Platform engineering and SRE cheat sheet (Site reliability)

1. 1Reliability basics: SLIs, SLOs and error budgetsHow to say how reliable a service should be in numbers users would recognise, and how to turn the gap between perfect and good enough into a budget you can spend.7 concepts

   1. 1SLI (service level indicator)A measurement of how a service is behaving from the user's side, best written as a ratio: good events divided by valid events, such as successful requests over…
   2. 2SLO (service level objective)A target for an SLI over a rolling window, for example 99.9 per cent of checkout requests succeed over 30 days.
   3. 3SLA (service level agreement)A contract with customers that states a level of service and what happens, usually service credits, if it is missed.
   4. 4Availability nines and allowed downtimeEach extra nine cuts allowed failure by ten times.
   5. 5Error budgetThe share of events allowed to fail within the SLO window: 1 minus the SLO.
   6. 6Error budget policyA short written agreement, signed by engineering and product, that says what happens when the budget runs low or runs out, such as freezing non-urgent releases…
   7. 7Composite availabilityWhen a request needs several components in series, their availabilities multiply.

   7 in this chapterOpen this chapter
2. 2Observability: metrics, logs and tracesGoing past a first dashboard: which signals to collect, how Prometheus stores them, PromQL you use every week, cardinality, the OpenTelemetry Collector and sampling.9 concepts

   1. 1Four golden signalsLatency, traffic, errors and saturation: the four measurements Google's SRE book says to watch first on any user-facing system.
   2. 2RED and USE methodsRED (rate, errors, duration) is a checklist for request-driven services.
   3. 3Counter, gauge and histogramA counter only goes up (requests served).
   4. 4PromQL basicsPrometheus's query language. rate() turns a counter into per-second change, sum by (label) aggregates across instances and histogram_quantile() reads a…
   5. 5Recording rulesPrometheus rules that evaluate a query on a schedule and store the result as a new, cheaper series, such as a 5 minute error ratio per service.
   6. 6CardinalityThe number of unique time series a metric produces: every distinct combination of label values is a separate series stored in memory.
   7. 7OpenTelemetry CollectorA vendor-neutral service that receives telemetry, processes it (batching, filtering, adding attributes, sampling) and exports it to one or more backends…
   8. 8Head and tail samplingHead sampling decides whether to keep a trace when it starts, for example keep 10 per cent.
   9. 9Correlating logs, metrics and tracesStamping every log line with the trace and span ID, and attaching example trace IDs (exemplars) to metrics, so you can jump from a spike on a graph to a trace…

   9 in this chapterOpen this chapter
3. Incident response and postmortemsAlerts that fire for the right reasons, the roles that keep an incident calm, mitigating before diagnosing, and postmortems that change something.8 conceptsOpens with credits
4. Kubernetes in productionWhat changes when a cluster runs real traffic: a troubleshooting flow, failure states, disruption budgets, spreading across zones, graceful shutdown, upgrades and guard rails.8 conceptsOpens with credits
5. Infrastructure as code at scaleTerraform beyond one folder: splitting state, versioned modules, plans in CI, policy as code, drift, and control planes that let teams request infrastructure themselves.6 conceptsOpens with credits
6. GitOps and progressive deliveryGit as the source of truth for what runs, an agent that reconciles the cluster to it, promotion between environments, and releasing gradually with automatic checks.7 conceptsOpens with credits
7. Internal developer platformsTreating the platform as a product for internal developers: golden paths, a software catalogue, templates, docs and the measures that show whether it helps.8 conceptsOpens with credits
8. Capacity, performance and costThe few sums behind capacity planning, why latency jumps near full utilisation, load testing, autoscaling at each layer and keeping the cloud bill honest.8 conceptsOpens with credits
9. Chaos and resilience testingProving the system copes with failure before real failure does: experiments, blast radius, game days and the patterns that stop one slow dependency taking everything down.8 conceptsOpens with credits
10. On-call without burnoutRotations, escalation and alert hygiene that keep a pager sustainable, plus toil: what it is, how to measure it and how to cut it.7 conceptsOpens with credits
11. 11Glossary, A to ZEvery concept in the sheet and the smaller terms it uses, in one list.124 terms

SLOs, observability, Kubernetes platforms.