Host routed overlay with deterministic host learning and localized integrated routing and bridging

1. A system comprising:

a plurality of bare metal servers;

a plurality of leaf nodes within a leaf-spine network topology;

a host routed overlay comprising at least one host virtual machine;

a leaf-server link established between a first bare metal server of the plurality of bare metal servers and a first leaf node of the plurality of leaf nodes; and

a routed backup link established between two leaf nodes of the plurality of leaf nodes;

wherein the routed backup link is pre-programmed as a backup failure path for overlay host routes facing the first bare metal server when the leaf-server link experiences a failure; and

wherein a plurality of virtual customer edge routers corresponding with the plurality of bare metal servers are configured with a same anycast gateway media access control (MAC) address to enable host virtual machine mobility across the host routed overlay.

2. The system of claim 1 , wherein the routed backup link is activated on the leaf-server link without use of a prefix associated with a virtual local area network (VLAN) encapsulation.

3. The system of claim 1 , wherein the plurality of virtual customer edge routers removes the leaf-server link from an equal-cost multipath routing path-set when the leaf-server link experiences the failure.

4. The system of claim 1 , wherein an ethernet virtual private network (EVPN) withdraw mechanism is extended to an Internet Protocol (IP) host route to enable the routed backup link to be activated on the leaf-server link without use of a prefix.

5. The system of claim 1 , further comprising a plurality of distributed anycast routers within a redundancy group, wherein an edge side includes (ESI) construct is associated with a set of layer-3 links within the redundancy group.

6. The system of claim 5 , wherein local reachability for the ESI construct is advertised via per-ESI ethernet access direct (EAD) to route traffic to one or more remote leaf nodes outside the redundancy group.

7. The system of claim 6 , wherein a forwarding indirection is established at the one or more remote leaf nodes via the routed backup link to enable convergence from the first leaf node following a failure of the ESI construct.

8. The system of claim 1 , wherein the routed backup link is configured to route traffic across layer-3 equal-cost multi-path (ECMP) links to an upstream leaf node within the leaf-spine network topology rather than being hashed across a layer-2 link aggregation (LAG).

9. The system of claim 8 , wherein the upstream leaf node serves as a layer-3 router that is free of any layer-2 bridging or integrated routing and bridging (IRB) functionalities.

10. The system of claim 1 , wherein the plurality of leaf nodes do not install any host MAC routes.

11. The system of claim 1 , wherein each of the plurality of bare metal servers comprises a virtual customer edge router and a host virtual machine; and

wherein the system further comprises a routed uplink from a first virtual customer edge router on the first bare metal server to one or more of the plurality of leaf nodes.

12. The system of claim 11 , further comprising a distributed anycast router, wherein the first virtual customer edge router is configured to enact host route learning between the first virtual customer edge router and the distributed anycast router; and

wherein the first virtual customer edge router provides localized integrated routing and bridging (IRB) for a first host virtual machine on the first bare metal server.

13. The system of claim 12 , wherein the first virtual customer edge router is in communication with the first host virtual machine on the first bare metal server, and wherein the first virtual customer edge router comprises memory for storing adjacencies for the first host virtual machine that is local to the first bare metal server.

14. The system of claim 1 , wherein the leaf-spine network topology comprises:

the plurality of leaf nodes; and

a plurality of spine nodes;

wherein each of the plurality of spine nodes comprises a link to each of the plurality of leaf nodes.

15. The system of claim 1 , wherein the first bare metal server comprises a first virtual customer edge router;

wherein the first virtual customer edge router comprises a plurality of subnets; and

wherein the leaf-server link between the first bare metal server and the first leaf node is established with a physical connection between the first leaf node and the first virtual customer edge router.

16. The system of claim 15 , wherein the leaf-server link experiences the failure when the physical connection between the first leaf node and the first virtual customer edge becomes inactive.

17. The system of claim 1 , wherein the first bare metal server comprises a first virtual customer edge router;

wherein the leaf-server link is established between the first leaf node and the first virtual customer edge router; and

wherein the first virtual customer edge router comprises memory for storing one or more of: local Internet Protocol (IP) entries for the host routed overlay, media access control (MAC) entries for the host routed overlay, or a default ECMP route to the host routed overlay.

18. The system of claim 1 , wherein the host routed overlay performs host routing for stretched subnets; and

wherein the host routed overlay is an Ethernet virtual private network (EVPN) host.

19. The system of claim 17 , wherein the first virtual customer edge router is a first hop gateway for the first leaf node prior to failure of the leaf-server link.