Junior Event-Driven & Streaming Interview Questions (50 MCQs)

Question 1

What is the main difference between synchronous and asynchronous communication between services?

Accepted Answer

In sync the caller waits for a reply; in async it sends a message and moves on

Answer

Sync always uses HTTP while async always uses raw TCP

Answer

In async messages are never lost, while in sync they may be lost

Answer

In sync services are decoupled, while async makes them tightly coupled

Question 2

What is the key benefit of an architecture where services talk through a message broker?

Accepted Answer

Reduced temporal coupling; the consumer need not be available at that moment

Answer

Completely removing the need for error handling and retries in services

Answer

Guaranteeing all data is always instantly consistent across all services

Answer

Lower latency for each single request compared to a direct HTTP call

Question 3

In a classic point-to-point message queue with several consumers, who usually receives each message?

Accepted Answer

Only one of the consumers

Answer

All consumers at the same time

Answer

Forever to the consumer that subscribed first

Answer

To no consumer until all of them are ready

Question 4

In the publish/subscribe model, who receives a published event?

Accepted Answer

Every subscriber receives its own copy of the event

Answer

Only the fastest subscriber receives it

Answer

Only the subscriber the publisher picks directly

Answer

The event is only stored and sent to nobody

Question 5

What is the difference between an event and a command in messaging systems?

Accepted Answer

An event states something happened; a command requests that something be done

Answer

An event is always sync, while a command is always sent async

Answer

An event has one receiver, while a command is broadcast to everyone

Answer

An event cannot be stored, while a command is stored in a log

Question 6

What is the role of a message broker in an event-driven architecture?

Accepted Answer

Receiving, storing, and routing messages between producers and consumers

Answer

Executing services' business logic on their behalf

Answer

Replacing each service's primary database for storing state

Answer

Load balancing HTTP requests across instances of a service

Question 7

In pub/sub systems, what is a 'topic'?

Accepted Answer

A named channel where messages of one category are published

Answer

A specific consumer responsible for processing all messages

Answer

A mandatory field in every message body used for authentication

Answer

A table in the producer's database for storing messages

Question 8

What does the fan-out pattern mean in event-driven architecture?

Accepted Answer

One message is delivered to several independent destinations for parallel processing

Answer

Several small messages are combined into one large message and sent together

Answer

Failed messages are moved to a separate queue for inspection

Answer

The consumer slows down the producer by sending a signal

Question 9

What is the main purpose of splitting a topic into multiple partitions (e.g. in Kafka)?

Accepted Answer

Enabling parallel processing and horizontal throughput scaling

Answer

Guaranteeing a global order across all messages in the topic

Answer

Encrypting messages with different keys in each partition

Answer

Automatically removing duplicate messages on write

Question 10

Within what scope does Kafka guarantee message ordering?

Accepted Answer

Within each partition

Answer

Across the whole topic, over all partitions

Answer

Across the whole cluster, over all topics

Answer

There is no ordering guarantee at all

Question 11

If all events for a specific order must be processed in order, what should we do?

Accepted Answer

Use `orderId` as the partition key

Answer

Spread messages across partitions round-robin

Answer

Create a separate topic for every event

Answer

Run more consumers than there are partitions

Question 12

In Kafka, how do consumers in one consumer group split a topic's messages?

Accepted Answer

Each partition is assigned to only one consumer in the group at a time

Answer

Every message is delivered to all consumers in the group

Answer

Each consumer randomly picks messages from any partition

Answer

Only the first consumer works and the rest are always on standby

Question 13

What does 'at-most-once' delivery semantics mean?

Accepted Answer

A message may be lost but is never delivered twice

Answer

A message is never lost but may be delivered twice

Answer

A message is delivered exactly once with no exceptions

Answer

A message is sent to the consumer at most once per day

Question 14

With 'at-least-once' delivery, what must the consumer be prepared for?

Accepted Answer

Receiving duplicate messages

Answer

Losing some of the messages

Answer

Receiving messages with corrupted content

Answer

Receiving messages from other topics

Question 15

What is an 'idempotent consumer'?

Accepted Answer

A consumer for which processing a message many times equals processing it once

Answer

A consumer that requests each message from the broker only once

Answer

A consumer that has no state or database of its own

Answer

A consumer that processes messages without acknowledging them

Question 16

Why is true 'exactly-once' delivery hard in distributed systems?

Accepted Answer

Because network failures and crashes leave the sender unsure if a message was processed

Answer

Because brokers do not allow sending more than one message per second

Answer

Because messages always change during encryption and decryption

Answer

Because consumers can only read messages in batches

Question 17

What does a consumer's acknowledgement (ack) tell the broker?

