How would an event consumer avoid processing the same message twice?

Use the inbox pattern to deduplicate consumed messages

Makes consumer-side deduplication atomic with the business change while supporting concurrent delivery, retention, replay, and failed processing.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I use a stable source event id and store an inbox row scoped to the consumer, for example (ConsumerName, MessageId). The handler inserts that row and applies the business change in the same database transaction. A unique constraint resolves concurrent duplicate deliveries: one transaction succeeds and the other recognizes the existing message. Recording the inbox row before or after the business change in separate commits would create either lost work or duplicate effects.

Full answer below

Practice focus

Inbox pattern, consumed message table, transactions, duplicate detection, retries, and idempotent consumers.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Why this question matters
The inbox records which messages a particular consumer has applied. Its value comes from sharing a transaction with the business change, not from performing a separate check before the handler runs.
I use a stable source event id and store an inbox row scoped to the consumer, for example (ConsumerName, MessageId). The handler inserts that row and applies the business change in the same database transaction. A unique constraint resolves concurrent duplicate deliveries: one transaction succeeds and the other recognizes the existing message. Recording the inbox row before or after the business change in separate commits would create either lost work or duplicate effects.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
I define whether the row means received, processing, completed, ignored, or failed. A simple handler may only need an atomic completed marker, while a long workflow needs durable state and a lease so a crashed attempt can resume. The broker message id must remain stable across redelivery; if a producer creates a new id on every retry, I use a business operation id instead. External calls still need provider idempotency or a separate reconciliation boundary because the local inbox transaction cannot make them atomic.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Checking for an inbox row and inserting later without a unique constraint.
Recording completion outside the business transaction.
Assuming the broker message id remains stable across producer retries.
Deleting inbox rows before replay and redelivery windows close.

How would an event consumer avoid processing the same message twice?

Practice focus

Answer path

Why this question matters

Review signals

Follow-up prompts

Continue the section

Tech stacks

Skill areas

Suggest an improvement