Current Series Release Notes¶

9.1.0-8¶

Fixed Arista EOS L2VNI configuration to create the VLAN before configuring VXLAN mappings. Previously, the driver configured BGP EVPN and VXLAN interface mappings without creating the VLAN at the global configuration level. Without the VLAN, switches would not advertise IMET routes, remote VTEPs would not be discovered, and VXLAN connectivity would fail. See LP#2146511 for details.

Fixed Arista EOS L2VNI configuration to advertise locally learned MACs via BGP EVPN. Previously, the BGP EVPN VLAN configuration was missing the redistribute learned command, which prevented locally learned MAC addresses from being advertised as EVPN Type 2 routes to remote VTEPs. This caused complete VXLAN data plane failure between nodes on different leaf switches. See LP#2146559 for details.

Fixed Arista EOS route-target command syntax for BGP EVPN L2VNI configuration. The driver was generating route-target both auto which is invalid on Arista EOS. The correct syntax requires separate export and import directives with the AS number parameter: route-target export auto <ASN> and route-target import auto <ASN>. See LP#2146500 for details.

Fixed Arista EOS VLAN-to-VNI unplug command syntax. The driver was generating no vxlan vlan {segmentation_id} which is incomplete and rejected by Arista EOS with “% Incomplete command”. The correct syntax requires the VNI parameter: no vxlan vlan {segmentation_id} vni {vni}. This prevented VLAN-to-VNI mappings from being removed during network deletion, causing stale configurations to accumulate on switches. See LP#2146508 for details.

Fixed intermittent connectivity issues with Cisco NX-OS VXLAN deployments where baremetal nodes would not receive ARP replies. The NVE interface VNI member configuration now includes suppress-arp to enable proper ARP handling via EVPN Type-2 routes. See LP#2144910 for details.

Fixed L2VNI VLAN-to-VNI mapping cleanup to use Neutron segment state as the authoritative source for cleanup decisions. Previously, cleanup was based solely on switch port state, which prevented cleanup when static trunk configurations existed on switches. The new hybrid approach checks if the VLAN segment exists in Neutron first - if deleted, L2VNI mappings are removed unconditionally. If the segment still exists, per-switch port checking determines cleanup. This enables static trunk configurations and proper cleanup in multi-switch deployments. See LP#2144875 for details.