Question 210
A high-throughput TCP transfer appears to use a single uplink despite four equal-cost paths, and an operator declares ECMP broken without checking flow-based hashing. Once the issue is resolved, which ongoing practice BEST prevents this type of failure?
Explanation
Correct Answer
This control reduces recurrence of the same class of failure: Common ECMP forwarding implementations use stable flow hashing to select paths, so one flow need not balance packet-by-packet across every next hop. Apply the recurring practice: Evaluate ECMP using diverse realistic flows and per-path counters rather than a single stream. The previous example was: A high-throughput TCP transfer appears to use a single uplink despite four equal-cost paths, and an operator declares ECMP broken without checking flow-based hashing.
Incorrect Answers
Document loopback allocation and verify its route advertisement during device replacement and uplink maintenance. That concerns VTEP loopback source identity stability, while the issue here is ECMP hash path versus per-packet round robin. Test several distinct source/destination/transport tuples, inspect effective ECMP next hops and compare aggregate usage before claiming imbalance.
EVPN designated-forwarder selection controls eligible BUM forwarding toward a shared Ethernet segment and helps avoid duplicate delivery. For this issue, test several distinct source/destination/transport tuples, inspect effective ECMP next hops and compare aggregate usage before claiming imbalance.
Include controlled link-failure ECMP convergence tests in rack commissioning and uplink changes. Here, common ECMP forwarding implementations use stable flow hashing to select paths, so one flow need not balance packet-by-packet across every next hop.