RabbitMQ quiz — 37 questions with answers and explanations

11 Beginner · 26 Intermediate

All 37 RabbitMQ quiz questions below come with a short explanation, visible right on the page — no sign-in needed. They're grouped by beginner and intermediate levels and 14 subtopics, so you can revise one part at a time. Choose Try RabbitMQ quiz above to take a scored run and review the questions you got wrong.

What the RabbitMQ quiz covers

Operations5
Routing5
Fundamentals4
Queues4
Reliability4
Acknowledgement3
Dead Lettering3
Patterns3
Client1
Comparison1
Consumer1
Exchange1
Producer1
Queue1

Want long-form answers? Read 51 RabbitMQ interview questions with answers — same topic, in more depth.

Beginner level — 11 RabbitMQ questions

Question 1Fundamentals

In RabbitMQ, where does a producer send messages?

A.Straight into the queue the consumer reads from
B.Into an exchange
C.To the broker, which picks a queue at random
D.Directly to the consumer over a separate TCP connection

Answer: B

A producer always publishes to an exchange, never directly to a queue. The exchange looks at the message's routing key and the bindings to decide which queue to route to. Even when the code appears to "send to a queue", it's actually sending through the default exchange with a routing key matching the queue name.

References

RabbitMQ docs — AMQP 0-9-1 concepts

Question 2Fundamentals

How do exchange and queue differ in role?

A.The queue routes messages to the appropriate exchange
B.The exchange stores messages while the queue is just the consumer's read pointer
C.The exchange routes while the queue holds messages for consumers
D.They are two names for the same component

Answer: C

Exchange = router, queue = storage buffer. A message goes producer → exchange → (per binding) → queue → consumer. The exchange doesn't hold messages; if no queue accepts it, the message disappears rather than staying in the exchange.

References

RabbitMQ docs — AMQP 0-9-1 concepts

Question 3Routing

You need to broadcast an event to EVERY bound queue, ignoring routing keys. Which exchange type?

A.Headers — match on headers instead of a routing key
B.Topic — use a pattern to match many queues
C.Fanout
D.Direct — bind every queue with the same routing key

Answer: C

Fanout. It copies the message to every bound queue and ignores the routing key entirely. Quick memory aid for the 4 types: direct = exact key match, topic = pattern match (*/#), fanout = broadcast to all, headers = match on message headers.

References

RabbitMQ tutorial — Publish/Subscribe (fanout)

Question 4Routing

With the following code, which queue receives the message?

await ch.assertExchange('logs', 'direct')
await ch.bindQueue('q_error', 'logs', 'error')
await ch.bindQueue('q_info', 'logs', 'info')

ch.publish('logs', 'error', Buffer.from('disk full'))

A.Both q_error and q_info, since they bind to the same exchange
B.No queue — the publish call is missing a queue parameter
C.Only q_error
D.q_info, because info is the exchange's default level

Answer: C

Only q_error. A direct exchange matches the message's routing key (error) exactly against each queue's binding key. q_info is bound with key info so it doesn't receive it. If multiple queues are bound to key error, they all get a copy.

References

RabbitMQ tutorial — Routing (direct exchange)

Question 5Acknowledgement

A consumer runs the following code and crashes mid-processing. What happens to that message?

ch.consume('orders', async (msg) => {
  await chargeCustomer(msg) // crashes here
}, { noAck: true })

A.The broker redelivers the message to another consumer on the queue
B.The message is lost
C.The message moves into the default dead-letter queue
D.The broker holds the message until the consumer restarts

Answer: B

The message is lost. noAck: true (autoAck) means the broker considers the message delivered as soon as it's pushed to the socket — it doesn't wait for the consumer to process it. If the consumer crashes afterwards, there's nothing left to redeliver. For important work, use manual ack: only ch.ack(msg) after processing finishes.

References

RabbitMQ docs — Consumer acknowledgements

Question 6Fundamentals

Two consumers read from the same queue. How does the broker dispatch messages by default?

A.The first consumer to connect gets everything and the second is a standby
B.The broker measures load and picks the least busy consumer
C.Both receive a copy of every message
D.Round-robin, alternating between the consumers

Answer: D

Round-robin: message n goes to consumer A, n+1 to consumer B, alternating evenly — each message is delivered to only one consumer. This is the work queue (competing consumers) model: adding consumers increases parallel processing capacity.

References

RabbitMQ tutorial — Work Queues

Question 7Queues

At what scope does RabbitMQ guarantee message ordering?

A.Across the whole exchange, by publish time
B.Within a consumer group, as in Kafka
C.No scope guarantees ordering
D.Within one queue, when there is a single consumer

Answer: D

RabbitMQ maintains ordering within each queue. If application-side processing/completion order must also be preserved, use a single consumer and process sequentially; requeue/nack can shift the observed order.

References

RabbitMQ docs — Queues (ordering)

Question 8Reliability

A consumer loses its connection while holding unacked messages. What happens to them?

A.The broker holds them waiting for that same consumer to reconnect
B.They are lost because they were already delivered
C.They go straight into the Dead Letter Exchange
D.The broker requeues them and delivers them to another consumer

Answer: D

The broker requeues all unacked messages and redelivers them to a surviving consumer. The redelivery carries the redelivered=true flag — a sign this message may have been partially processed before.

References

RabbitMQ docs — Reliability guide

Question 9Dead Lettering

What is a Dead Letter Exchange (DLX) used for?

A.Preserving the messages of a queue that has just been deleted
B.Keeping a copy of every acknowledged message for reconciliation
C.Receiving messages removed from a queue so they can be handled separately
D.Returning failed messages to the producer that published them

Answer: C

A DLX is an exchange that receives messages dropped from a queue — instead of vanishing, they are routed to another queue for investigation or retry. It is declared via the x-dead-letter-exchange argument on the original queue.

