Did You get an answer to this ?
I have done a two-site FS5300 10GB RDMA/PBHA implementation and in the case of Direct-connect You can only connect one 10GB port pr node in System FS01 to one 10GB port in a node in System FS02. You can not have two pr. node From System FS01 to System FS02. That only works with Fiber channel.
Originally, I had on FS01 and FS02 these four ports:
Link Config
1 FS01 P1 Node1 (10.1.1.10/24) -> FS02 P1 Node1 (10.1.1.20/24)
2 FS01 P2 Node1 (10.1.2.10/24) -> FS02 P2 Node1 (10.1.2.20/24)
3 FS01 P1 Node2 (10.1.3.10/24) -> FS02 P1 Node2 (10.1.3.20/24)
4 FS01 P2 Node2 (10.1.4.10/24) -> FS02 P2 Node2 (10.1.4.20/24)
I physically connected them so that the IP subnets aligned (The port on A - 10.1.1.10 - connected to the port on B - 10.1.1.20) and so forth
I was able to ping all IP addresses on FS02 from FS01
I created two portsets on both system, each containing two ports each on FS01 (P1N1+P1N2, P2N1+P2N2) and then the same on FS02
The partnership came up (I used two portsets - PBHA-PS01, PBHA-PS02 - in the partnership link) but immediately went to a "Degraded link" status.
This is because the target Discovery service tried to connect from both the ports on FS01 in PBHA-PS1 to the first address in FS02's PBHA-PS01.
This, of course fails since this is direct-connect and there is no switch/router in between. So, as soon as I removed one link pr. portset, the RDMA partnership turned green.
In any case, use FC cards for Direct-connect when implementing PBHA, no matter what You use for host connections (Ethernet or FC). Or, use a switch/switch-stack on each side and keep the ip-adresses in the same subnet.