4 ms·
A week ago I might have thought this would be good for our microservices, as they were already using short timeouts with a retry but abandoning the original req
by karmakaze 9y ago
A week ago I might have thought this would be good for our microservices, as they were already using short timeouts with a retry but abandoning the original request. This is better. However it has limitations and isn't the solution. It suffers from request amplification if the called service also depends on other services and applies the same retry policy. It it still request latency bound. The main takeaway I've learned for bounding microservices is that the inflow of dependent data should be asynchronous and request latency dependent only on the service's data.