3 ms·
I'm suspicious of load shedding not mentioned in the article. Combine that with exp backoff in the caller and you got yourself a pretty robust starting point
by maxchisto 15d ago
I'm suspicious of load shedding not mentioned in the article. Combine that with exp backoff in the caller and you got yourself a pretty robust starting point
- otterley 15d ago429 (and sometimes 503) errors returned by servers might well be a symptom of intentional load shedding. Perhaps it's just not explicitly called out as a server behavior that induces client retries.
- hxtk 15d agoI’d like to mention it since Uber has a really cool load shedder [1], also implemented similarly by Netflix [2] and failsafe-go [3]. It basically looks for points where more requests per second suddenly cause a significant increase in latency and calls that the concurrency limit. 1: https://www.uber.com/us/en/blog/cinnamon-using-century-old-tech-to-build-a-mean-load-shedder/ https://www.uber.com/us/en/blog/cinnamon-using-century-old-t... 2: https://github.com/Netflix/concurrency-limits https://github.com/Netflix/concurrency-limits 3: https://failsafe-go.dev/adaptive-limiter/ https://failsafe-go.dev/adaptive-limiter/