Problem:

I have a client app in Java that needs to adhere to multiple rate limits. For example, req/second, req/minute, req/hour…etc. The app will only permit a call if the app is within all of the limits.

When I implement the rate limiters using either Resilence4J or Google Guava, it ends up consuming permits for all rate limits that returned true, even though at least one of the rate limits returned false.

Example (assuming Guava):

If I have an REQ_PER_SECOND RateLimit and a REQ_PER_MINUTE RateLimit defined, it’s possible that REQ_PER_SECOND.tryAcquire() returns true, but REQ_PER_MINUTE.tryAcquire() returns false.

In this case, an API call will not be made, but it still ended up consuming permits for REQ_PER_SECOND.

This results in the app throttling itself all the time.

Note: we can’t just throw this problem to the server due to cost and performance reasons.

Question

Is there a way to do this with Guava or Resilience4J libraries? If not, what’s the best way to handle this problem?

Guava has a setRate(int rate) method, but I think that only sets the stable rate, but does not actually reset the rates already consumed. So that wouldn’t work.

Any suggestion is greatly appreciated!