Re: RAC and Data Guard Configuration Question
Date: Wed, 07 May 2008 08:18:05 -0500
As Martin pointed out, for the "normal" case where DR is just performing recovery, only one instance will be used to apply all redo. However, the case you should be more concerned about is what happens when you actually activate the DR site and try to run production there. If your production workload requires 4 or 5 nodes in your current 6-node production site, then trying to place that workload on 3 nodes will result in a complete outage and negate any DR benefits you were trying to achieve. In fact, it's almost worse since you'll be trying to run on the DR site, failing, and likely will be troubleshooting the resulting issues on the DR site. Meanwhile, no one is working on trying to get the primary production site back up and running.
In no case would I consider running two instances on the DR servers since that will probably just make things worse by carving up each server's resources (CPU, RAM, I/O) for two instances. I don't think I'd worry about instance naming since all applications should connect via services (where service_name <> instance_name) anyway. I would use the same service names on the DR site as I have on production, though. Shouldn't be a requirement, just might make things easier.
Chris Dunscombe wrote:
> On the DR site we only have 3 nodes, we'd like 6 but 3 is all that we're going
> to get.
> The question is should we configure DR to have 1 instance per node (same as
> production) but this leaves only a total of 3 instances and hence 3 redundant
> UNDO tablespaces etc or have 2 instances per node?
> Also is it best practice to have the instance names on DR the same as in
> Any experiences, advice, "best practice" etc. for this is most welcome.