Hi,
I am bit confused with EVPN MAC-IP Synchronisation and re-origination topic, can some one please clarify below queries
In the below evpn topology PE1, PE2 are All Active mode, Host H1 is multi-homed to PE1, PE2 nodes
CE device when hash the ARP Request to PE1, so PE1 create a local MAC-IP route and advertise to PE2. When PE2 receive the MAC-IP route from PE1 it identifies the ESI is matching to locally attached and synchronises the MAC-IP ROUTE to forwarding (EVPN_SYNC) on PE2 , locally inject/creates the synced MAC-IP Route and advertise/re-originate to all EVPN Peers.
PE1 will have local learnt path(best) and L2vpn path from PE2(next best), In ARP table for Host H1 only local flag will be set PE2 has synced path(best) and L2vpn path from PE1. In ARP table for Host H1 only EVPN_SYNC flag will be set. PE3 will have IP ecmp path for H1 with nexthops PE1, PE2.
I would like to understand the behaviour at each node in these below scenarios. I will try to illustrate the behaviour down please correct me if my understanding is wrong
- BVI interface down 2. Host is alive, ARP learning changed to PE2, 3. Host is permanently dead
- Now when BVI Interface goes down on PE1 so it initiates with draw of local mac-ip route and try to program the next best path as remote to forwarding(i.e from PE2) is this right ? PE2 receive UNREACH from PE1 and remove the EVPN_SYNC route and generate UNREACH. so PE3 will have no route to host H1 even though there is physical reachability to HOST H1 via PE2 . TIll PE2 learn the HOST H1 locally it will not have the path. Is this what happens on the cisco nod ?
- When Host is alive but ARP age out happen for host H1 on PE1 and learning on PE2. what will be the behaviour at PE1? DOES PE1 Generate UNREACH in this case? and program the next best path to forwarding as EVPN_SYNC and re-originate the mac-ip route again ? or it avoids UNREACH and only generate REACH.? On PE2 it would have learnt the dynamic arp entry locally from CE device and when it may receives UNREACH message from PE1 now it cleanup synced flag in arp table and process the following REACH message again from PE1. So PE3 has would have process UNREACH from PE1 and LATER from PE1?
- When Host is dead and ARP age out happen for host H1 on PE1 and NO learning on PE2. what will be the behaviour at PE1? DOES PE1 Generate UNREACH in this case? and program the next best path to forwarding as EVPN_SYNC and re-originate the mac-ip route again ? or it avoids UNREACH and and generate only generate REACH? On PE2 when it receives UNREACH message from PE1 now it cleanup synced flag in arp table and process the REACH message again from PE1. So PE3 has would have process UNREACH from PE1 and again REACH from PE1 and PE3 will have ecmp path for the host H1 from PE1, PE2(both are EVPN_SYNC re-originated paths) which is dead and it can cause permanent traffic drop right ?
