Mastering Distributed Tracing

Observability via a service mesh

If we place a service mesh in the path of every network request between the services in an application, it makes it an ideal place to collect consistent, standardized telemetry about the application. This standardized observability alone may be enough to compensate for a small performance loss introduced by going through a proxy for every request because it dramatically enhances the operational properties of the system, making it easier to monitor and troubleshoot:

• The sidecar can emit uniformly named metrics about the traffic going in and out of a service instance, such as throughput, latency, and error rates (also known as the RED method (Rate, Error, Duration). These metrics can be used to monitor the health ...