Observability Mastery Path¶
Most "observability" tutorials show you a pretty Grafana dashboard and call it done, without ever explaining the one idea that makes the whole stack make sense: a metric, a log line, and a trace span are three different, lossy projections of the same underlying event, and the real skill is knowing which projection answers which question — and how to connect them back together when one alone isn't enough. This course builds that skill by running a real Prometheus server, a real Grafana instance, and real OpenTelemetry-instrumented code on an ordinary laptop, and deliberately breaking things to watch each tool actually earn its place.
This course was written and run against a real local Prometheus 3.x server, a real local Grafana instance, and real Python services instrumented with the OpenTelemetry SDK — no cloud observability vendor account, no SaaS trial, no Docker Compose stack of pre-built dashboards. Real PromQL queries against real scraped metrics, real alert rules that really fire, and a real trace reconstructed from real exported spans.
It assumes basic programming and command-line comfort, and benefits from — but doesn't require — the service-level instrumentation built in the Microservices Mastery Path, whose own tracing and structured-logging lessons this course builds on directly and links back to rather than repeating.
How the program is organized¶
| Level | Focus | Modules | What you can do after |
|---|---|---|---|
| 1 · Metrics & Prometheus Fundamentals | The Prometheus data model, instrumenting real metrics, PromQL, recording rules, the TSDB | 10 | Instrument a real app and query its metrics correctly |
| 2 · Dashboards, Alerting & Logs | Grafana dashboards, the RED/golden-signals/USE methods, Alertmanager, centralized logging | 10 | Build dashboards and alerts that actually page the right person |
| 3 · Distributed Tracing & OpenTelemetry | Spans, context propagation, the Collector, sampling, cardinality, correlating all three pillars | 10 | Trace a real request end to end and correlate it with metrics and logs |
| 4 · SLOs & Operating Observability at Scale | SLIs/SLOs/error budgets, burn-rate alerting, capacity planning, scaling Prometheus, on-call | 10 | Design and operate an SLO-driven observability stack for a real service |
Each level has 9 teaching modules plus a hands-on project (module 10), for 40 lessons total.
How to use this site¶
- Run it yourself. Download Prometheus and Grafana directly from their own sites, or install both with one
brew install prometheus grafanacommand — no account, no cloud bill, no managed observability platform needed for the core lessons. - Read the "what was run" note at the top of each lesson. The overwhelming majority of this course's command output, PromQL results, and dashboard screenshots come from a real, observed, local stack; anything that genuinely couldn't be run locally (a managed vendor's UI, a petabyte-scale long-term-storage backend) says so plainly.
- Do the "predict, then run" exercises. Prometheus has several behaviors (counter resets, histogram bucket math, PromQL's staleness handling) that are easy to get wrong from reading alone and genuinely clarifying to watch happen.
- Do the projects. Each one is a real, runnable instrumentation, dashboard, or alerting pipeline, not a screenshot.
More from the Mastery Path series¶
Free, structured, module-wise training across 97 other languages, platforms and disciplines — expand a category to browse: