Offset management in Kafka - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

What is the output of this Kafka consumer offset commit code?

Consider the following Kafka consumer code snippet that commits offsets synchronously. What will be printed if the commit succeeds?

Kafka

consumer.subscribe(['my_topic'])
try:
    records = consumer.poll(timeout_ms=1000)
    for record in records:
        print(f"Received: {record.value}")
    consumer.commit()
    print("Offsets committed successfully")
except Exception as e:
    print(f"Commit failed: {e}")

A No output, consumer.poll returns empty

B Commit failed: OffsetOutOfRangeException

C Received: <message_value>\nOffsets committed successfully

D Offsets committed successfully

💡 Hint

The commit() call happens after messages are received and printed.

The consumer polls messages and prints each received message's value. After processing, commit() commits the offsets synchronously. If successful, the message 'Offsets committed successfully' is printed. So the output includes the received message(s) followed by the commit success message.

🧠 Conceptual

intermediate

Which offset reset policy causes the consumer to start from the earliest offset when no committed offset exists?

Kafka consumers can be configured with an offset reset policy. Which of the following values for 'auto.offset.reset' causes the consumer to read from the earliest offset if no committed offset is found?

A "auto"

B "latest"

C "none"

D "earliest"

💡 Hint

Think about starting from the beginning of the topic.

Setting 'auto.offset.reset' to 'earliest' tells the consumer to start reading from the earliest available offset if no committed offset exists. 'latest' starts from the end, 'none' throws an error, and 'auto' is not a valid value.

❓ Predict Output

advanced

What error does this Kafka consumer code raise?

Given the following code snippet, what error will be raised when running it?

Kafka

consumer.assign([TopicPartition('my_topic', 0)])
consumer.seek(TopicPartition('my_topic', 0), 100)
records = consumer.poll(timeout_ms=1000)
for record in records:
    print(record.value)

A TimeoutException: Poll timed out

B No error, prints messages starting from offset 100

C OffsetOutOfRangeException

D IllegalStateException: Consumer not subscribed to topic

💡 Hint

Assigning partitions manually allows seeking to offsets.

Using assign() manually assigns partitions, so the consumer is not subscribed but can seek offsets. If offset 100 exists, poll returns messages from there without error. No subscription error occurs because assign() is used instead of subscribe().

❓ Predict Output

advanced

What is the output of this asynchronous offset commit example?

Examine this Kafka consumer code that commits offsets asynchronously with a callback. What will be printed if the commit fails?

Kafka

def on_commit(err, partitions):
    if err:
        print(f"Commit failed for partitions: {partitions}")
    else:
        print("Commit succeeded")

consumer.subscribe(['my_topic'])
consumer.poll(timeout_ms=1000)
consumer.commit_async(callback=on_commit)

A Commit failed for partitions: [TopicPartition(topic='my_topic', partition=0)]

B Commit succeeded

C No output, commit_async is non-blocking

D SyntaxError due to callback parameter

💡 Hint

The callback prints a message only if there is an error.

If the commit fails, the callback receives an error and prints the failure message with the partitions. If it succeeds, it prints 'Commit succeeded'. The question asks about the failure case, so the failure message is printed.

🧠 Conceptual

expert

How many committed offsets are stored if a consumer group consumes 3 partitions and commits offsets only for 2 partitions?

A Kafka consumer group reads from 3 partitions of a topic. It commits offsets only for 2 partitions and never commits for the third. How many committed offsets will Kafka store for this group?

A 2 committed offsets, only for the partitions committed

B 1 committed offset, the latest among all partitions

C 3 committed offsets, one per partition

D 0 committed offsets, because all partitions must be committed together

💡 Hint

Offsets are stored per partition independently.

Kafka stores committed offsets per partition per consumer group. If offsets are committed only for 2 partitions, only those 2 offsets are stored. The third partition has no committed offset until committed.