← All roadmaps
Study Roadmap
Site Reliability Engineering
observability · SLOs · incident response · infra-as-code
A path into running systems that stay up: measure them honestly, set reliability targets you can defend, respond when they break, and manage the whole thing as code. This roadmap curates the field's best resources; the taught, from-first-principles versions of the deep material live in the Zero Series' Ubuntu Zero, Observability Zero, Kubernetes Zero and Automation Zero books, which the relevant stages link to as canonical homes.
12 Stages
0/51 Resources done
— Day streak
00
STAGE 00 · what SRE actually is
Foundations
you finish able to → the SRE mental model: error budgets, toil, and reliability as a feature
Book
1. Seeking SRE (Blank-Edelman, ed.)essays from across the industry — where the field is going
01
STAGE 01 · know the box
Linux & Systems Performance
you finish able to → the ability to find why a machine is slow, not just that it is
Web
Ubuntu Zero — the machine from zero
series canonical home for Linux fundamentals; start here if the terminal is new
Book
1. Systems Performance, 2e (Brendan Gregg)how the machine behaves under load, methodically
Book
2. BPF Performance Tools (Gregg)modern Linux observability at the kernel level
02
STAGE 02 · measure everything
Observability I · Metrics
you finish able to → a Prometheus + Grafana stack and PromQL you can write from memory
Book
1. Observability Engineering (Majors, Fong-Jones, Miranda)the three pillars and the case for high-cardinality events
Web
Observability Zero — metrics & Prometheus, taught
series canonical home for the observability curriculum (in progress)
03
STAGE 03 · the other three signals
Observability II · Logs, Traces & Profiles
you finish able to → a full LGTM stack — logs, traces, metrics, profiles — you wired yourself
Lab
Roll-your-own LGTM stack↗
Lab setup
Stack: the official Grafana OTel-LGTM image (Loki, Grafana, Tempo, Mimir + an OTel collector) run locally with podman — one container to explore all four signals.Instrument: point a sample app's OpenTelemetry SDK at the collector, then swap the collector for Grafana Alloy to see the production shape.Hardware: any laptop or a small homelab node; 4GB RAM is enough to start.Guardrail: generate load so there's real signal to explore, and set log/trace retention low locally so you don't fill the disk.
04
STAGE 04 · reliability as a number
SLOs & Error Budgets
you finish able to → defensible SLOs with burn-rate alerts that actually page correctly
Book
1. Implementing Service Level Objectives (Hidalgo)the definitive practical SLO book
Lab
SLO + burn-rate lab↗
Lab setup
Stack: the Prometheus/Grafana stack from Stage 02 plus a sample service, all local under podman .Define: pick an SLI (request success rate or latency), set an SLO, and generate multi-window burn-rate alerts with Sloth.Load & break: drive traffic with Grafana k6 (link) and inject failures so the burn-rate alert actually fires.Guardrail: an untested alert is not an alert — prove it pages before you trust it, and prove it stays quiet on healthy traffic.
05
STAGE 05 · when it breaks
Incident Response & On-call
you finish able to → a real on-call practice: roles, comms, and blameless postmortems
Book
1. Incident Management for Operations (Schnepp, Vidal, Bhatia)
Lab
Chaos & game-day lab↗
Lab setup
Software: a local k8s cluster (kind or k3s) plus Chaos Mesh (link) or LitmusChaos to inject pod/network/IO faults.Hardware: a homelab node or a laptop with 16GB+ RAM for the cluster.Run a game day: break something on purpose, practice the on-call response against your Stage 03 dashboards, then write a blameless postmortem.Guardrail: only ever run chaos experiments against a throwaway cluster — never anything shared or production.
06
STAGE 06 · declarative systems
Infrastructure as Code
you finish able to → environments you can destroy and recreate from a git repo
Book
1. Terraform: Up & Running (Brikman)the practical on-ramp to declarative infra
Book
2. Infrastructure as Code, 2e (Kief Morris)the patterns, tool-agnostic
07
STAGE 07 · the substrate SRE runs on
Containers & Orchestration
you finish able to → a GitOps-managed cluster you can reason about and roll back
Web
Kubernetes Zero — orchestration, Helm, operators & GitOps
series canonical home for the container-orchestration curriculum
Book
1. Kubernetes Up & Running (Hightower, Burns, Beda)
08
STAGE 08 · engineer the ops away
Automation & Toil Reduction
you finish able to → the reflex to replace a repeated task with code that does it
Book
1. Ansible for DevOps (Jeff Geerling)config management you can put to work the same day
Web
Automation Zero — Ansible, AWX & event-driven automation
series canonical home for the automation curriculum
09
STAGE 09 · the genuinely hard part
Distributed Systems & Scaling
you finish able to → the vocabulary to reason about failure, consistency, and overload
Book
1. Designing Data-Intensive Applications (Kleppmann)the modern classic on distributed data systems
Book
2. Database Reliability Engineering (Campbell & Majors)
10
STAGE 10 · SRE meets the model
Reliability for ML & Data
you finish able to → the ability to keep data and model systems reliable, not just up
Web
AI Ops Zero — model serving, evals & drift monitoring
series canonical home for running models reliably in production
11
STAGE 11 · keep it up and safe
Security & Reliability
you finish able to → systems that are dependable against failure and adversary alike