Accepted Answer

The message was processed successfully and can be considered delivered

Answer

The consumer is ready to take the next message exactly one minute later

Answer

The message must immediately be sent to all other consumers too

Answer

The producer must resend the message with a higher priority

Question 18

What is a dead-letter queue (DLQ) used for?

Accepted Answer

Holding messages that failed processing after several attempts

Answer

Keeping successfully processed messages for auditing

Answer

Sending high-priority messages before all others

Answer

Temporarily buffering messages while the producer is offline

Question 19

What is a 'poison message'?

Accepted Answer

A message that fails every time it is processed, e.g. due to an invalid format

Answer

A message containing malicious code to attack the broker

Answer

A message that has waited in the queue longer than allowed

Answer

A message the producer sent twice in a row

Question 20

Why are failed messages usually retried with exponential backoff?

Accepted Answer

To give a struggling dependency time to recover instead of flooding it

Answer

To fully preserve message order across the entire topic

Answer

To automatically remove duplicate messages from the queue

Answer

To get messages to the consumer faster than normal

Question 21

When incoming traffic suddenly spikes, how does a queue between the API and workers help?

Accepted Answer

It absorbs the spike so workers can process at their own steady rate

Answer

It automatically rejects excess requests to keep the system healthy

Answer

It multiplies the number of workers without needing autoscaling

Answer

It reduces each message's processing time by compressing it

Question 22

What is backpressure in messaging and streaming systems?

Accepted Answer

A mechanism for a slow consumer to limit the rate of data production

Answer

Compressing messages to reduce network bandwidth usage

Answer

Sending messages back from consumer to producer for verification

Answer

Automatically increasing partition count as load grows

Question 23

What is a key feature of a distributed log (like Kafka) that traditional queues usually lack?

Accepted Answer

Messages remain after consumption and can be re-read

Answer

Each message is read by only one consumer in the whole system

Answer

Messages are deleted immediately after delivery to a consumer

Answer

The broker tracks per-message ack state for every consumer

Question 24

In Kafka, what is an 'offset'?

Accepted Answer

The sequential number of a message in a partition, marking read position

Answer

The time gap between the producer sending and receiving a message

Answer

The number of messages a consumer receives per batch

Answer

The unique identifier of each broker in the cluster

Question 25

What does 'consumer lag' indicate?

Accepted Answer

The gap between the latest written message and the consumer's last processed one

Answer

The time needed for the consumer to connect to the broker

Answer

The number of consumers in the group that have failed

Answer

The time a message spends traveling between brokers

Question 26

In event sourcing, how is the current state of an entity obtained?

Accepted Answer

By replaying all events recorded for that entity

Answer

By reading the latest updated row in a relational table

Answer

By querying other services and merging their responses

Answer

By keeping state only in memory without any storage

Question 27

What is the core idea of CQRS?

Accepted Answer

Separating the write model (commands) from the read model (queries)

Answer

Using a single shared database for all services

Answer

Caching every query at the CDN layer

Answer

Turning all write operations into synchronous calls

Question 28

When services sync data through events, what kind of consistency do we usually get?

Accepted Answer

Eventual consistency; data converges after a short delay

Answer

Strong consistency; all services always see the same data

Answer

Full linearizability across all services and databases

Answer

No consistency; data never converges

Question 29

How does a webhook work?

Accepted Answer

The source service sends an HTTP request to a registered URL when an event occurs

Answer

The client asks the server for status every few seconds

Answer

The client keeps a permanent bidirectional connection open with the server

Answer

The server writes events to a file for the client to download

Question 30

What is the main downside of short polling for receiving updates?

Accepted Answer

Many empty requests and latency tied to the polling interval

Answer

Requiring browser support for special, new protocols

Answer

Inability to use HTTP and requiring raw TCP

Answer

Being limited to sending data only from client to server

Question 31

How does long polling differ from short polling?

Accepted Answer

The server holds the request open until new data arrives or a timeout

Answer

The client sends requests at longer time intervals

Answer

The connection is upgraded to WebSocket and becomes bidirectional

Answer

The server sends data in batches once per hour

Question 32

What is the main characteristic of WebSocket?

Accepted Answer

A persistent, full-duplex connection between client and server

Answer

A one-way HTTP request that closes after each response

Answer

A server-side message queue for communication between microservices

Answer

A compression format to shrink JSON responses

Question 33

What scenario are Server-Sent Events (SSE) suited for?

Accepted Answer

One-way streaming of updates from server to browser over HTTP

Answer

Ultra-low-latency bidirectional communication for multiplayer games

Answer

Uploading large binary files from client to server

Answer

Async communication between microservices via a broker

Question 34

For a chat app where users constantly send and receive messages, which browser transport fits best?

