As part of testing this, we have tried to kill the apps listener on one of
the nodes to see if a failover is initiated. But it does not happen. We also
tried to kill the FNDSM process on one of the node to see if a failover is
initiated. It does not happen. We log a tar for this with Oracle and the
analyst says we have to physically shutdown the node in order for this to
work (but note 743716.1 does not say this).
Secondly, if you have PCP setup, you would setup some managers to be running
on node1 and some managers running on node 2 and you would expect the
managers running on node 1 to failover to node 2 and vice versa in the event
of a crash. The analyst says the purpose of node 2 is to run only the apps
listener, FNDSM and Internal monitor and that does not make sense to us.
That might be the default configuration. But customers may want to
distribute the middle tier load (or even database load in RAC) by forcing
some managers to run on different nodes (and differet two_tasks) and I think
that should be one of the advantages of implementing PCP.
How is PCP testing done at your place when this is implemented?
Thank you for your input