How do I design retries and dead-letter handling?

Definition:

Retries repeat failed work in the hope that a temporary problem has passed. Backoff spreads attempts over time, jitter avoids synchronized retry bursts, and a dead-letter queue stores work that exceeded its retry policy. Without limits, retries can turn a small outage into sustained overload.

Key points:

• Retry temporary failures with exponential backoff, random jitter and a limit.

• Respect provider rate limits; avoid retrying permanent validation failures.

• Send repeatedly failing messages to a dead-letter queue for investigation and controlled replay.

• A timeout after sending an external request means the result may be unknown, not that the external action definitely failed.

Interview point:

• Classify failures as permanent, temporary or uncertain before deciding whether and how to retry.