Personally I'd write a KM that checks to see if active node, if not turn thresholds off. And turn them on if it becomes active. That way you Patrol Central/Classic would not be all Red. Another alternative would be to use a virtual agent that switches, see Patrol Agent reference manual for details of clustered agents.
Sent from my iPhone
Another option here is to use "virtual" Patrol agents, which is something we use in my company. The Patrol software is installed as normal and runs on the standard port on each cluster server, but you then create a third (or more) agent that runs as a cluster resource with its logs/configuration held on a shared SAN disk. The cluster software (such as Veritas Cluster Server) controls this agent, which has a separate IP address (which can be configured using the BindToIPAddress variable in the agent configuration).
Using a virtual agent such as this means you can configure it to monitor specific resources for the cluster application and not alert on them when they are moved, if you ensure that the virtual Patrol agent is shutdown first and brought up last when the cluster application is moved. You can then exclude the clustered processes and filesystems from the static agents.
There is some information in the agent manuals about running virtual agents which will be worth assessing, and I should be able to provide more info if needed.