References

RabbitMQ docs — Dead Letter Exchanges

Question 10Operations

In the management UI, what does a message in a queue's "Unacked" column mean?

A.A message not yet written to disk despite being declared persistent
B.A message the consumer nacked that is waiting to be requeued
C.A malformed message the broker cannot deliver to a consumer
D.A message delivered to a consumer but not yet acknowledged

Answer: D

"Unacked" = messages delivered to a consumer but not yet acked. "Ready" = messages still sitting in the queue waiting to be delivered. The sum of the two columns is the queue's real depth. Abnormally high Unacked is a sign that consumers are slow or forgot to ack.

References

RabbitMQ docs — Management plugin

Question 11Operations

What are vhosts in RabbitMQ used for?

A.Routing messages between exchanges with different names
B.Creating standby copies of a queue on another node in the cluster
C.Raising throughput by running the broker across several virtual machines
D.Isolating queues, exchanges, bindings and permissions per vhost

Answer: D

A vhost is a logically isolated namespace inside a broker: each vhost has its own queues, exchanges, bindings, and user permissions, and same-named resources in two vhosts are unrelated. Like a database inside a PostgreSQL server. Used to separate environments (staging/dev) or teams/applications sharing one broker.

References

RabbitMQ docs — Virtual hosts

Intermediate level — 26 RabbitMQ questions

Question 1Routing

With the topic exchange below, which queue receives the message?

await ch.assertExchange('events', 'topic')
await ch.bindQueue('q1', 'events', 'order.*')
await ch.bindQueue('q2', 'events', 'order.#')

ch.publish('events', 'order.payment.failed', Buffer.from('...'))

A.Both q1 and q2, since both start with order
B.Only q2
C.Only q1, because * matches more broadly than #
D.No queue, because the routing key has three parts

Answer: B

Only q2. In a topic exchange: * substitutes exactly one word (a word = a segment between dots), # substitutes zero or more words. order.payment.failed has two words after order, so order.* misses while order.# matches. If order.created were published, both queues would receive it.

References

RabbitMQ tutorial — Topics

Question 2Routing

A message is published to an exchange but its routing key matches no binding. What happens by default?

A.The message waits in the exchange until a matching binding appears
B.The message is dropped silently
C.The broker raises a channel exception back to the producer
D.The message lands in the exchange's default queue

Answer: B

The message is silently dropped — no error, no warning. A producer that wants to know must enable the mandatory flag (the broker returns the message via basic.return) or configure an alternate exchange to catch unroutable messages into a separate queue.

References

RabbitMQ docs — Unroutable message handling

Question 3Acknowledgement

A consumer hits a malformed message and calls `basic.nack` with requeue=true. What is the risk?

A.The queue is locked until that message is acknowledged
B.An infinite loop: the message is redelivered and fails again
C.The message is lost, since nack means rejecting it permanently
D.The broker gradually increases the delay between redeliveries

Answer: B

Poison message loop: a permanently failing message (wrong format, missing field) is redelivered immediately and fails again, consuming CPU and flooding logs endlessly — RabbitMQ has no backoff or default redelivery limit. For permanent failures, nack with requeue=false plus a dead-letter exchange.

References

RabbitMQ docs — Negative acknowledgement and requeueing

Question 4Acknowledgement

Why should you set prefetch (`basic.qos`) when a consumer uses manual ack?

A.It stops the broker piling unlimited unacknowledged messages onto one consumer
B.Prefetch enables batch compression to raise network throughput
C.It ensures messages are written to disk before being delivered
D.Prefetch is a prerequisite for manual ack to work

Answer: A

Without setting prefetch, the broker pushes messages without limit to a consumer: one consumer holds thousands of unacked messages in memory while others sit idle, and if that consumer crashes everything must be redelivered. prefetch(n) caps the maximum n unacknowledged messages per consumer — once the limit is reached the broker waits for an ack before delivering the next one.

References

RabbitMQ docs — Consumer prefetch

Question 5Fundamentals· select all that apply

Which settings are required for a message to survive a broker restart? (select all that apply)

A.The queue is declared durable
B.The message is published with delivery_mode=2 (persistent)
C.The consumer uses manual ack rather than autoAck
D.The exchange is a direct type rather than fanout

Answer: A, B

You need BOTH: a durable queue and a persistent message (delivery_mode=2, in amqplib { persistent: true }). Missing either one means loss: a non-durable queue makes the whole queue disappear; a non-persistent message leaves the queue present but empty inside.

References

RabbitMQ docs — Queue durability

Question 6Routing

Three different services each need ALL events from one producer. How do you design the queues?

A.One exchange per service, with the producer publishing three times
B.One queue per service, all bound to one fanout exchange
C.One shared queue with the queue's broadcast option enabled
D.All three services consume one shared queue

Answer: B

One separate queue per service, all three bound to a single fanout exchange — each message is replicated into all three queues. The foundational rule: within ONE queue, consumers share messages (work queue); if you want everyone to receive all (pub/sub), split the queues. Within a service you can still scale many instances reading the same shared queue for that service.

References

RabbitMQ tutorial — Publish/Subscribe

Question 7Reliability

A producer uses a confirm channel as below. What does a successful `waitForConfirms()` mean?

const ch = await conn.createConfirmChannel()
ch.sendToQueue('orders', Buffer.from(payload), { persistent: true })
await ch.waitForConfirms()

A.The message has been written to disk on every node of the cluster
B.The broker has received the message and taken responsibility for it
C.The consumer has received and finished processing the message
D.An AMQP transaction has been committed on the broker

Answer: B

The broker has received the message and taken responsibility for it — for persistent messages on a durable queue, the confirm fires only after it has been written to disk. Without a confirm, publishing is fire-and-forget: a mid-flight network drop loses the message without the producer ever knowing.

