ACE - L3 - Specialist Data Center Engineering Free Sample Questions

For environments that strictly require massive Layer 2 domains across all racks without overlay technologies, a Layer 2 Leaf-Spine (L2LS) architecture is required. In Arista L2LS designs, spines are typically configured as an MLAG pair to provide active-active redundant Layer 2 paths from the leafs, preventing STP blocking while extending the broadcast domain. Pure L3LS without an overlay cannot provide L2 adjacency. While EVPN-VXLAN could provide this, an L2LS fabric is the direct, traditional answer for native L2 scaling without overlay complexity.

In an L2LS architecture where routing is centralized at the spine layer, Switch Virtual Interfaces (SVIs) must be used. SVIs (configured as 'interface Vlan X') provide the Layer 3 gateway for hosts within a Layer 2 broadcast domain that spans multiple physical links and switches. Routed sub-interfaces are only applicable to point-to-point Layer 3 links and cannot act as a distributed gateway for a trunked VLAN spanning multiple leaf switches.

In an Arista L2LS architecture using MLAG at the spine layer, the MLAG spine pair should act as a single logical unit and MUST be the STP root for all VLANs. Best practice dictates setting the STP priority of both MLAG spine switches to 0 (or the lowest possible value) to ensure they always win the root election. Leaf switches should be left at the default priority (32768) or higher to prevent them from ever becoming the root bridge, which would cause suboptimal traffic flow and potential outages during topology changes.

LACP load balancing relies on a hashing algorithm to pin flows to specific physical links. If traffic consists of 'elephant flows' (large, long-lived streams), a simple MAC or IP hash might map all traffic to one link. Including Layer 4 port information increases the entropy of the hash, allowing different sessions between the same IPs to be distributed across multiple links. Additionally, enabling dynamic load balancing (if supported by the platform/profile) can rebalance flowlets.

In an Arista MLAG environment, if the peer-link goes down but the heartbeat remains up, the system recognizes a split-brain scenario. To prevent loops and MAC flapping, the MLAG secondary switch will automatically place all of its MLAG member ports into an errdisable (down) state. Orphan ports (ports not part of an MLAG) will remain up and active. If the heartbeat also fails, both switches would assume the primary role, which is why the heartbeat is critical.

False. Unlike VRRP, which relies on hello packets and an election process to determine an active Master router, VARP (Virtual ARP) is completely stateless. There is no control plane protocol or election process. All switches configured with the same VARP virtual MAC and virtual IP address will actively route traffic destined to that MAC address, providing an inherent active-active gateway without protocol overhead.