How Can OpenStack VM Live Migration Work Seamlessly in EVPN-VXLAN Fabrics with SONiC?

The blog outlines how OpenStack live migration can work with minimal disruption in EVPN-VXLAN fabrics by focusing on consistent SONiC configuration, EVPN MAC/IP route updates, and automated pre-production validation. For IT and security leaders, the networking dependency affects session continuity, uptime, and operational risk during VM mobility.

Research Overview

OpenStack live migration moves a running VM between physical servers with minimal service interruption. In EVPN-VXLAN data center fabrics, the blog frames it as a network requirement as well as a virtualization feature, because the fabric needs to preserve the VM’s IP, MAC, gateway reachability, and traffic path.

The post targets operators using SONiC-based open networking and describes how SONiC EVPN-VXLAN supports seamless VM mobility. It also discusses how FTAS is used to validate the fabric before live migration is used at scale.

Key Findings

The blog states that seamless migration depends on alignment of EVI, VNI, gateway, and MTU settings across the EVPN-VXLAN fabric. If the source and destination compute nodes are not in the same EVPN instance and VNI, the VM cannot retain the same Layer-2 behavior after migration.

It also describes that the destination leaf detects the VM MAC on a new server-facing port, advertises an updated EVPN Type-2 MAC/IP route, and that Neutron or OVS triggers Gratuitous ARP or Neighbor Advertisement to refresh ARP and ND tables. For validation, the blog says Aviz Fabric Test Automation Suite covers live migration under load and during failures, including checks for route movement, withdrawal, blackholing, and control-plane stability.

Technical Breakdown

The blog says consistent VLAN-to-VNI mappings are required across SONiC ToR leaf switches and that all leaves must participate in the same EVPN instance. It adds that Anycast Gateway should be configured consistently so the VM continues using the same default gateway before, during, and after migration.

For packet handling, the post states that VTEPs must support VXLAN encapsulation and that the underlay MTU must account for VXLAN overhead to avoid fragmentation or packet loss. During mobility, it explains that EVPN tracks MAC/IP location, then redirects traffic after the VM moves by updating EVPN Type-2 MAC/IP routing information.

Operational Impact

The blog ties live migration reliability to coordination across OpenStack and the physical network, including Nova, Neutron, OVS/OVN, and SONiC EVPN-VXLAN components. It describes FTAS as validating production readiness for live migration at scale rather than only confirming behavior in ideal conditions.

FTAS validation scenarios named in the blog include basic live migration with shared storage, migration under CPU and memory load, Layer-2 and Layer-3 continuity, concurrent migrations, failure handling, and fabric-level EVPN/VXLAN validation. The post says FTAS confirms MAC/IP route movement to the new VTEP, prompt withdrawal of old routes, absence of traffic blackholing, and stability of the control plane during mobility events.

Overall, the blog presents live migration in EVPN-VXLAN as a system-wide networking problem that relies on consistent SONiC and EVPN/VXLAN configuration, EVPN Type-2 MAC/IP route updates, gateway consistency, and ARP/ND refresh behavior, validated via FTAS before broad rollout. This Blog Signals brief is a fact-based summary of the vendor blog.

Full content from aviznetworks.com.