References

RabbitMQ docs — Publisher Confirms and Consumer Acknowledgements

Question 8Reliability

Why should a RabbitMQ consumer be written to be idempotent?

A.Because message ordering between queues is not guaranteed
B.Because a fanout exchange copies the message to every consumer of one queue
C.Because the broker retries every message, including acknowledged ones
D.Because at-least-once means one message may be delivered several times

Answer: D

Because RabbitMQ delivers at-least-once: the consumer finishes processing but the ack never reaches the broker (crash, network drop), so the message is redelivered and processed a second time. The standard fix: attach a unique message id on publish, and have the consumer store processed ids and skip duplicates.

References

RabbitMQ docs — Reliability guide (consumer side)

Question 9Dead Lettering· select all that apply

A queue has a DLX configured. Which cases send a message to the DLX? (select all that apply)

A.The consumer loses its connection with the message unacknowledged
B.The consumer acknowledges the message successfully
C.The queue exceeds its max-length
D.The message's TTL expires while it sits in the queue
E.The consumer rejects/nacks with requeue=false

Answer: C, D, E

Three paths into the DLX: reject/nack with requeue=false, TTL expiry, and queue hitting max-length. A lost connection causes the message to be requeued, not dead-lettered — telling these two paths apart is the common mix-up.

References

RabbitMQ docs — Dead Letter Exchanges

Question 10Dead Lettering

You need to retry a message after 30 seconds without installing plugins. What is the common RabbitMQ approach?

A.Attach an x-delay header of 30000 when publishing to an ordinary queue
B.The consumer holds the message, waits 30 seconds with setTimeout, then processes and acks
C.Nack with requeue=true and configure the broker to delay 30 seconds before redelivering
D.Publish into a waiting queue with a 30s TTL whose DLX points back at the main queue

Answer: D

Use a delay queue: a waiting queue with no consumer, declared with x-message-ttl=30000 and x-dead-letter-exchange pointing back at the main queue's exchange. A failing message is published to the waiting queue, sits there for 30 seconds, and when the TTL expires the broker dead-letters it back to the main queue.

References

RabbitMQ docs — Dead Letter Exchanges

Question 11Reliability

What is wrong with this consumer code?

ch.consume('payments', async (msg) => {
  try {
    await handle(JSON.parse(msg.content))
    ch.ack(msg)
  } catch (err) {
    ch.nack(msg, false, true)
  }
})

A.A broken message is requeued and fails in an infinite loop
B.It must ack before processing so the broker does not deliver duplicates
C.It is missing waitForConfirms so the broker may not have received the ack
D.It should enable auto-ack so acking need not be manual

Answer: A

Every error gets nack(msg, false, true) — i.e. unconditional requeue. A message with bad data (malformed JSON) will fail exactly the same way again, creating a poison-message loop that burns CPU and blocks the queue. Permanent failures (requeue=false, send to DLX) need to be separated from transient ones.

References

RabbitMQ docs — Negative acknowledgement and requeueing

Question 12Queues

What is the difference between `x-message-ttl` and `x-expires` on a queue declaration?

A.x-expires drops the oldest messages when the queue exceeds its length limit
B.They are the same; x-expires is the old name
C.x-message-ttl expires messages; x-expires deletes the queue itself when unused
D.x-message-ttl is for durable queues and x-expires for temporary ones

Answer: C

x-message-ttl is the lifetime of each message: once expired it's discarded (dead-lettered if a DLX is configured). x-expires is the lifetime of the whole queue: if there's no consumer and nothing touches it for that period, the broker deletes the queue along with every message in it.

References

RabbitMQ docs — Time-To-Live and Expiration

Question 13Queues

A queue has `x-max-length=10000` and is full. By default, what happens when a new message is published?

A.The oldest message at the head is dropped to make room (drop-head)
B.The broker blocks the producer's connection until the queue has room
C.The new message is staged on disk until the queue drains
D.The producer receives a basic.nack and the new message is rejected

Answer: A

The default is drop-head: the oldest message is discarded to make room for the new one — and the producer is never notified. If the queue has a DLX, the discarded message is dead-lettered with reason maxlen instead of being lost.

References

RabbitMQ docs — Queue Length Limit

Question 14Queues

A queue accumulates millions of backlogged messages and the RabbitMQ node consumes lots of RAM. What is the right broker-level approach?

A.Split the backlog across a fanout exchange so several queues share it
B.Add RAM to the node because messages must live in memory
C.Move messages to disk early instead of holding them in RAM (lazy behaviour)
D.Enable auto-ack so messages leave the queue faster

Answer: C

Get messages out to disk early instead of holding them in RAM. This used to be x-queue-mode=lazy; since RabbitMQ 3.12, classic queues behave lazily by default (keeping only a small in-flight portion in RAM), and quorum queues manage memory the same way. A million-message backlog then only costs disk, and won't push the node into a memory alarm.

References

RabbitMQ docs — Lazy Queues / classic queue storage

Question 15Comparison

When should you pick RabbitMQ over Kafka?

A.When you must write millions of events per second into one data stream
B.When you need a task queue with complex routing where messages are deleted on consumption
C.When you need a new consumer to replay the whole event history
D.When you must retain data for days for several independent consumer groups

Answer: B

Pick RabbitMQ when the problem is a work queue: each message is handled by one worker and then disappears, and you need complex routing (topic/headers exchange), priority, TTL, per-message ack. Pick Kafka when the data is an event log that must be retained for replay, with many independent consumers and very high throughput.

