Question about rdma_join_multicast() retransmission and timeout during SM failover

Infrastructure & Networking Software And Drivers Mellanox OFED
,
1

Hello,

I am using InfiniBand UD multicast and have a question regarding the behavior of rdma_join_multicast().
As far as I understand, this function internally sends a MAD packet to the Subnet Manager (SM). If the SM is not in the Master state (e.g., during failover transition), the SM would not be able to process the packet.

  • How is retransmission handled in this case?

  • Is there a configurable or fixed timeout for the operation?

  • In my testing, I consistently see a “multicast error” CM event about 3 seconds after calling rdma_join_multicast(). Is this the expected behavior?

Any clarification on the retransmission/timeout mechanism during SM failover would be greatly appreciated.

Thank you.