At-Most-Once Delivery
Pronunciation: at MOHST wuns dih-LIV-er-ee
Also known as: At Most Once Message Delivery
Definition
At-Most-Once Delivery is a delivery approach in which each message is attempted no more than once, preventing redelivery but allowing messages to be lost when transmission or processing fails. It favors duplicate avoidance and low overhead, unlike at-least-once delivery, which retries and expects idempotent handling.
Overview
Meaning and scope
At-Most-Once Delivery is a delivery approach in which each message is attempted no more than once, preventing redelivery but allowing messages to be lost when transmission or processing fails. It favors duplicate avoidance and low overhead, unlike at-least-once delivery, which retries and expects idempotent handling.
Mechanism and operation
The production boundary for At-Most-Once Delivery should identify the authoritative system, responsible owner, accepted states, and recovery path. A production implementation should use it only where occasional loss is acceptable, make loss visible through counters or reconciliation, preserve authoritative state elsewhere, and document which events are noncritical. Monitoring for At-Most-Once Delivery should distinguish transport success, processing success, and the final external or financial result. Useful measures include send failures, estimated loss rate, reconciliation gaps, consumer disconnects, unobserved state changes, and recovery actions triggered from the source of truth. Operational metrics for At-Most-Once Delivery should use stable denominators and separate technical activity from successful business completion.
Practical payment example
A notification system sends each event no more than once, accepting that an outage might mean a merchant never receives that particular notification.
Operational checks
For At-Most-Once Delivery, identifiers and timestamps should remain stable enough to trace the technical action to its final business outcome. Changes to At-Most-Once Delivery should be tested against normal, failed, delayed, duplicate, and recovery paths that apply to the operation. Evidence for At-Most-Once Delivery should preserve the input, configuration version, actor or service, decision, downstream reference, and final outcome.
Distinctions and limits
Key risks include silent event loss, incomplete audit trails, missed status updates, stale caches, irrecoverable notifications, and business actions that never occur after a transient failure. The At-Most-Once Delivery runbook should define who can retry, cancel, replay, reconcile, communicate, and approve an exception.
Related reading
Related definitions: At-Least-Once Delivery, Message Queue.
Example
A notification system sends each event no more than once, accepting that an outage might mean a merchant never receives that particular notification.
Comparison
Key risks include silent event loss, incomplete audit trails, missed status updates, stale caches, irrecoverable notifications, and business actions that never occur after a transient failure. It favors duplicate avoidance and low overhead, unlike at-least-once delivery, which retries and expects idempotent handling.
Key Takeaway
Use it only where occasional loss is acceptable, make loss visible through counters or reconciliation, preserve authoritative state elsewhere, and document which events are noncritical.
Sources
- Apache Kafka Documentation: Message Delivery Semantics — Apache Software Foundation (2026-08-03)
- CloudEvents Specification — Cloud Native Computing Foundation (2026-08-03)
- Reactive Streams Specification — Reactive Streams Initiative (2026-08-03)