Partition count strategy in Kafka - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a partition in Kafka?

A partition is a division of a Kafka topic that stores a subset of the messages. It allows Kafka to scale and distribute data across multiple brokers.

beginner

Why is the number of partitions important in Kafka?

The number of partitions affects parallelism, throughput, and fault tolerance. More partitions allow more consumers to read in parallel but can increase complexity.

intermediate

What happens if you increase the partition count after topic creation?

You can increase partitions, but existing data stays in old partitions. New partitions start empty. This can affect message ordering and consumer logic.

beginner

How does Kafka assign messages to partitions by default?

Kafka uses a partitioner that assigns messages based on the message key's hash. If no key is provided, it uses round-robin to distribute messages evenly.

intermediate

What is a good strategy to decide the number of partitions?

Consider expected throughput, number of consumers, and hardware limits. Start with enough partitions to allow parallelism but not too many to avoid overhead.

What does increasing the number of partitions in a Kafka topic do?

A Reduces the topic size

B Deletes old messages automatically

C Allows more consumers to read in parallel

D Prevents message duplication

More partitions allow more consumers to read messages in parallel, increasing throughput.

If a Kafka message has no key, how is its partition chosen?

A Round-robin distribution across partitions

B Always assigned to partition 0

C Based on the message timestamp

D Randomly assigned to any partition

Kafka uses round-robin to evenly distribute messages without keys across partitions.

Can you decrease the number of partitions in a Kafka topic after creation?

A Yes, anytime without issues

B No, Kafka does not support decreasing partitions

C Only if the topic is empty

D Only with a broker restart

Kafka does not allow decreasing the number of partitions once a topic is created.

What is a risk of having too many partitions in Kafka?

A Increased overhead and complexity

B Consumers cannot connect

C Messages get lost

D Lower parallelism

Too many partitions increase overhead for brokers and consumers, making management harder.

Which factor is NOT important when deciding partition count?

A Expected message throughput

B Broker hardware limits

C Number of consumer instances

D Color of the server hardware

The color of server hardware does not affect partition count decisions.

Explain how Kafka uses partitions to improve message processing.

Think about how dividing data helps multiple consumers work at the same time.

Describe the considerations when choosing the number of partitions for a Kafka topic.

Balance performance needs with system complexity.