Lab: RED metrics and traces without code using Beyla

What you will learn in this TP

• Deploy Beyla alongside an application
• Read the RED metrics
• Visualise the traces

Introduction #

What you will learn in this section

• Deploy Beyla alongside an application
• Read the RED metrics
• Visualise the traces

Context. A third-party application exposes no metrics and cannot be modified. Yet you want its throughput, errors and latency.

Prerequisites: Docker Compose on a Linux host  ·  Estimated time: 45 min

Metrics and traces without a single line of code.

Hands-on #

Guided steps, with their solutions.

1. 1. Write the Compose file

   Why? Beyla shares the application's PID namespace (pid: service:app) so it can see it without being privileged across the whole host.

   Create a Compose file with a web application (nginx), Beyla, Prometheus, Tempo and Grafana.

2. 2. Generate traffic

   Why? Without requests, Beyla has nothing to measure.

   Start the stack and send requests, some of which return a 404 error.

3. 3. Read the RED metrics

   Why? The metrics follow the OpenTelemetry semantic conventions: they are identical whatever the application's language.

   Query Prometheus: total throughput, 4xx error rate, p95 latency.

4. 4. View the traces

   Why? Each request becomes a span: useful for tracking down a specific slow request.

   Add Tempo to Grafana (http://tempo:3200) and search for boutique-front traces in error.

Difficulty level : ●●●○○ (3/5)