Topology is as follows;
Client ------ DUT [ex4400 VC] ---- aeX ------ Intermediate Device 1 ------- aeA --------- Device X -------- Client | | |---------------- aeY ------ Intermediate Device 1 ------- aeB ------------|
The above is EVPN/VXLAN with IPV6 UNDERLAY/OVERLAY.
ae bundle is configured with ports from two PICs (et-0/2/0 and et-2/2/0 for example).
UNDERLAY is configured between ae peers IPV6 address.
OVERLAY is configured between DUT and Device X loopback address.
Inter vlan (inter irb) traffic is sent across clients through EVPN/VXLAN.
When we do PIC restart of one of the port in ae bundle, the traffic egressing out is completely lost. When we do packet capture on the peer ingress (ae X on Intermediate Device 1) we see vlan tag getting added with vlan id 4095.
Need to reboot the device to recover.
Traffic loss will be observed if the packet having a VLAN tag of 4095 when egressing out is getting dropped after the PIC restart in the EVPN-VXLAN IPv6 underlay environment. The issue happens when the AE interface is a next-hop, once the PIC gets restarted few member interfaces go down, and once back up unable to handle the packets with a VLAN tag of 4095 due to a programming issue for PFE will not receive any new next-hop addition to PFE as the AE interface is already present in PFE as next-hop.
On EX4400, EX4100, EX4650, and QFX5120 platforms in the EVPN-VXLAN (Ethernet VPN - Virtual Extensible LAN) IPv6 (Internet Protocol Version 6) underlay environment, after the PIC (Physical Interface Card) restart few AE (Aggregated Ethernet) member interfaces go down and comes back up traffic loss is observed if the packet has a VLAN tag of 4095. The issue happens because PFE (Packet Forwarding Engine) will not receive any new next-hop addition to PFE as the AE interface is already present in PFE as next-hop. You can refer below PR for more details; https://prsearch.juniper.net/problemreport/PR1788573