Symptom

Broadcast, unknown unicast and multicast (BUM) traffic are dropped at the border gateway (BGW) in some VLANs as it arrives at a site on the BGW device. This will result in failure of both the data plane and control plane traffic such as ARP and HSRP in one or more VLANs between sites.

Conditions

*9500 end of rack platform [EoR] cloudscale line card. *NXOS 9.2.3 or earlier *Multisite VLXAN with multicast replication in each site and ingress replication between data center interconnects (DCI). *Fabric DCI and outgoing interface (OIF) on different line cards. *Can be triggered by any one of the followiing: 1. Line card (LC) reload. 2. DCI or fabric interface changes. (add/remove/flap DCI links or fabric links,)

Workaround

One of the following will temporarily recover the condition on the BGW: 1. Remove and re-add VLAN. 2. Remove and reconfigure (re-add) the VNI under the NVE interface. 3. Bounce (shut/no shut) the NVE interface. Note that these workarounds do not prevent the issue from occurring again if any of the triggers are invoked. Issue resolved in Cisco NX-OS 9.3.2 and later.

Further Problem Description