8.2 Rate Limiter Failures and Multiple Regions

Every request to the API in the previous lesson passes through one limiter that allows 100 requests per minute per API key. A sliding window counter in a Lua script makes the check atomic, and on Redis 8.8 and later INCREX does the same for a fixed window in one command. That design is correct while one healthy Redis deployment serves one region.

This lesson covers what happens outside that case. It decides the policy when Redis is slow or unavailable, gives the limit a meaning when one client sends requests to several data centers, and ends with sharding, hot keys and the signals to monitor