References

RabbitMQ docs — Queues

Question 16Operations

Why should an application open one TCP connection with many channels instead of many connections?

A.The broker hard-limits each client to a single connection
B.Channels are more durable than connections and survive network faults
C.Messages sent over a channel are compressed while those over a connection are not
D.Channels multiplex over one connection so opening one is far cheaper

Answer: D

A channel is a virtual multiplexed channel over one TCP connection. Opening a new connection costs a TCP + TLS handshake and broker resources, while opening a channel is nearly free. Standard pattern: one connection per process, one channel per thread — never share a channel between threads.

References

RabbitMQ docs — Channels

Question 17Patterns

In the RPC pattern below, why must the consumer check `correlationId` on the response?

const { queue: replyQueue } = await ch.assertQueue('', { exclusive: true })
const correlationId = crypto.randomUUID()

ch.consume(replyQueue, (msg) => {
  if (msg.properties.correlationId === correlationId) {
    console.log('result:', msg.content.toString())
  }
})

ch.sendToQueue('rpc_queue', Buffer.from('30'), {
  replyTo: replyQueue,
  correlationId,
})

A.To verify the response really came from a legitimate RPC server
B.To match a response to the right request when the reply queue receives several
C.So the broker knows which reply queue to route the response to
D.To prevent duplicate publishes caused by server-side retries

Answer: B

The client sends a request with replyTo (the queue for results) and correlationId (the request identifier). When the server finishes, it publishes the response to replyTo with the same correlationId attached. Since one reply queue can receive many responses, the client uses correlationId to match a response to the right pending request.

References

RabbitMQ tutorial 6 — RPC

Question 18Operations· select all that apply

Queue depth keeps rising during peak hours. Which responses are reasonable?

A.Check whether each message's processing time has risen unusually
B.Purge the queue to bring the depth back to 0
C.Raise the queue's x-max-priority
D.Add more consumers to the queue
E.Raise the prefetch if it is set too low

Answer: A, D, E

Rising depth means the inbound rate is greater than the outbound rate. Three valid directions: add consumers, tune prefetch if the bottleneck is round-trips, and investigate why consumers have slowed down. Purge only removes the symptom and loses data; priority does not change overall throughput.

References

RabbitMQ docs — Consumer prefetch

Question 19Operations

What problem do RabbitMQ heartbeats solve?

A.Telling the broker that the consumer has finished processing a message
B.Measuring the network latency between client and broker to tune the prefetch automatically
C.Detecting a dead connection that the TCP layer does not report promptly
D.Keeping messages from being expired by TTL while the consumer is alive

Answer: C

Heartbeats are small frames both sides send periodically to detect a dead connection in cases where TCP goes silent: sudden network drop, firewall/NAT cutting an idle connection, the peer losing power. After several missed heartbeats, the other side treats the connection as dead and closes it.

References

RabbitMQ docs — Heartbeats

Question 20Patterns

A message must only be processed 5 minutes after being sent. Without the delayed-message plugin, what is the standard approach?

A.Publish into a waiting queue with a 5-minute TTL whose DLX points at the main exchange
B.Set an x-delay header on the message; every broker understands it
C.Have the consumer receive the message and sleep 5 minutes before processing
D.Use amqplib's delay option on sendToQueue

Answer: A

The TTL + DLX pattern: publish to a waiting queue with no consumer, set x-message-ttl: 300000 and x-dead-letter-exchange pointing to the main processing exchange. After 5 minutes the message expires and is dead-lettered to the destination — a delayed message without any plugin.

References

RabbitMQ docs — Dead letter exchanges

Question 21Patterns

What is a queue declared with `{ exclusive: true }` suited for?

A.A priority queue that only important messages may enter
B.A queue that must survive broker restarts intact
C.A temporary reply queue for RPC, tied to one connection's lifetime
D.A durable work queue needing several consumers to share the load

Answer: C

An exclusive queue belongs to a single connection: only the connection that declared it can use it, and the queue is deleted when that connection closes. It suits an RPC client's reply queue, or a per-session queue — things that are meaningless once the owner disappears.

References

RabbitMQ docs — Exclusive queues

Question 22Producer

How do publisher confirms differ from AMQP transactions?

A.Confirms are asynchronous so they are much faster
B.Confirms only work with durable queues
C.Transactions support several queues while confirms do not
D.Confirms guarantee the message reached a consumer

Answer: A

Transaction is synchronous: the producer waits for each commit to complete, so throughput is very low. Confirm is asynchronous: the producer keeps sending and the broker sends confirmations later with sequence numbers. With the same safety, confirm is many times faster.

References

RabbitMQ Docs — Publisher confirms

Question 23Consumer

What should determine a consumer prefetch value?

A.The processing time per message and the network latency
B.How many queues the consumer subscribes to
C.How many messages are currently in the queue
D.The broker's memory capacity

Answer: A

Base it on per-message processing time and network latency: the value must be large enough that the consumer is never idle waiting for the next message. Fast processing with slow network needs a large value; long processing needs a small value, often one, for fair distribution.

References

RabbitMQ Docs — Consumer prefetch

Question 24Client

An app loses its broker connection during a broker restart. What must the client handle?

A.Restart the whole application
B.Nothing, since the broker reconnects the client itself
C.Reconnect with backoff and redeclare the objects
D.Just retry the failed publish call

Answer: C

The client must reconnect with backoff, reopen the channel, redeclare queues and bindings if the client created them, and re-register to receive messages. Many libraries have built-in automatic recovery, but it must be enabled and you need to understand what it restores.

