In article <38625689.22EB5527_at_Unforgettable.com>,
BlueSax_at_Unforgettable.com says...
> Brad wrote:
> >
> > In article <3862258B.29F24895_at_Unforgettable.com>,
> > BlueSax_at_Unforgettable.com says...
> > > I have been asked to investigate the possability of setting up a
> > > fault-tolerant Oracle instance on a Sun 450/SunOS5.6 system. The
> > > requirements are somewhat fuzzy, but maybe someone will recognize what I'm
> > > trying to accomplish.
> > >
> > > The system architects want to set up the file system so that it is mirrored
> > > to another file system. They would like to have Oracle set up so that if
> > > the primary database fails that another instance will be instantly
> > > available which uses the mirrored files.
> > >
> > > In a way this seems to be backwards from parallel server since that schema
> > > would use two different instances against the same files. In my case it
> > > would be an instance with the same name as the primary instance except that
> > > it would be running on another machine (in a cluster configuration).
> > >
> > > I do see some tremendous problems with this since all of the applications
> > > programs are actually running on NT workstations as client applications.
> > > Even if I could do the database end I don't see how I could instantly
> > > switch users over to the alternate instance without diddling with the
> > > tnsnames.ora. If it were simply a matter of having them log in different by
> > > specifying a difference database alias I could see how that could be
> > > handled, but I didn't write the login screens and the database alias is
> > > hardcoded into one of the programs and is handled by a global configurator
> > > for the other applications. Considering that the current usages is about
> > > 80-100 call-center agents, it would appear to me that the deck is stacked
> > > against me for any type of automated switchover.
> >
> > The machine that takes over needs to takeover the IP address of the
> > crashed machine.
>
> The problem with that is that the primary machine itself might not crash.
> There might only be a problem with the database, the file system or with
> networking. Whatever I do, I can only effect the database.
Well then your applications will have to know how to pick the second
server for their connections if the first one doesn't respond.
Received on Thu Dec 23 1999 - 11:37:14 CST
