implement-a-circuit-breaker

# implement-a-circuit-breaker ## When to use - Use around a dependency whose failure can consume scarce resources or cascade. - Do not count caller cancellation or valid business rejection as dependency failure. ## Preconditions - Define operation, timeout, failures, volume, fallback, retry, idempotency, and recovery. ## Procedure 1. Classify success, slow, retryable, terminal, caller, and dependency outcomes. 2. Choose scope by operation, tenant, region, or endpoint without global collateral damage. 3. Define closed, open, and half-open transitions with minimum volume and bounded windows. 4. Coordinate timeout, retry, concurrency, and backoff. 5. Provide a truthful safe fallback or explicit unavailability. 6. Test flapping, low volume, partial failure, probe concurrency, and restart state. ## Failure plan - Allow authorized bounded bypass only with monitoring and resource protection. ## Worked example A payment breaker opens on upstream timeouts but not declined cards and permits one idempotent probe. ## Done - A circuit-breaker implementation records scope, classifications, thresholds, states, fallback, and telemetry - Low-volume, timeout, retry, probe, flapping, restart, bypass, and idempotency tests verify behavior