Data Center Design, Routing, and Troubleshooting
Core concepts
Design starts with requirements: bandwidth, oversubscription, failure domains, and scale. Know how to size a spine-leaf fabric (spine count for bandwidth, leaf count for ports, ECMP paths = number of spines). Understand routing protocols: eBGP is the preferred underlay for large fabrics (scalable, multipath, policy-rich), with OSPF used in smaller designs. Know BGP best-path basics and how ECMP is enabled (maximum-paths).
Concepts to master
MLAG design (peer link, peer-keepalive, dual-homed servers), VXLAN control plane options (multicast vs EVPN), and EVPN route types (Type 2 MAC/IP, Type 3 IMET, Type 5 IP prefix). Troubleshooting methodology: verify physical, then L2, then L3, then overlay. Use show ip bgp summary, show ip route, show vxlan address-table, and ping/traceroute.
Traps
Questions often assume OSPF scales to huge fabrics (eBGP is preferred at scale). Watch for EVPN route-type mix-ups: Type 2 is MAC/IP advertisement, Type 3 is multicast/IMET, Type 5 is IP prefix. Another trap: forgetting that ECMP requires equal-cost paths and maximum-paths configured.
Worked example
A host on leaf A cannot reach a host on leaf B in a VXLAN/EVPN fabric. Check: (1) underlay BGP neighbors up on both leafs and spines; (2) VTEP loopbacks reachable; (3) show vxlan address-table shows the remote MAC learned via EVPN Type 2; (4) MTU end-to-end. If the MAC is missing, the EVPN control plane or VNI mapping is the fault, not the underlay.