VIP in 11g RAC with CRS

Question

Hi all,
My system is a three nodes 11.1.0.7 RAC cluster with CRS (same version).
I have a question about how VIP works.
Notes:
eth0 - public interface
eth0:1 - vip interface
eth1: private interface
If I go down with eth0 (ifconfig eth0 down), VIP is relocated to other of the two remain network living nodes. Everything is OK at this moment and only listener resource in CRS stack is OFFLINE, even INST, GSD and ONS are ONLINE and remains in the first node.
Then I go up with the public interface (ifconfig eth0 up). Now everything still remains in the same state... Nothing Change!!!
Questions:
- Node1 VIP resource should go back to the first node???
- Listener resource should go online again???
I can manually start listener resource or relocate vip to the node1 without a problem... but it should be done without manual word, isn't it????
Thanks in advance.

Answer 1

user605547 wrote:
I know that listener is nodeapps (vip) dependent... so if VIP is not running up, listener can't start
All my issue is about automatic VIP relocating after a network failure, so I'm trying to understad what happens and what it should happens.
Scenario 1:
Step 1:
NODE1: ifconfig eth0 down => so vip fails over node2 and listener on node1 also goes down. Any other resource in node1 remains ONLINE.
Step 2:
NODE1: ifconfig eth0 up => nothing happens and everything remains in same state. NODE1 VIP is still running in node2
Question: NODE1 VIP should automatic relocate to node 1 (from node2)??????
11gR1, NODE1 VIP not automatic to node01, we still relocate manual.
This is actually the default default behavior in 10.2.0.4 and 11.1.
Step 3:
NODE1: srvctl start listener -n node1 -l listener_node1 => node1 vip relocates to node 1 and listener_node1 goes upnot sure, perhaps srvctl check about public, then relocate VIP
Scenario 2:
Step 1:
NODE1: ifconfig eth0 down => so vip fails over node2 and listener on node1 also goes down. Any other resource in node1 remains ONLINE.
Step 2:
NODE1: ifconfig eth0 up => nothing happens and everything remains in same state. NODE1 VIP is still running in node2
Step 3:
NODE3: srvctl stop listener -n node3 -l listener_node3 (changing state of any resource not involved with node1) =>
=> listener in node3 goes down, node1_vip relocates from node2 to node1 and listener in node1 goes up
Is it normal??????? It looks like any operation in CRS forces to check NODE1_VIP availability and make it goes up in original node??? Expected behaviour???
When we use srvctl, instance check is performed, that check public ip -> resolved and then relocate VIP
How about ?
ORA_RACG_VIP_FAILBACK in the racgwrap script in the ORACLE_HOME