Subject | RE: [firebird-support] Looking for some suggestions...Vertical market apps |
---|---|
Author | Helen Borrie |
Post date | 2004-08-02T14:35:15Z |
At 12:16 AM 3/08/2004 +1000, Alan McDonald wrote:
you can protect your users' sites from "rogue"
sysdbas. Carelessly-deployed installations have problems. Those who know
and use the sysdba login should be tightly restricted - *especially* if
you're deploying a database that is going to have hands-off metadata updates.
integrity relationship, you have to be very careful. Non-nullable fields
should be defined with a default and set to that default as soon as the
metadata change is committed.
If you take the update route as recommended by Alan, it's not necessarily
going to be a piece of cake. If you take the build-anew route recommended
by Lester, with good data extraction and pumping tools, it will take longer
but you have plenty of fall-back and can also log the causes of
failures. But the dump-and-pump approach won't be practicable with very
large databases if you're doing it often.
People are using both approaches successfully.
/heLen
> > b. Is there a way that you know, where an application can detect ifThis is misleading. Shutdown doesn't have "its own problems", as long as
> > other users are logged into the database, to be sure that they can run any
> > utility scripts without logged in users getting in the way?
>
>if you are not using sysdba for all your connections, i.e. you use the
>security db for your users and not your own schema, then you can use the
>shutdown db method but this has it's own problems.
you can protect your users' sites from "rogue"
sysdbas. Carelessly-deployed installations have problems. Those who know
and use the sysdba login should be tightly restricted - *especially* if
you're deploying a database that is going to have hands-off metadata updates.
>If you are adding fields you don't have to worry at all.Oh yes you do. If you add a non-nullable field or add one that alters an
integrity relationship, you have to be very careful. Non-nullable fields
should be defined with a default and set to that default as soon as the
metadata change is committed.
If you take the update route as recommended by Alan, it's not necessarily
going to be a piece of cake. If you take the build-anew route recommended
by Lester, with good data extraction and pumping tools, it will take longer
but you have plenty of fall-back and can also log the causes of
failures. But the dump-and-pump approach won't be practicable with very
large databases if you're doing it often.
People are using both approaches successfully.
/heLen