Message-Queue Delivery Semantics & Failure Handling Quiz

Q1. EASY: What does the 'at-least-once' delivery guarantee imply for message delivery? It ensures each message is delivered at most one time or not at all.
It ensures each message is delivered exactly one time without duplication or loss.
It ensures no message is lost, though a message may be delivered more than once.
It provides no guarantee; messages might be lost or duplicated.

Q2. EASY: What is the purpose of an idempotency key (unique message ID) in a messaging system? To detect and avoid processing duplicate messages.
To ensure messages are delivered in a specific order.
To count how many times a message has been delivered.
To increase the speed of message delivery.

Q3. EASY: What is a Dead-Letter Queue (DLQ) used for in message queuing systems? To temporarily hold messages before their initial delivery attempt.
To isolate messages that repeatedly fail processing (poison messages) for later review.
To store higher priority messages for faster handling.
To automatically replicate messages to backup queues.

Q4. EASY: Why are exponential backoff and jitter used when retrying message processing failures? To retry failures as quickly as possible without delay.
To ensure all failing messages are retried at the same fixed interval.
To decrease the wait time after each consecutive failure.
To gradually increase delay and add randomness to avoid retry collisions.

Q5. EASY: In a message queue with acknowledgments, what happens if a consumer fails to acknowledge a message within the visibility timeout? It is removed from the queue and lost.
It moves immediately to a dead-letter queue.
It becomes visible again on the queue for another consumer.
It gets automatically marked as processed by the broker.

Q6. MEDIUM: What is a trade-off of using multiple partitions for a message queue instead of a single FIFO queue? Losing the guarantee of a single global ordering of messages.
Significantly reducing the maximum throughput of the system.
Preventing any ordering of messages even within a partition.
Removing any delivery guarantee for messages.

Q7. MEDIUM: What happens to a message when it reaches its TTL (time-to-live) in a message queue? It stays in the queue until a consumer finally picks it up.
It expires and is removed from the queue (possibly sent to a dead-letter queue).
It triggers an immediate delivery attempt to a consumer, then resets the TTL.
It moves to the front of the queue for priority processing.

Q8. MEDIUM: Which statement is true regarding duplicate message detection in messaging systems? Brokers always automatically drop any duplicate message deliveries.
Duplicate messages cannot be detected by any system once sent.
Some brokers support duplicate detection, but consumers may also need idempotency to handle duplicates.
Only consumers can eliminate duplicates; brokers do not offer deduplication features.

Q9. MEDIUM: In message queuing, what is commonly referred to as a 'poison message'? A message containing malicious content used to attack consumer systems.
A message that is too large and exceeds the queue's size limit.
A message with invalid formatting that gets deleted on the first failure.
A message that consistently fails processing and ends up in a dead-letter queue.

Q10. MEDIUM: In a transactional messaging setup, what problem does the 'Outbox' pattern address? It ensures that a database transaction and a message publish either both succeed or both fail together.
It bundles multiple messages together to send in one batch for efficiency.
It orders messages from multiple producers into a single sequence.
It allows consumers to process messages without acknowledging them.

Q11. MEDIUM: In RabbitMQ, what is a Dead Letter Exchange (DLX)? A built-in exchange that automatically broadcasts every message to all queues.
A mechanism for encrypting messages in RabbitMQ.
An exchange that routes messages to a dead-letter queue when they are rejected or expire.
A special queue that stores all unacknowledged messages by default.

Q12. HARD: Which of the following is required to achieve exactly-once delivery semantics in Apache Kafka? Enable idempotent producers and use transactions so that writes and acknowledgments are atomic.
Use at-least-once delivery and implement duplicate handling on the consumer side.
Use a single partition topic to naturally achieve exactly one delivery.
Set the producer acknowledgment setting (acks) to 0 to avoid resending messages.

Q13. HARD: For an Amazon SQS FIFO queue, how long is a message's deduplication ID retained to prevent duplicate deliveries? About 5 minutes.
Around 24 hours.
Roughly 60 seconds.
Indefinitely, until the queue is purged.

Q14. HARD: If messages for the same key arrive out of order from parallel consumers or partitions, how can their correct order be restored before processing? Count on the broker to merge all partitions into a single ordered stream.
Have consumers buffer messages and sort them by key or sequence before processing.
Switch to at-most-once delivery so messages are never retried out of order.
Attach timestamps to messages to guarantee automatic ordering.

Q15. HARD: Why is a two-phase commit (2PC) often avoided when integrating a database and message broker for exactly-once processing? It cannot guarantee atomicity across a database and a message broker.
It only works if the system uses at-most-once delivery semantics.
It duplicates messages by design as part of its protocol.
It adds a lot of complexity and can lock resources, severely hurting performance.

software-architecture