On Thursday, 1. August 2002 10:09, you wrote:

> At 09:50 AM 01-08-02 +0200, you wrote:
> >Hello
> >
> >I want to set up two (or more) database server machines redundantly,
> >supporting a "failover" scenario where a "standby" machine takes over
> >operation from the "lead" machine as soon as the "lead" dies. This can be
> >done for example with "HA Linux" (www.linux-ha.org).
> >
> >Now the obvious problem with the database is that I need to keep current a
> >data file on the standby, i. e., somehow send all updates done to the
> > "lead" database also to the standby database.
>
> Have you investigated the database shadowing capability?

As far as I understand, shadow files must be local. I have to update the data
on a second machine.

> If shadowing can't provide the coverage you need then what you
> are probably looking at is some kind of replication solution involving
> triggers that fire when a DML operation occurs and use the event mechanism
> to signal a server-based application to update tables on the "foreign"
> server..

Triggers/events are ok. But how can that server-based application know which
update to send to the foreign server? Is there a way to let an external
application know the SQL command used in an update?
Or could we somehow apply all updates of a transaction to a remote database?

Thanks!
Bernhard

P. S. The kind functionality I am trying to achieve is something like this:
http://www.mysql.com/doc/R/e/Replication_Intro.html
Obviously I don't want to use mySQL, though :-)