On Thu, 2003-07-31 at 10:42, Nickolay Samofatov wrote:

> Hello, Paul,
>
> > struct backup_id {
> > unsigned long master_uid; // set from a generator in fbsecurity.fdb
> ^^^^^^^^^^^^ This may be replaced original database name
>
> > unsigned short backup_id; // increments on each full backup
> > unsigned short increment_id; // increments on each incremental backup, 0
> ^^^^^^^^^^^^ This is not needed. I've got needed SCN's and levels to check
> order and detect missing parts of backup
> > on full
> > };
>
> If you want to improve incremental backup utility, feel free to do it.
> I'll not have much time for that and I do not have all variety of
> backup devices in my lab here.

I generally don't like incremental backups, simply because they move the
time element from backup time to restore time, and when a critical
database is down, the time to restore one tape is bad enough, without
having to restore a whole bunch. Because dollars to donuts, that
database is going to fail at 2:30pm on a Friday afternoon, when you are
going on vacation the next day, and your wife has already made the hotel
reservations at a too expensive hotel, six hours drive away, and you
give up the room, but still get to pay for it at 10pm. So being able to
restore it, and still leave on time is critical.

I would rather spend time on trying to get a whole-database backup and
restore to run fast enough that the average database can be backed up
within a reasonable time.

PaulS