Practice incident response with a simulated system outage

Run an interactive outage simulation so I can practice incident response. System: [ARCHITECTURE - e.g. "web app behind a load balancer, Postgres primary and replica, Redis cache, a message queue, 3 microservices"] My role: [on-call engineer / incident commander / SRE] Difficulty: [beginner / intermediate / advanced] Failure type to focus on: [e.g. "cascading failures", "database problems", "bad deploys", "third-party outage", "surprise me"] Maximum turns: [e.g. 10] Simulation rules: - Secretly pick a root cause that sets off a chain of failures (e.g. connection pool exhaustion, retry storm, cache stampede, full disk, expired certificate, bad config push). Don't reveal it. - Open with the alert and the symptoms an on-call engineer would actually see: dashboard readings, error rates, latency, and a few log lines. - Each turn I say what I check or do. Return realistic results only for what I checked. Actions have consequences: restarts, rollbacks, failovers, and scaling can help, do nothing, or make things worse. - Track elapsed time and customer impact. Let new symptoms appear as the failure spreads. - If I go two turns without progress, give me one subtle hint. - Stay in the scenario and don't solve it for me. When the incident is resolved or the turns run out, debrief me: 1. The real root cause and the full failure chain. 2. What I did well and where I lost time. 3. The status updates and escalations I should have sent, and when. 4. Missing alerts, runbooks, or safeguards (timeouts, circuit breakers, backoff with jitter, rate limits, capacity headroom). 5. A score out of 5 for diagnosis, mitigation, and communication, with one thing to practice next.