Hello
I would like to establish RoCEv2 communication between ServerA (192.168.10.10) and ServerC (192.168.20.10), which are on different subnets, using the following network configuration.
All servers (ServerA, ServerB, and ServerC) are using Mellanox ConnectX-6 Dx (2-port) network adapters.
ServerA (192.168.10.10/24) —— (192.168.10.20/24) ServerB (192.168.20.20/24) —— ServerC (192.168.20.10/24)
The following directly connected sections can successfully communicate using rping:
ServerA (192.168.10.10) and ServerB (192.168.10.20)
ServerB (192.168.20.20) and ServerC (192.168.20.10)
However, rping fails in the following section between different subnets, even though routing is configured and ping works successfully:
ServerA (192.168.10.10) and ServerC (192.168.20.10)
rping Execution Results Between ServerA and ServerC:
Receiver (ServerA)
$ rping -s -a 192.168.10.10 -C 2 -d mlx5_1 -v
verbose
created cm_id 0x557532db48c0
rdma_bind_addr successful
rdma_listen
Sender (ServerC)
$ rping -c -a 192.168.10.10 -C 2 -d mlx5_1 -v
verbose
created cm_id 0x5627ed413f40
cma_event type RDMA_CM_EVENT_ADDR_RESOLVED cma_id 0x5627ed413f40 (parent)
cma_event type RDMA_CM_EVENT_ROUTE_RESOLVED cma_id 0x5627ed413f40 (parent)
rdma_resolve_addr - rdma_resolve_route successful
created pd 0x5627ed4142d0
created channel 0x5627ed410630
created cq 0x5627ed414330
created qp 0x5627ed417998
rping_setup_buffers called on cb 0x5627ed410830
allocated & registered buffers…
cq_thread started.
I would appreciate it if you could provide a solution to resolve this issue.