Pattern 20 / RESILIENCE

Retry with Exponential Backoff

Use this when handling transient failures without giving up immediately.

Pressure: Handling transient failures without giving up immediately
Mechanism: Retry after increasing delays and add jitter so clients do not retry in lockstep
Toll: Poorly bounded retries amplify outages and increase tail latency

Diagram for Retry with Exponential Backoff — Architecture plate20

Executive brief

Retry with Exponential Backoff fits when handling transient failures without giving up immediately. Mechanism: retry after increasing delays and add jitter so clients do not retry in lockstep. Use it for temporary network errors, 429/503 responses, cloud APIs, and idempotent calls. The toll: poorly bounded retries amplify outages and increase tail latency.

Use when

Temporary network errors, 429/503 responses, cloud apis, and idempotent calls.

Example

Object upload retrying after 1s, 2s, 4s with jitter.

Review framing

Describe the pressure first, then the mechanism, then the cost. That keeps the design grounded.

Same pressure family

Reliability Patterns

19Circuit Breaker 21Bulkhead 22Timeout 23Idempotency 24Dead Letter Queue (DLQ)25Graceful Degradation