Accepted Answer

WebSocket, since it offers low-latency two-way messaging on one connection

Answer

Short polling every 30 seconds, since it is simpler to implement

Answer

Webhooks, since the server can call the browser directly

Answer

Email, since it always guarantees message delivery

Question 35

What problem does the saga pattern solve?

Accepted Answer

Managing a business transaction spread across services with separate databases

Answer

Speeding up read queries using a distributed cache

Answer

Distributing request load across several instances of a service

Answer

Encrypting messages while they pass through the broker

Question 36

In a choreography-based saga, how are the steps coordinated?

Accepted Answer

Each service reacts to events and publishes the next event

Answer

A central service tells each service what to do

Answer

All services participate in one shared database transaction

Answer

The user manually approves and triggers each step

Question 37

In an orchestration-based saga, who controls the flow?

Accepted Answer

A central orchestrator that sends commands to services

Answer

Each service independently, unaware of the others

Answer

The message broker using its internal routing rules

Answer

A shared database via triggers and stored procedures

Question 38

What is a compensating transaction in a saga?

Accepted Answer

An operation that semantically undoes the effect of a previously completed step

Answer

An automatic database rollback that reverts all services at once

Answer

A transaction that runs only when all steps succeed

Answer

A transaction that refunds message processing cost to the consumer

Question 39

What is the 'dual write' problem? (writing to the database and then publishing an event to a broker)

Accepted Answer

One may succeed and the other fail, leaving data inconsistent

Answer

Data is stored twice in the database, wasting space

Answer

The broker sends the message twice to all consumers

Answer

Two services lock the same record and deadlock occurs

Question 40

What is Change Data Capture (CDC)?

Accepted Answer

Capturing database changes and publishing them as a stream of events

Answer

Taking full database backups at regular intervals

Answer

Encrypting sensitive data before storing it in the database

Answer

Compressing old tables to reduce database size

Question 41

Why should event schemas be managed carefully like a contract?

Accepted Answer

Because many, sometimes unknown, consumers depend on its structure

Answer

Because brokers always reject events without a schema

Answer

Because the schema determines which partition an event goes to

Answer

Because events cannot be encrypted without a schema

Question 42

Which change to an event schema is usually safe for existing consumers?

Accepted Answer

Adding a new optional field with a default value

Answer

Renaming a widely used existing field

Answer

Changing a field's type from number to string

Answer

Removing a required field that consumers read

Question 43

What is the difference between stream processing and batch processing?

Accepted Answer

Stream processes data continuously as it arrives; batch works on bounded sets periodically

Answer

Stream is only for small data and batch is only for big data

Answer

Stream is always more accurate than batch since it reads data twice

Answer

Batch needs a broker while stream reads directly from files

Question 44

What is a tumbling window in stream processing?

Accepted Answer

Fixed-size, non-overlapping time windows, like every 1 minute

Answer

Windows that advance by one event with every new event

Answer

Windows that close after a period of user inactivity

Answer

A single window that stays open from stream start forever

Question 45

What is the difference between event time and processing time?

Accepted Answer

Event time is when it actually happened; processing time is when the system processes it

Answer

Event time is when it was sent to the broker; processing time is when it hit disk

Answer

Both are always equal unless server clocks are out of sync

Answer

Event time applies only to batch and processing time only to streaming

Question 46

Why do we include a correlation ID in messages?

Accepted Answer

To trace one workflow across multiple services and async messages

Answer

To decide which partition the message is stored in

Answer

To encrypt the message with a per-user key

Answer

To raise the message's priority in the broker queue

Question 47

In queues like Amazon SQS, what does the 'visibility timeout' do?

Accepted Answer

It temporarily hides a received message from other consumers

Answer

It permanently deletes messages older than a set duration

Answer

It sets how long the producer waits for broker confirmation

Answer

It limits how long a consumer's connection stays open

Question 48

What is the simplest way to increase processing throughput of a queue?

Accepted Answer

Adding more consumers that compete on the same queue

Answer

Increasing each message's size to carry more information

Answer

Turning the queue into a topic with a single subscriber

Answer

Disabling acks so messages are removed faster

Question 49

What is the difference between 'event notification' and 'event-carried state transfer'?

Accepted Answer

The first just signals a change with an ID; the second carries the full needed data

Answer

The first is sent over HTTP and the second only through Kafka

Answer

The first is used for commands and the second for queries

Answer

The first is stored while the second is deleted after delivery

Question 50

What is needed so messages are not lost when the broker restarts?

Accepted Answer

Messages must be persistent and written to disk (ideally several replicas)

Answer

Messages must be kept only in broker memory with lots of RAM

Answer

Consumers must ack messages before processing them

Answer

The producer must send messages without waiting for confirmation