Jitter
Random variation added to retry delays to prevent multiple clients from retrying simultaneously after a shared failure event. Without jitter, all clients using the same exponential backoff schedule retry at the same instant, creating coordinated bursts that can re-trigger the original failure. Full jitter picks a random value between 0 and the calculated backoff, while decorrelated jitter uses a different formula to spread retries more evenly across time.
関連プロトコル
ガイドで言及されています
How to Implement and Handle Rate Limiting (429)
API Design & Best Practices
Debugging 429 Too Many Requests
Debugging & Troubleshooting
Essential Network Debugging Tools: ping, traceroute, mtr, and tcpdump
DNS & Networking
Implementing Retry with Exponential Backoff
Error Handling Patterns
API Gateway Retry and Timeout Policies
API Gateway Patterns
Chaos Engineering for APIs: Injecting Faults and Status Codes
Testing & Mocking