Skip to content

SRE and Reliability

Define reliable service from the user's perspective, trace and measure a live system with the right signal for the right question, spend risk deliberately, and recover through practiced systems.

flowchart LR J[Junior: signals and first response] --> M[Middle: instrument and budget] M --> S[Senior: incident command and audit] --> P[Professional: platform governance]
flowchart LR UserNeed --> SLI --> SLO --> ErrorBudget --> Decision --> Incident --> Telemetry --> Learning --> UserNeed

This topic covers both halves of running a system in production: the SRE discipline (SLIs, SLOs, error budgets, incident response, postmortems) and the observability tooling it runs on (logs, metrics, traces, profiles, crash reports, sampling). It is the operational counterpart to Debug-Thinking, which covers the reasoning process — pattern recognition, bisection, hypothesis testing — you apply through this tooling.

Level Guide You are done when
Junior First incident response You can identify impact, follow a runbook, choose the right signal for a question, and verify recovery from the user path.
Middle Instrument and budget You can instrument a service with RED/USE metrics, propagate trace context, use error-budget burn to prioritize work, and reduce toil.
Senior Incident command and audit You can own SLOs across dependencies, lead incident roles, design telemetry for degraded state, and run blameless postmortems that produce durable fixes.
Professional Platform governance You can govern SLOs and telemetry cost across a shared platform, and design multi-window alerting and observability infrastructure at scale.

Practice rule

Define reliability as measurable user outcomes, not infrastructure uptime alone. Pick the observability signal — log, metric, trace, or profile — that answers the specific question you're asking, not the one that's easiest to add.

  • Debug-Thinking — the reasoning process this topic's tooling supports.
  • Performance — profiling and hot-path measurement this topic's telemetry feeds into.
  • Release — the deploys whose blast radius this topic's SLOs and error budgets bound.