References

RabbitMQ Docs — Connections and recovery

Question 25Queue

How does a lazy queue differ from the default mode?

A.It only delivers messages when a consumer explicitly asks
B.It delays delivering messages for a period
C.It writes messages to disk early rather than holding them in memory
D.It only writes messages to disk when the broker is shutting down

Answer: C

This mode writes messages to disk immediately instead of keeping them in memory. That lets a queue hold millions of messages without exhausting broker memory — suitable for queues that occasionally back up heavily. The trade-off is higher latency when consumers read.

References

RabbitMQ Docs — Lazy queues

Question 26Exchange

What does a headers exchange route on, and when should it be used?

A.On the order in which the messages were sent
B.On the message's size
C.On header attributes, when several criteria are needed
D.On the name of the producer that sent the message

Answer: C

It matches on header attributes instead of a routing key string, and can be declared to match all or just one attribute. Suitable when routing depends on multiple independent criteria that would be hard to read if crammed into a single key string.

References

RabbitMQ — AMQP concepts and exchange types

13 more Advanced questions for RabbitMQ

Advanced goes deep on real-world scenarios and the technical pitfalls candidates actually run into. Included with Pro.

Upgrade to Pro

RabbitMQ quiz — 37 questions with answers and explanations

11 Beginner · 26 Intermediate

Try RabbitMQ quiz

What the RabbitMQ quiz covers

Operations5
Routing5
Fundamentals4
Queues4
Reliability4
Acknowledgement3
Dead Lettering3
Patterns3
Client1
Comparison1
Consumer1
Exchange1
Producer1
Queue1

Want long-form answers? Read 51 RabbitMQ interview questions with answers — same topic, in more depth.

Beginner level — 11 RabbitMQ questions

Question 1Fundamentals

In RabbitMQ, where does a producer send messages?

A.Straight into the queue the consumer reads from
B.Into an exchange
C.To the broker, which picks a queue at random
D.Directly to the consumer over a separate TCP connection

Answer: B

References

RabbitMQ docs — AMQP 0-9-1 concepts

Question 2Fundamentals

How do exchange and queue differ in role?

A.The queue routes messages to the appropriate exchange
B.The exchange stores messages while the queue is just the consumer's read pointer
C.The exchange routes while the queue holds messages for consumers
D.They are two names for the same component

Answer: C

References

RabbitMQ docs — AMQP 0-9-1 concepts

Question 3Routing

You need to broadcast an event to EVERY bound queue, ignoring routing keys. Which exchange type?

A.Headers — match on headers instead of a routing key
B.Topic — use a pattern to match many queues
C.Fanout
D.Direct — bind every queue with the same routing key

Answer: C

References

RabbitMQ tutorial — Publish/Subscribe (fanout)

Question 4Routing

With the following code, which queue receives the message?

await ch.assertExchange('logs', 'direct')
await ch.bindQueue('q_error', 'logs', 'error')
await ch.bindQueue('q_info', 'logs', 'info')

ch.publish('logs', 'error', Buffer.from('disk full'))

A.Both q_error and q_info, since they bind to the same exchange
B.No queue — the publish call is missing a queue parameter
C.Only q_error
D.q_info, because info is the exchange's default level

Answer: C

References

RabbitMQ tutorial — Routing (direct exchange)

Question 5Acknowledgement

A consumer runs the following code and crashes mid-processing. What happens to that message?

ch.consume('orders', async (msg) => {
  await chargeCustomer(msg) // crashes here
}, { noAck: true })

A.The broker redelivers the message to another consumer on the queue
B.The message is lost
C.The message moves into the default dead-letter queue
D.The broker holds the message until the consumer restarts

Answer: B

References

RabbitMQ docs — Consumer acknowledgements

Question 6Fundamentals

Two consumers read from the same queue. How does the broker dispatch messages by default?

A.The first consumer to connect gets everything and the second is a standby
B.The broker measures load and picks the least busy consumer
C.Both receive a copy of every message
D.Round-robin, alternating between the consumers

Answer: D

References

RabbitMQ tutorial — Work Queues

Question 7Queues

At what scope does RabbitMQ guarantee message ordering?

A.Across the whole exchange, by publish time
B.Within a consumer group, as in Kafka
C.No scope guarantees ordering
D.Within one queue, when there is a single consumer

Answer: D

References

RabbitMQ docs — Queues (ordering)

Question 8Reliability

A consumer loses its connection while holding unacked messages. What happens to them?

A.The broker holds them waiting for that same consumer to reconnect
B.They are lost because they were already delivered
C.They go straight into the Dead Letter Exchange
D.The broker requeues them and delivers them to another consumer

Answer: D

References

RabbitMQ docs — Reliability guide

Question 9Dead Lettering

What is a Dead Letter Exchange (DLX) used for?

A.Preserving the messages of a queue that has just been deleted
B.Keeping a copy of every acknowledged message for reconciliation
C.Receiving messages removed from a queue so they can be handled separately
D.Returning failed messages to the producer that published them

Answer: C

References

RabbitMQ docs — Dead Letter Exchanges

Question 10Operations

In the management UI, what does a message in a queue's "Unacked" column mean?

A.A message not yet written to disk despite being declared persistent
B.A message the consumer nacked that is waiting to be requeued
C.A malformed message the broker cannot deliver to a consumer
D.A message delivered to a consumer but not yet acknowledged

Answer: D

References

RabbitMQ docs — Management plugin

Question 11Operations

What are vhosts in RabbitMQ used for?

