Notes on Microservice Networking

This is a placeholder post used to preview the blog UI. Replace it with real content.

The first time a microservice system fails in production, it’s rarely the code. It’s the network between the services — the timeouts that were never set, the retries that amplified a small blip into an outage, the service discovery that pointed at a dead instance for thirty seconds too long.

Set timeouts everywhere

A request with no timeout is a request that can hang forever. Every hop should have a budget, and the budgets should add up to less than the timeout of the call that started the chain. Otherwise you get cascading failures that are miserable to debug.

Retries need backoff and jitter

Naive retries are how you turn a brief hiccup into a thundering herd. Exponential backoff with jitter spreads the load out so your recovering service isn’t immediately knocked over again.

Make failure visible

Dashboards for latency percentiles, error rates, and saturation will tell you more than any amount of log spelunking after the fact. Instrument first, debug second.