agent, clusters and failover

Maybe a trivial question:
how would you set oracle agents (10g in particular) in a cluster environment?

In a simple failover system? Would you keep an agent active on every node or only on the one which is giving service?

And for mutual failover? Would you bind the agent at the service(s) or at the node?

I see more advantages having an agent on every node but I'd welcome different opinions.
Anybody with experience in clustered services monitored by agents?

Thank you
Fabrizio Magni

-- 

Fabrizio Magni

fabrizio.magni_at_mycontinent.com

replace mycontinent with europe