Observability and Incident Diagnosis
The question for this group
How do I know users are affected, and what evidence narrows down the cause?
Build the habit of diagnosing from evidence before changing configuration.
Topics to explore
- Operational signals: logs, metrics, traces, structured events, and correlation IDs.
- CloudWatch: log groups, retention, metric dimensions, dashboards, alarms, and Logs Insights.
- End-to-end diagnosis: frontend errors, synthetic checks, request tracing, and OpenTelemetry concepts.
- Audit and network evidence: CloudTrail, VPC Flow Logs, deployment history, and access changes.
- Service health: latency percentiles, traffic, error rate, saturation, SLIs, SLOs, and alert quality.
- Incident workflow: triage, mitigation, rollback, communication, runbooks, and blameless postmortems.
Practice target
Break a lab deployment deliberately, detect the symptom, trace the cause, restore service, and write a short incident note with evidence.