1.) HP-UX in general gets very confused if you configure two separate NICs to use two IP addresses within the same network segment. This configuration is not supported. (It can be made to work, but this configuration can be unpredictable in failure situations. Not recommended for a HA environment.)
2.) If your heartbeat goes through only one segment, there may be a Single Point of Failure: even one failed NIC sending a constant stream of garbage packets can block the entire segment, unless your network switches are smart enough to isolate a NIC with this type of fault on their own.
Two heartbeat IP segments are obviously better than one. But if you can have only one, you might want to use APA or ServiceGuard's Local LAN Failover on the heartbeat NICs.
With APA, you can join two NICs into an active/active "team", so both NICs are used for heartbeats, even though there is only one heartbeat IP address per node.
With SG's Local LAN Failover, the pair of NICs must be in active/passive configuration, so only one NIC at a time is used. The other one will be idle until the first NIC loses its network link or can otherwise be detected as failed. At that point, the IP address of the first NIC will be switched over to the second NIC.
In configurations like this, if your heartbeat NICs are not all connected to a single switch (to avoid a single point of failure), you should test the failover very carefully when setting up the cluster, e.g. by disconnecting one LAN cable at a time and verifying the nodes can still reach each other.
The most challenging test is probably when Node1's active HB NIC is connected to switch A, Node2's active HB NIC is connected to switch B and the interconnection between the switches is lost.
MK
MK
