The customer applied the following configuration to build EVPN AS topology with IRB/VRRP virtual gateway:
MX204 Gateway 1:
user@router1> show configuration | match 667 | display set
set interfaces ae2 unit 667 description "EVPN - TEST VLAN 667 - with virtual-gateway-v4-mac"
set interfaces ae2 unit 667 encapsulation vlan-bridge
set interfaces ae2 unit 667 vlan-id 667
set interfaces ae2 unit 667 esi 00:94:43:05:00:00:00:00:06:67
set interfaces ae2 unit 667 esi single-active
set interfaces irb unit 667 virtual-gateway-accept-data
set interfaces irb unit 667 description "INET - TEST VLAN 667 - with virtual-gateway-v4-mac"
set interfaces irb unit 667 family inet rpf-check
set interfaces irb unit 667 family inet address 10.66.67.2/29 preferred
set interfaces irb unit 667 family inet address 10.66.67.2/29 virtual-gateway-address 10.66.67.1
set interfaces irb unit 667 virtual-gateway-v4-mac 00:50:00:44:06:67
set routing-instances EVPN-TEST protocols evpn extended-vlan-list 667-668
set routing-instances EVPN-TEST bridge-domains TEST-VLAN-667 description "INET - TEST VLAN 667"
set routing-instances EVPN-TEST bridge-domains TEST-VLAN-667 vlan-id 667
set routing-instances EVPN-TEST bridge-domains TEST-VLAN-667 interface ae2.667
set routing-instances EVPN-TEST bridge-domains TEST-VLAN-667 routing-interface irb.667
MX204 Gateway 2:
user@router2> show configuration | match 667 | display set
set interfaces irb unit 667 family inet address 10.66.67.3/29 preferred
set interfaces irb unit 667 family inet address 10.66.67.3/29 virtual-gateway-address 10.66.67.1
The customer has tried to disable the active link to simulate the cable issue, and the downstream switch will notice the issue since LACP is configured between EVPN routers and downstream devices. The downstream MAC table will also be updated and point to correct MX device.
user@router1# show | compare
[edit interfaces ae2]
- disable;
When the customer re-enable the router1 ae2 link, it becomes the active-link as expected, and VLAN 667 (example) status in router1 will be in forwarding mode, however, the VLAN 667 status in router2 will be in block mode, this is also expected due to the EVPN Active-Standby configuration. However, two downstream switch devices does not aware this kind of change, the downstream devices only notice router1 ae2 link is up due to LACP protocol, but they do not know the active link switches to router1, there is no protocol or message available to tell them what is happening, there is no other L2 protocol enabled by default. So there is a legacy MAC entry pointed to router2, which caused packet dropped.The issue will be resolved after the MAC expired and the new MAC address re-learnt from the active EVPN PE router.
In a standard design, the EVPN multi-homing peering PE routers are supposing to connect to a single CE device, however, there are two CE switches in the customer design and there is no other L2 protocol running between EVPN MX routers and downstream switches, except LACP protocol.
So it is an expected behavior due to the EVPN Active Standby technology limitation.
The customer may try the EVPN-VPWS to see whether it is working for this scenario.