RDS · Replication · Advanced
Replica lag serves stale truth
Users read stale data immediately after successful writes.
INITIAL ALERT
Users read stale data immediately after successful writes.
You are primary on-call. Establish the failure domain, protect healthy capacity, and choose the safest next action.
Primary writes commit and database health is green.
The decisive evidence and solution remain locked inside the simulation.WHAT THIS TESTS
Production judgment across RDS · Replication.
Distributed failures amplify through queues, caches, replicas, alerts, and retries. Operators need to reason about system behavior rather than one isolated metric.
- 01Correlate service-level and dependency signals
- 02Control feedback loops and backlog growth
- 03Turn incidents into measurable safeguards
SIMULATION STRUCTURE
Four evidence layers. No answer shown here.
The full exercise unlocks each layer as you commit to decisions.
Alert
Confirm the reported symptom and customer impact.
Baseline
Compare the failing path with a known-good condition.
Finding
Identify the evidence that separates plausible causes.
Change
Choose a controlled recovery and a durable prevention step.
Can you make the call without seeing the solution?
Try three incidents free. Paid incidents unlock only after Stripe confirms lifetime access.