I'm supporting a customer just installing a GI 126.96.36.199.3 on Aix 6.1, I was asked for HAIP requirements in order to properly set up the IC interfaces and network.
Can you confirm me that all statement shown below are true?
[as an example: we have 2 physical interfaces per machine to be user for IC]
- all interfaces on different subnets
- all interfaces have to communicate with each other
in other words, in order to let all interfaces to communicate with each other (even if they are on different subnets) we must implement such a routing network
configuration (layer 3) ... layer 2 network configuration is not enough.
11gR2 Grid Infrastructure Redundant Interconnect and ora.cluster_interconnect.haip (Doc ID 1210883.1)
That should clarify all your questions. Especially this extract:
NOTE: If using the 188.8.131.52 (and above) Redundant Interconnect/HAIP feature (as documented in CASE 2 below) - At present it is REQUIRED that all interconnect interfaces be placed on separate subnets. If the interfaces are all on the same subnet and the cable is pulled from the first NIC in the routing table a rebootless-restart or node reboot will occur.
Edited by: Sebastian Solbach (DBA Community) on Oct 8, 2012 6:50 AM
so, as per MOS note you posted, once "it is REQUIRED that all interconnect interfaces be placed on separate subnets", in order to let all interfaces
communicate to each other (even if they are on different subnets) it's obvious that a layer 2 network configuration is not enough, it's needed also
a layer 3 network configuration with such a routing mechanism configured, isn't it?
"Multiple switches can be deployed if there's more than one private network adapters on each node, in case one network adapter fails, the HAIP on that network segment will be failed over to others on all nodes"
If you use multiple switches for redundancy and they are not interconnected (as described above), both HAIP addresses of a subnet will failover to the other subnet.