API Gateway Down

API Gateway Down#

The API Gateway sits in front of the rate limiter. If it goes down, requests never reach the rate limiter at all.

Impact#

User → API Gateway (down) → ✗

Users see connection errors or timeouts. The rate limiter, Redis, Rule DB — all completely unaffected and healthy. But no traffic reaches them because the entry point is gone.

This is a full user-facing outage but it is outside the rate limiter's scope to solve. The API Gateway is its own system with its own redundancy requirements.

API Gateway Redundancy (Not Rate Limiter's Problem)#

In a production system, the API Gateway is never a single instance. It runs as multiple instances behind a load balancer or as a managed cloud service (AWS API Gateway, Kong, Nginx cluster):

Internet → Load Balancer → API Gateway Instance 1
                         → API Gateway Instance 2
                         → API Gateway Instance 3

One instance going down routes to others. The LB health checks detect failure within seconds.

The rate limiter does not need to handle API Gateway failures — that is the API Gateway's own fault isolation responsibility.

Summary#

Impact on rate limiter : none — rate limiter never receives traffic
Impact on users        : full outage — connection errors
Responsibility         : API Gateway team, not rate limiter
Solution               : multiple API Gateway instances behind LB
                         managed cloud API GW (inherently redundant)