A rate limiter runs as 10 replicas, and each keeps its own counter with a limit of 100 requests a minute per client. What happens?

Options

• ✗ The limit works exactly as intended
• ✓ A client spread across replicas can get up to 1,000 requests a minute Correct
• ✗ Each client is limited to 10 a minute
• ✗ Requests are rejected at random

Why

Each replica sees only its share of a client's traffic. Keep the counter in a shared store and use an atomic increment (such as Redis INCR) before the decision, so two replicas cannot both admit the last request.

Review this in your daily deck → All cards in Resilience