Delivery Semantics in Kafka: At Most Once, At Least Once, Exactly Once

When working with Apache Kafka, one of the most important questions is:
How many times can a message be delivered to a consumer?

This is called delivery semantics. Kafka supports three modes: at most once, at least once, and exactly once.

1. At Most Once

• Meaning: Each message is delivered 0 or 1 time.
• If there is a failure, the message may be lost.
• No duplicates, but risk of data loss.

How it happens:

• Consumer commits the offset before processing the message.
• If processing fails, Kafka thinks it’s done and skips it.

Example:

• A payment system that commits offset before charging the customer.
• If the app crashes, the payment message is lost and never retried.

Benefits

• Fastest performance (low overhead).
• No duplicates ever appear.
• Simpler logic for consumers.

Drawbacks

• Risk of data loss if a failure happens after offset commit but before processing.
• Not safe for critical workflows.

Use Case:

• OK for non-critical data like metrics, logs, or monitoring signals.

2. At Least Once

• Meaning: Each message is delivered 1 or more times.
• No data loss, but there can be duplicates.

How it happens:

• Consumer commits the offset after processing the message.
• If processing succeeds but commit fails, the same message will be reprocessed.

Example:

• A billing system charges a customer and then crashes before committing.
• On restart, it charges again → duplicate billing.

Benefits

• No message loss (every message is eventually processed).
• Default and safest option in most Kafka clients.

Drawbacks

• Duplicates may occur if the same message is processed multiple times.
• Requires idempotent consumers (processing must handle duplicates safely, e.g., avoid double-charging).

Use Case:

• Most common in Kafka.
• Safe for critical data if your processing is idempotent (can handle duplicates safely).