!!

Answer 1

I'm a little confused about your chronology of events. It would help to
have the exact commands you are typing. But first, perhaps a little
shaft of light when all around is dark...
ufsdump has no idea how many tapes are actually needed. Assuming
that ufsrestore would know is also untrue. This explains why it keeps
asking for the next volume when you thought there shouldn't be quite so
many volumes.
The way I do this looks something like:
dump script:
ufsdump 0uf /dev/rmt/0cn /
ufsdump 0uf /dev/rmt/0cn /var/spool
restore process is then:
mt -f /dev/rmt/0 rew (yes, I know that /dev/rmt/0 is usually the default :-)
ufsrestore rf /dev/rmt/0cn (restores /)
cd /var/spool
ufsrestore rf /dev/rmt/0cn (restores /var/spool)
What happens is that for each file system dumped a separate tape file
(watch out for terminology overloading here) is created. The "b" option
(see man mtio(7i)) won't fsf the tape to the next file, which is why I
don't like to use it. If you use /dev/rmt/0bn for the ufsrestore you will
need to put in a "mt fsf" to get to the next tape file to read your
/var/spool ufsdump.
HTH, if not send more info.
-- richard