I would like to be able to define DPA alerts as cluster wide or instance specific, rather than being limited to being defined only against individual instances.
For example, if I have a three node RAC cluster called ORCL with instances ORCL1, ORCL2 and ORCL3 and I activate the default Failed Logins alert for all instances, I'll get three notifications for a failed login event. It seems more appropriate that event like this should definable as being cluster-wide.
It may seem appropriate to only define the event to run against a single instance of the cluster (ORCL1 for example) but if the individual node is down there are situations where the alert would be lost.
It comes down to how connection strings are defined. For a RAC cluster there should be a SCAN (or VIP) definition to accommodate cluster wide events, and host (or VIP) definition to specify individual instances within the cluster. This would then allow events to be more granularly defined at levels appropriate to their function/intent.
