> > I've never heard of NTBACKUP - ... And I also have a system tool called
> merely "Backup" which I never use.
>
> This is what was called NTBACKUP since NT 4.0.
> It is really powerful, since it is afaik the only backup programm
> which can
> make use of volume shadow copies at that time and can also backup Exchange
> servers without additional software - if the exchange admin tools are
> installed on the machine.

This was merely an agent to shutdown exchange services for the time taken to
backup the critical files which were locked for writing by exchange server.

>
> > I'm not sure I understand you but you appear to be confirming
> what I have
> said. i.e. running processes continue while the copying is taking place.
>
> This is true, but it is NO problem. Why should it? You can keep updating
> rows while other transactions still see old versions in an RDBS. This is
> true for volume shadow copies also. They appear to be some kind of
> transactions to the file system. The backup programm can read the NOT
> FURTHER MODIFYING file, while the other processes can keep modifing the
> files - including databases.

It IS a problem - a real one. Don't confuse DB server transactions and the
way the FB engine writes and reads from the file under it's sole control
iwht wht the OS does when it copies a file. They are not reading the file in
the same way. The OS only knows disk sectors while the db engine knows db
pages as well.

The only way NTBACKUP would work successfully all the time is for it to have
an "agent" which stops the fb server while it copies the gdb file, then
starts the fb server again.

>
> > The file in total is not locked but disk sector by disk sector it IS
> locked while copying, this is an OS thing which will not change.
>
> Are you sure about this? I don't really know, how volume shadow copies are
> implemented, but my understanding of all the white papers I have read
> suggest, that it's primary design goal was not to change the behaviour of
> already running processes - including having no additional locks. This CAN
> be done by the os, if it simply reads the existing blocks before
> overwriting
> them and stalling the writing process for that short time.

I'm sure about this - it's a problem.

>
> > I think you might be confusing other things with regard to power failure
> and the forced writes property.
>
> Because of the semantics of volume shadow copies of drives like
> transactions
> in database systems, the state of the file (visible for the
> backup process)
> will stall at the time, the volume shadow copy was created. Just as if the
> power had failed at that very time.
>
> Regards,
> Steffen

Again - they may be like db transactions semantically but they are not the
same. The OS and NTBACKUP do not know about the way FB structures it's paged
data.

Alan