Q1: During a deployment, NLB traffic splits 64%/36% between two cells despite identical healthy target counts. This is consistent across ALL callers. What level is the imbalance occurring at?
- Client-side connection affinity
- NLB-level traffic distribution behavior
- DNS round-robin imbalance
- Target group health check timing
Show Answer
B — Since the imbalance is consistent across all callers, it's NLB-level behavior rather than client-specific. NLB distribution can be uneven even with equal healthy targets.
Q2: The traffic imbalance causes per-host TPS to exceed rated capacity (3,649 vs 3,500 limit), triggering auto-rollback. What's the operational impact?
- Minor — just retry the deployment
- Significant — requires ~709 extra hosts across regions to handle worst-case cell imbalances
- None — auto-rollback protects the service
- Only affects the specific deployment
Show Answer
B — If NLB imbalance is unpredictable, teams must over-provision to handle worst-case scenarios, significantly increasing infrastructure costs across all regions.
Q3: What DNS-level solution could provide more deterministic traffic distribution across NLBs?
- Lower the DNS TTL
- Use weighted round-robin DNS records across multiple NLBs instead of a single 100%-weight record
- Enable DNSSEC
- Use geolocation routing
Show Answer
B — Weighted round-robin across multiple NLBs gives DNS-level control over traffic distribution, providing more predictable splitting than relying on a single NLB's internal distribution.