A.Routing messages between exchanges with different names
B.Creating standby copies of a queue on another node in the cluster
C.Raising throughput by running the broker across several virtual machines
D.Isolating queues, exchanges, bindings and permissions per vhost

Answer: D

References

RabbitMQ docs — Virtual hosts

Intermediate level — 26 RabbitMQ questions

Question 1Routing

With the topic exchange below, which queue receives the message?

await ch.assertExchange('events', 'topic')
await ch.bindQueue('q1', 'events', 'order.*')
await ch.bindQueue('q2', 'events', 'order.#')

ch.publish('events', 'order.payment.failed', Buffer.from('...'))

A.Both q1 and q2, since both start with order
B.Only q2
C.Only q1, because * matches more broadly than #
D.No queue, because the routing key has three parts

Answer: B

References

RabbitMQ tutorial — Topics

Question 2Routing

A message is published to an exchange but its routing key matches no binding. What happens by default?

A.The message waits in the exchange until a matching binding appears
B.The message is dropped silently
C.The broker raises a channel exception back to the producer
D.The message lands in the exchange's default queue

Answer: B

References

RabbitMQ docs — Unroutable message handling

Question 3Acknowledgement

A consumer hits a malformed message and calls `basic.nack` with requeue=true. What is the risk?

A.The queue is locked until that message is acknowledged
B.An infinite loop: the message is redelivered and fails again
C.The message is lost, since nack means rejecting it permanently
D.The broker gradually increases the delay between redeliveries

Answer: B

References

RabbitMQ docs — Negative acknowledgement and requeueing

Question 4Acknowledgement

Why should you set prefetch (`basic.qos`) when a consumer uses manual ack?

A.It stops the broker piling unlimited unacknowledged messages onto one consumer
B.Prefetch enables batch compression to raise network throughput
C.It ensures messages are written to disk before being delivered
D.Prefetch is a prerequisite for manual ack to work

Answer: A

References

RabbitMQ docs — Consumer prefetch

Question 5Fundamentals· select all that apply

Which settings are required for a message to survive a broker restart? (select all that apply)

A.The queue is declared durable
B.The message is published with delivery_mode=2 (persistent)
C.The consumer uses manual ack rather than autoAck
D.The exchange is a direct type rather than fanout

Answer: A, B

References

RabbitMQ docs — Queue durability

Question 6Routing

Three different services each need ALL events from one producer. How do you design the queues?

A.One exchange per service, with the producer publishing three times
B.One queue per service, all bound to one fanout exchange
C.One shared queue with the queue's broadcast option enabled
D.All three services consume one shared queue

Answer: B

References

RabbitMQ tutorial — Publish/Subscribe

Question 7Reliability

A producer uses a confirm channel as below. What does a successful `waitForConfirms()` mean?

const ch = await conn.createConfirmChannel()
ch.sendToQueue('orders', Buffer.from(payload), { persistent: true })
await ch.waitForConfirms()

A.The message has been written to disk on every node of the cluster
B.The broker has received the message and taken responsibility for it
C.The consumer has received and finished processing the message
D.An AMQP transaction has been committed on the broker

Answer: B

References

RabbitMQ docs — Publisher Confirms and Consumer Acknowledgements

Question 8Reliability

Why should a RabbitMQ consumer be written to be idempotent?

A.Because message ordering between queues is not guaranteed
B.Because a fanout exchange copies the message to every consumer of one queue
C.Because the broker retries every message, including acknowledged ones
D.Because at-least-once means one message may be delivered several times

Answer: D

References

RabbitMQ docs — Reliability guide (consumer side)

Question 9Dead Lettering· select all that apply

A queue has a DLX configured. Which cases send a message to the DLX? (select all that apply)

A.The consumer loses its connection with the message unacknowledged
B.The consumer acknowledges the message successfully
C.The queue exceeds its max-length
D.The message's TTL expires while it sits in the queue
E.The consumer rejects/nacks with requeue=false

Answer: C, D, E

References

RabbitMQ docs — Dead Letter Exchanges

Question 10Dead Lettering

You need to retry a message after 30 seconds without installing plugins. What is the common RabbitMQ approach?

A.Attach an x-delay header of 30000 when publishing to an ordinary queue
B.The consumer holds the message, waits 30 seconds with setTimeout, then processes and acks
C.Nack with requeue=true and configure the broker to delay 30 seconds before redelivering
D.Publish into a waiting queue with a 30s TTL whose DLX points back at the main queue

Answer: D

References

RabbitMQ docs — Dead Letter Exchanges

Question 11Reliability

What is wrong with this consumer code?

ch.consume('payments', async (msg) => {
  try {
    await handle(JSON.parse(msg.content))
    ch.ack(msg)
  } catch (err) {
    ch.nack(msg, false, true)
  }
})

A.A broken message is requeued and fails in an infinite loop
B.It must ack before processing so the broker does not deliver duplicates
C.It is missing waitForConfirms so the broker may not have received the ack
D.It should enable auto-ack so acking need not be manual

Answer: A

References

RabbitMQ docs — Negative acknowledgement and requeueing

Question 12Queues

What is the difference between `x-message-ttl` and `x-expires` on a queue declaration?

A.x-expires drops the oldest messages when the queue exceeds its length limit
B.They are the same; x-expires is the old name
C.x-message-ttl expires messages; x-expires deletes the queue itself when unused
D.x-message-ttl is for durable queues and x-expires for temporary ones

Answer: C

References

RabbitMQ docs — Time-To-Live and Expiration

Question 13Queues

A queue has `x-max-length=10000` and is full. By default, what happens when a new message is published?

