Question: How to handle dynamically changing retry (think api throttling)

[ravensorb]

I am curious if there is a way to handle dynamically change the retry wait period based on something that happens within the process being executed. The specific example I have is an API that has a rate limit on it and it returns in the HTTP headers the recommended amount of time to wait before retying the call.

Any thoughts or suggestions on how to approach?

[ahgraber]

I'm also interested in this!

[basilisk487]

you can use a custom wait callback:

def my_wait(retry_state: RetryCallState):
    ex = retry_state.outcome.exception()
    if isinstance(ex, requests.exceptions.HTTPError) and ex.response.status_code == 429:
        retry_after = ex.response.headers.get("Retry-After")
        try:
            return int(retry_after)
        except (TypeError, ValueError):
            pass
    return 0