Concept questionHard~9 minEvent Streaming / 5 of 7

Interview question

How would you respond if someone says the broker gives exactly-once delivery?

Explain the exactly-once messaging caveat

Explains why exactly-once claims do not remove the need for idempotent side effects and consistency boundaries.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I would ask exactly-once for what boundary. A broker may provide strong guarantees inside its own log or producer-consumer protocol, but the application still writes to databases, sends emails, calls payment providers, and updates caches. Those side effects can succeed or fail around the broker acknowledgement. I would still design handlers to be idempotent.

Full answer below

Practice focus

Exactly-once semantics, broker boundaries, side effects, idempotent consumers, transactions, and practical guarantees.

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
Exactly-once guarantees apply only inside a defined broker or transactional boundary. Database writes, payments, email, caches, and acknowledgements still create failure windows that application idempotency must handle.
I would ask exactly-once for what boundary. A broker may provide strong guarantees inside its own log or producer-consumer protocol, but the application still writes to databases, sends emails, calls payment providers, and updates caches. Those side effects can succeed or fail around the broker acknowledgement. I would still design handlers to be idempotent.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
I would avoid dismissing broker guarantees, but I would not extend them beyond their scope. If a stream processor writes back to the same transactional log, exactly-once semantics may be meaningful for that pipeline. If the consumer updates SQL and then commits an offset, there is still a failure window. The practical design uses idempotency keys, unique constraints, inbox/outbox records, and retry-safe handlers.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
In production I document the guarantee at each boundary: broker publication, broker consumption, database mutation, external side effect, and acknowledgement or offset commit. Crash tests stop the worker before and after each boundary to prove whether work is lost, repeated, or safely deduplicated. Transactional broker features are valuable when all state remains inside their supported scope, but SQL updates, payments, and email still need inbox, outbox, unique constraints, or provider idempotency. Success is measured as exactly-once business effect where required, not as a broker label.

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

Believing exactly-once removes all duplicate side effects.
Rejecting broker guarantees without understanding their scope.
Ignoring database writes and external calls.
No crash testing around acknowledgement windows.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Cannot define the boundary of the guarantee.
Uses exactly-once as a slogan.
Has no idempotency story.
Ignores external side effects.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

Exactly-once between which components?
What failure window still exists with SQL writes?
How would you prove duplicate side effects are prevented?

Continue the section

Next in Event Streaming: Retention and Compaction

Step 6 of 7

Set event retention and compaction for consumer recovery

Opening drill

Concept questionHard~9 minEvent Streaming / 5 of 7

Interview question

How would you respond if someone says the broker gives exactly-once delivery?

Explain the exactly-once messaging caveat

Explains why exactly-once claims do not remove the need for idempotent side effects and consistency boundaries.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

Full answer below

Practice focus

Exactly-once semantics, broker boundaries, side effects, idempotent consumers, transactions, and practical guarantees.

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
Exactly-once guarantees apply only inside a defined broker or transactional boundary. Database writes, payments, email, caches, and acknowledgements still create failure windows that application idempotency must handle.
I would ask exactly-once for what boundary. A broker may provide strong guarantees inside its own log or producer-consumer protocol, but the application still writes to databases, sends emails, calls payment providers, and updates caches. Those side effects can succeed or fail around the broker acknowledgement. I would still design handlers to be idempotent.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
I would avoid dismissing broker guarantees, but I would not extend them beyond their scope. If a stream processor writes back to the same transactional log, exactly-once semantics may be meaningful for that pipeline. If the consumer updates SQL and then commits an offset, there is still a failure window. The practical design uses idempotency keys, unique constraints, inbox/outbox records, and retry-safe handlers.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
In production I document the guarantee at each boundary: broker publication, broker consumption, database mutation, external side effect, and acknowledgement or offset commit. Crash tests stop the worker before and after each boundary to prove whether work is lost, repeated, or safely deduplicated. Transactional broker features are valuable when all state remains inside their supported scope, but SQL updates, payments, and email still need inbox, outbox, unique constraints, or provider idempotency. Success is measured as exactly-once business effect where required, not as a broker label.

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

Believing exactly-once removes all duplicate side effects.
Rejecting broker guarantees without understanding their scope.
Ignoring database writes and external calls.
No crash testing around acknowledgement windows.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Cannot define the boundary of the guarantee.
Uses exactly-once as a slogan.
Has no idempotency story.
Ignores external side effects.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

Exactly-once between which components?
What failure window still exists with SQL writes?
How would you prove duplicate side effects are prevented?

Continue the section

Next in Event Streaming: Retention and Compaction

Step 6 of 7

Set event retention and compaction for consumer recovery

How would you respond if someone says the broker gives exactly-once delivery?

Practice focus

Answer path

Why this question matters

Review signals

Follow-up prompts

Continue the section

How would you respond if someone says the broker gives exactly-once delivery?

Practice focus

Answer path

Why this question matters

Review signals

Follow-up prompts

Continue the section