A.The oldest message at the head is dropped to make room (drop-head)
B.The broker blocks the producer's connection until the queue has room
C.The new message is staged on disk until the queue drains
D.The producer receives a basic.nack and the new message is rejected

Answer: A

References

RabbitMQ docs — Queue Length Limit

Question 14Queues

A queue accumulates millions of backlogged messages and the RabbitMQ node consumes lots of RAM. What is the right broker-level approach?

A.Split the backlog across a fanout exchange so several queues share it
B.Add RAM to the node because messages must live in memory
C.Move messages to disk early instead of holding them in RAM (lazy behaviour)
D.Enable auto-ack so messages leave the queue faster

Answer: C

References

RabbitMQ docs — Lazy Queues / classic queue storage

Question 15Comparison

When should you pick RabbitMQ over Kafka?

A.When you must write millions of events per second into one data stream
B.When you need a task queue with complex routing where messages are deleted on consumption
C.When you need a new consumer to replay the whole event history
D.When you must retain data for days for several independent consumer groups

Answer: B

References

RabbitMQ docs — Queues

Question 16Operations

Why should an application open one TCP connection with many channels instead of many connections?

A.The broker hard-limits each client to a single connection
B.Channels are more durable than connections and survive network faults
C.Messages sent over a channel are compressed while those over a connection are not
D.Channels multiplex over one connection so opening one is far cheaper

Answer: D

References

RabbitMQ docs — Channels

Question 17Patterns

In the RPC pattern below, why must the consumer check `correlationId` on the response?

const { queue: replyQueue } = await ch.assertQueue('', { exclusive: true })
const correlationId = crypto.randomUUID()

ch.consume(replyQueue, (msg) => {
  if (msg.properties.correlationId === correlationId) {
    console.log('result:', msg.content.toString())
  }
})

ch.sendToQueue('rpc_queue', Buffer.from('30'), {
  replyTo: replyQueue,
  correlationId,
})

A.To verify the response really came from a legitimate RPC server
B.To match a response to the right request when the reply queue receives several
C.So the broker knows which reply queue to route the response to
D.To prevent duplicate publishes caused by server-side retries

Answer: B

References

RabbitMQ tutorial 6 — RPC

Question 18Operations· select all that apply

Queue depth keeps rising during peak hours. Which responses are reasonable?

A.Check whether each message's processing time has risen unusually
B.Purge the queue to bring the depth back to 0
C.Raise the queue's x-max-priority
D.Add more consumers to the queue
E.Raise the prefetch if it is set too low

Answer: A, D, E

References

RabbitMQ docs — Consumer prefetch

Question 19Operations

What problem do RabbitMQ heartbeats solve?

A.Telling the broker that the consumer has finished processing a message
B.Measuring the network latency between client and broker to tune the prefetch automatically
C.Detecting a dead connection that the TCP layer does not report promptly
D.Keeping messages from being expired by TTL while the consumer is alive

Answer: C

References

RabbitMQ docs — Heartbeats

Question 20Patterns

A message must only be processed 5 minutes after being sent. Without the delayed-message plugin, what is the standard approach?

A.Publish into a waiting queue with a 5-minute TTL whose DLX points at the main exchange
B.Set an x-delay header on the message; every broker understands it
C.Have the consumer receive the message and sleep 5 minutes before processing
D.Use amqplib's delay option on sendToQueue

Answer: A

References

RabbitMQ docs — Dead letter exchanges

Question 21Patterns

What is a queue declared with `{ exclusive: true }` suited for?

A.A priority queue that only important messages may enter
B.A queue that must survive broker restarts intact
C.A temporary reply queue for RPC, tied to one connection's lifetime
D.A durable work queue needing several consumers to share the load

Answer: C

References

RabbitMQ docs — Exclusive queues

Question 22Producer

How do publisher confirms differ from AMQP transactions?

A.Confirms are asynchronous so they are much faster
B.Confirms only work with durable queues
C.Transactions support several queues while confirms do not
D.Confirms guarantee the message reached a consumer

Answer: A

References

RabbitMQ Docs — Publisher confirms

Question 23Consumer

What should determine a consumer prefetch value?

A.The processing time per message and the network latency
B.How many queues the consumer subscribes to
C.How many messages are currently in the queue
D.The broker's memory capacity

Answer: A

References

RabbitMQ Docs — Consumer prefetch

Question 24Client

An app loses its broker connection during a broker restart. What must the client handle?

A.Restart the whole application
B.Nothing, since the broker reconnects the client itself
C.Reconnect with backoff and redeclare the objects
D.Just retry the failed publish call

Answer: C

References

RabbitMQ Docs — Connections and recovery

Question 25Queue

How does a lazy queue differ from the default mode?

A.It only delivers messages when a consumer explicitly asks
B.It delays delivering messages for a period
C.It writes messages to disk early rather than holding them in memory
D.It only writes messages to disk when the broker is shutting down

Answer: C

References

RabbitMQ Docs — Lazy queues

Question 26Exchange

What does a headers exchange route on, and when should it be used?

A.On the order in which the messages were sent
B.On the message's size
C.On header attributes, when several criteria are needed
D.On the name of the producer that sent the message

Answer: C

References

RabbitMQ — AMQP concepts and exchange types

13 more Advanced questions for RabbitMQ

Advanced goes deep on real-world scenarios and the technical pitfalls candidates actually run into. Included with Pro.

Upgrade to Pro

Beginner level — 11 RabbitMQ questions

In RabbitMQ, where does a producer send messages?

How do exchange and queue differ in role?

You need to broadcast an event to EVERY bound queue, ignoring routing keys. Which exchange type?

With the following code, which queue receives the message?

A consumer runs the following code and crashes mid-processing. What happens to that message?

