Curriculum Spec & Implementation Plan

Observability Mastery — 100 Days, OSS-First

A hands-on program to take an experienced Staff engineer (strong on Node.js, AWS, Kafka, Kubernetes) to deep, employable observability expertise targeting Grafana Labs, Chronosphere, Honeycomb, Datadog and platform-engineering roles. The bias is 80% hands-on / 20% theory, every lab ships as its own GitHub repo, and every tool is open source from the Grafana LGTM stack, OpenTelemetry, Prometheus, Cilium and k6 — culminating in a production-grade, fully-instrumented distributed system survived against a chaos suite.

100
days, day01 → day100
11
phases / repos
4
certs: PCA·OTCA·CKA·CCA
4
signals correlated

Read me firstWhat this document is

Goals & non-goals scope contract

This is the spec / acceptance contract for the curriculum, not the day-by-day timeline (that artifact lives separately). It defines what "done" means at every phase gate, how the certs sequence, and what the capstone must survive. Read it as the rubric a hiring panel — or a future you — would grade the program against.

Goals

  • Reach production-operator depth on the full Grafana LGTM + OpenTelemetry + eBPF stack, not tutorial-level familiarity.
  • Ship one self-contained GitHub repo per lab — runnable by a stranger via README + docker compose up or make.
  • Pass 4 certifications in priority order (PCA → OTCA → CKA → CCA) as forcing functions, not as the goal itself.
  • Be able to debug a real incident using telemetry alone — no SSH, no print statements, no guessing.
  • Build interview-grade artifacts and OSS contributions that demonstrate the work publicly.

Non-goals

  • Not a survey of commercial SaaS consoles (Datadog/New Relic UI clickthroughs); concepts transfer, vendor UIs do not.
  • Not re-teaching Node.js, Kafka, Kubernetes, or AWS fundamentals — those are assumed.
  • Not a frontend/RUM deep-dive; browser telemetry appears only where the capstone touches it.
  • Not chasing every signal vendor — depth on OSS reference implementations over breadth.
  • Not a certification-cram; certs validate the hands-on work, they do not replace it.

The work, phase by phase11 phases · one repo each

Forcing functionsCertification plan

Certification plan 4 exams, priority-ordered

Certs are sequenced so each exam lands at the end of the phase that prepares it, with a buffer day. They validate the lab work rather than driving it.

PriorityCertFull name~Exam dayPrepared by phase
1PCAPrometheus Certified Associate~Day 20P2 — Metrics (Prometheus & Mimir)
2OTCAOpenTelemetry Certified Associate~Day 55P5 — OpenTelemetry & Collector
3CKACertified Kubernetes Administratorafter P6 windowP6 — Kubernetes Observability
4CCACilium Certified Associateafter P8 windowP8 — eBPF & zero-code (Beyla/Cilium)

PCA and OTCA are tightly time-boxed inside their phases. CKA and CCA are broader and may schedule after Day 100 — see Risk R1.

Day 100Capstone acceptance criteria

Production capstone — acceptance criteria repo: day100-production-capstone

A realistic e-commerce flow: Frontend → API Gateway → Order → Inventory → Payment → Kafka → Workers → PostgreSQL → Redis, instrumented with all four signals plus Beyla eBPF, deployed up a progressive target ladder, load-tested with k6, paged via OnCall, and governed by SLOs. The capstone is accepted only when every box below is true.

Deploy target ladder
Docker Compose→ Kind→ K8s (self-managed)→ AWS EKS
Platform & instrumentation
    Chaos suite — each fault debugged with telemetry only

      Surfaced, not buriedRisks & open decisions

      Risks & open decisions 5 entries, each with a recommendation

      The barSuccess metrics after 100 days

      Success metrics after 100 days how we grade it