I read the concept of SCAN in 11g R2 Grid Infrastructure. In that i come accross the below statements
" 3 SCAN IP's and 3 SCAN Listeners will be up in the Cluster even if onlu ONE NODE IS UP with Clusterware active. This is BECAUSE SCAN listeners are not like NODE LISTENERS; they simply reroute connections so they do not need an instance to be available beneath them to make a database connections".
Could some one please explan me the difference between Scan Listener and Node listener?
The SCAN listener is the new feature in 11g. RAC in 9i and 10g had only node listeners. In earlier releases, the clients had to know each of the node listeners and define load balancing / failover between the node listener.
Each node in the RAC cluster has a TNS Listener listening on the VIP.
The SCAN is a single cluster-wide listener (although actually running as three separate processes, one on each node).. The client now needs to know only the SCAN listener. It does not need information of the node listeners to be maintained in the client-side tnsnames.ora file. SCAN allows for automatic load balancing as the client is directed to the nodes that are "up and able to take new connections" only. Also, as new nodes are added to the cluster, SCAN keeps track of them, the client tnsnames.ora does not need to be updated for new nodes.
It is the SCAN listeners that reroute connections --- to the local listener of the least loaded node (instance).
Each database instance registers with the SCAN listeners as remote listeners. The SCAN listeners are independent of nodes and instances. That is why they "do not need an instance to be available beneath them".
You would have node listeners registered as local listeners on each instance, but running on the VIPs.