Two consumers read from the same queue. How does the broker dispatch messages by default?

At what scope does RabbitMQ guarantee message ordering?

A consumer loses its connection while holding unacked messages. What happens to them?

What is a Dead Letter Exchange (DLX) used for?

In the management UI, what does a message in a queue's "Unacked" column mean?

What are vhosts in RabbitMQ used for?

Intermediate level — 26 RabbitMQ questions

With the topic exchange below, which queue receives the message?

A message is published to an exchange but its routing key matches no binding. What happens by default?

A consumer hits a malformed message and calls basic.nack with requeue=true. What is the risk?

Why should you set prefetch (basic.qos) when a consumer uses manual ack?

Which settings are required for a message to survive a broker restart? (select all that apply)

Three different services each need ALL events from one producer. How do you design the queues?

A producer uses a confirm channel as below. What does a successful waitForConfirms() mean?

Why should a RabbitMQ consumer be written to be idempotent?

A queue has a DLX configured. Which cases send a message to the DLX? (select all that apply)

You need to retry a message after 30 seconds without installing plugins. What is the common RabbitMQ approach?

What is wrong with this consumer code?

What is the difference between x-message-ttl and x-expires on a queue declaration?

A queue has x-max-length=10000 and is full. By default, what happens when a new message is published?

A queue accumulates millions of backlogged messages and the RabbitMQ node consumes lots of RAM. What is the right broker-level approach?

When should you pick RabbitMQ over Kafka?

Why should an application open one TCP connection with many channels instead of many connections?

In the RPC pattern below, why must the consumer check correlationId on the response?

Queue depth keeps rising during peak hours. Which responses are reasonable?

What problem do RabbitMQ heartbeats solve?

A message must only be processed 5 minutes after being sent. Without the delayed-message plugin, what is the standard approach?

What is a queue declared with { exclusive: true } suited for?

How do publisher confirms differ from AMQP transactions?

What should determine a consumer prefetch value?

An app loses its broker connection during a broker restart. What must the client handle?

How does a lazy queue differ from the default mode?

What does a headers exchange route on, and when should it be used?

13 more Advanced questions for RabbitMQ

Beginner level — 11 RabbitMQ questions

In RabbitMQ, where does a producer send messages?

How do exchange and queue differ in role?

You need to broadcast an event to EVERY bound queue, ignoring routing keys. Which exchange type?

With the following code, which queue receives the message?

A consumer runs the following code and crashes mid-processing. What happens to that message?

Two consumers read from the same queue. How does the broker dispatch messages by default?

At what scope does RabbitMQ guarantee message ordering?

A consumer loses its connection while holding unacked messages. What happens to them?

What is a Dead Letter Exchange (DLX) used for?

In the management UI, what does a message in a queue's "Unacked" column mean?

What are vhosts in RabbitMQ used for?

Intermediate level — 26 RabbitMQ questions

With the topic exchange below, which queue receives the message?

A message is published to an exchange but its routing key matches no binding. What happens by default?

A consumer hits a malformed message and calls basic.nack with requeue=true. What is the risk?

Why should you set prefetch (basic.qos) when a consumer uses manual ack?

Which settings are required for a message to survive a broker restart? (select all that apply)

Three different services each need ALL events from one producer. How do you design the queues?

A producer uses a confirm channel as below. What does a successful waitForConfirms() mean?

Why should a RabbitMQ consumer be written to be idempotent?

A queue has a DLX configured. Which cases send a message to the DLX? (select all that apply)

You need to retry a message after 30 seconds without installing plugins. What is the common RabbitMQ approach?

What is wrong with this consumer code?

What is the difference between x-message-ttl and x-expires on a queue declaration?

A queue has x-max-length=10000 and is full. By default, what happens when a new message is published?

A queue accumulates millions of backlogged messages and the RabbitMQ node consumes lots of RAM. What is the right broker-level approach?

When should you pick RabbitMQ over Kafka?

Why should an application open one TCP connection with many channels instead of many connections?

In the RPC pattern below, why must the consumer check correlationId on the response?

Queue depth keeps rising during peak hours. Which responses are reasonable?

What problem do RabbitMQ heartbeats solve?

A message must only be processed 5 minutes after being sent. Without the delayed-message plugin, what is the standard approach?

What is a queue declared with { exclusive: true } suited for?

How do publisher confirms differ from AMQP transactions?

What should determine a consumer prefetch value?

An app loses its broker connection during a broker restart. What must the client handle?

How does a lazy queue differ from the default mode?

What does a headers exchange route on, and when should it be used?

13 more Advanced questions for RabbitMQ

A consumer hits a malformed message and calls `basic.nack` with requeue=true. What is the risk?

Why should you set prefetch (`basic.qos`) when a consumer uses manual ack?

A producer uses a confirm channel as below. What does a successful `waitForConfirms()` mean?

What is the difference between `x-message-ttl` and `x-expires` on a queue declaration?

A queue has `x-max-length=10000` and is full. By default, what happens when a new message is published?

In the RPC pattern below, why must the consumer check `correlationId` on the response?

What is a queue declared with `{ exclusive: true }` suited for?

A consumer hits a malformed message and calls `basic.nack` with requeue=true. What is the risk?

Why should you set prefetch (`basic.qos`) when a consumer uses manual ack?

A producer uses a confirm channel as below. What does a successful `waitForConfirms()` mean?

What is the difference between `x-message-ttl` and `x-expires` on a queue declaration?

A queue has `x-max-length=10000` and is full. By default, what happens when a new message is published?

In the RPC pattern below, why must the consumer check `correlationId` on the response?

What is a queue declared with `{ exclusive: true }` suited for?