Redundant Interconnect Usage

Question

Hello everyone,
This is regarding configuring Redundant Interconnect Usage on 11gr2 RAC on linux.
The RAC is already installed and so I think that I've to use "oifcfg" to configure the same. However, I've few questions regarding the same.
oifcfg iflist -p -n
eth0 10.1.1.0 PRIVATE 255.255.255.0
eth0  169.254.0.0  UNKNOWN  255.255.0.0*
eth2 172.16.2.0 PRIVATE 255.255.254.0
My questions are:
- The eth3 is the new interface provided by system admins. So I think below command will work for me
- oifcfg setif -global eth3/169.254.0.0:clusster_interconnect
- 169.254.0.0 is chosen as the same is current subnet used by cluster interconnect as mentioned above.
- Also, just like eth0 does have a IP statically assigned and there is another IP given again to it by oracle for "redundant interconnect usage", I guess eth3 too needs a IP on each node statically assigned and there will be another IP assigned and managed by CRS for RIU. Am I correct ?
thanks

Answer 1

jjk wrote:
This is regarding configuring Redundant Interconnect Usage on 11gr2 RAC on linux.Why not use bonding? 2 interfaces. Each cabled to a separate private switch. Switches connected to one another. Bond the 2 interface as a single interface. Use the bonded interface as private Interconnect interface.
With this config you can loose a cable or an entire switch - but the Interconnect will still function.
This is a much more robust method than using 2 separate interfaces individually. A network socket on a bonded interface survives an underlying infrastructure failure as the logical bonded interface does not fail.
Using 2 interfaces means that a network socket on the failed interface, also fails. And the s/w needs to be intelligent enough to perform a (potentially complex) failover to the 2nd interface, re-establish communication, handshake and then resume the conversation from where it failed. This is a lot more moving parts than using a bonded interface running on top of redundant paths and hardware.