Resilience

Platform Resilience & Fallback Handling Skill

Purpose

Standardize fault tolerance, circuit breaking, retry policies, and ambiguous failure modeling across external payment gateways and integrations.

1. Rules of Engagement

MANDATORY

• MUTATING RETRY SAFETY RULE: Never blindly retry a financially mutating external operation (e.g. payment initiation, charge, refund).
  • Apply @Retry ONLY when:
    1. The downstream endpoint explicitly guarantees idempotency support, OR
    2. The operation is strictly read-only / idempotent by HTTP semantics.
• FAILURE TRIAGE MODEL: Financial integrations MUST explicitly distinguish between three outcome states:
  1. Business Failure: Downstream explicitly rejected payment (e.g. INSUFFICIENT_FUNDS, CARD_EXPIRED). Status -> FAILED.
  2. Technical Failure: Integration call failed before execution (e.g. DNS failure, HTTP 400). Status -> REJECTED.
  3. Unknown Outcome: Request timed out after transmission (e.g. HTTP 504 Gateway Timeout). Status -> PENDING_VERIFICATION / UNKNOWN. (Do NOT immediately fail or retry!)

2. Circuit Breaker Configuration Pattern

resilience4j.circuitbreaker:
  instances:
    gatewayService:
      slidingWindowSize: 20
      minimumNumberOfCalls: 5
      failureRateThreshold: 50
      waitDurationInOpenState: 10000ms