# Brokers, Replication and KRaft — Apache Kafka: Event Streaming from Basics to Production

Source: https://www.geekswithgeeks.com/en/kafka/kb-brokers-replication

> Explain how a cluster of brokers stores partition replicas with a leader and followers.

## Copies for safety

A Kafka **cluster** is a set of **brokers** (servers). Each partition has a **replication factor** (commonly 3): one broker holds the **leader** replica that serves all reads and writes for that partition, and the others hold **follower** replicas that copy it. The followers that are fully caught up form the **in-sync replicas (ISR)**. If the leader fails, one ISR member becomes the new leader. Cluster metadata (which broker leads which partition) is managed by **KRaft**, Kafka's own consensus protocol; ZooKeeper was removed in Kafka 4.0, so new clusters use KRaft.

## Partition layout from a real broker

I started a single-node Kafka 3.9.0 (KRaft) in Docker and ran `kafka-topics.sh --describe`. Each partition shows its leader, replicas and ISR. With one broker, replication factor is 1, so each partition is led by broker 1 with only itself in the ISR. In a real cluster you would see several broker IDs.

```bash
kafka-topics.sh --bootstrap-server localhost:9092 --describe --topic orders
```

Output:

```
Topic: orders	TopicId: aWjqjntwSYmI_PLyTqQcLA	PartitionCount: 6	ReplicationFactor: 1	Configs: segment.bytes=1073741824
	Topic: orders	Partition: 0	Leader: 1	Replicas: 1	Isr: 1
	Topic: orders	Partition: 1	Leader: 1	Replicas: 1	Isr: 1
```

## Use replication factor 3 in production

A factor of 1 (fine for a laptop) means losing that broker loses the data. Three replicas across different racks or availability zones is the common production baseline.

**Quiz:** What is the ISR?

- [ ] The producer buffer
- [ ] The list of consumers
- [ ] A kind of topic
- [x] The replicas that are fully caught up with the leader

*Answer:* The replicas that are fully caught up with the leader. In-sync replicas are eligible to take over as leader without losing acknowledged data.