Detailed explanation with the sample topology below:
- When ARP request is received on ISC port, and the switch does not have FDB entry of the sending host, the switch does not reply to the ARP request.
- Because of this behavior, the first ARP request from the host is not replied, and the host has to retransmit the ARP request.
- By default, FDB learning is disabled on the ISC port as per the MLAG implementation.
- With the topology above, When ARP is received in SW1 from SW3, the ARP will be flooded to ISC port in hardware and a copy will be lifted to SW1 CPU for learning.
- Note: ARP will be flooded in hardware faster than the FDB getting synced by software between the MLAG peers.
- SW2 will not update the ARP table because the corresponding FDB is not yet present in the switch. Hence the ARP entry is not getting updated on the first packet.
- Later FDB will be check pointed by SW1 and FDB will be programmed in SW2 as well.
- Now when ARP entry (second packet) is received on ISC port, SW2 will update the entry in ARP table.