Telemetry for Free
Because every request crosses a proxy, the mesh sees all traffic. That vantage point gives you the three pillars of observability without instrumenting application code.
The Three Pillars
- Metrics: request rate, error rate, and latency, often called the golden signals, per service and per route.
- Logs: structured access logs for each connection.
- Traces: spans that follow a request across services.
The proxy records these uniformly, so a Python service and a Rust service produce identical telemetry.
Distributed Tracing Caveat
Tracing needs one small cooperation from the app. The proxy can start a trace, but to link spans across hops the application must forward the trace headers it receives. Without that, each hop looks like an unrelated request.
Why It Matters
Uniform, automatic telemetry means you can build a service dependency map, spot a slow downstream, and alert on error spikes without asking every team to add libraries. The mesh turns observability into a platform feature.
Key idea
Every request passing through a proxy lets the mesh emit uniform metrics, logs, and traces automatically, though apps must forward trace headers to link spans across hops.