Kafka KRaft: 1 out of 3 brokers is always down

I'm trying to create a cluster with KRaft which has 3 brokers and 3 controllers. But each time one broker is going down while the previous downed one is getting up.

This is my docker configs:

version: "3.7"
services:
  controller-1:
    image: confluentinc/cp-kafka:latest
    hostname: controller-1
    container_name: controller-1
    environment:
      KAFKA_NODE_ID: 1
      KAFKA_PROCESS_ROLES: controller
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@controller-1:9093,2@controller-2:9093,3@controller-3:9093"
      KAFKA_LISTENERS: CONTROLLER://0.0.0.0:9093
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: CONTROLLER:PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER # Set your generated UUID her
    volumes:
      - controller_1_data:/var/lib/kafka/data
    user: 1000:1000
    networks:
      - kafka-connector
    restart: always

  controller-2:
    image: confluentinc/cp-kafka:latest
    hostname: controller-2
    container_name: controller-2
    environment:
      KAFKA_NODE_ID: 2
      KAFKA_PROCESS_ROLES: controller
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@controller-1:9093,2@controller-2:9093,3@controller-3:9093"
      KAFKA_LISTENERS: CONTROLLER://0.0.0.0:9093
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: CONTROLLER:PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
    volumes:
      - controller_2_data:/var/lib/kafka/data
    networks:
      - kafka-connector
    restart: always

  controller-3:
    image: confluentinc/cp-kafka:latest
    hostname: controller-3
    container_name: controller-3
    environment:
      KAFKA_NODE_ID: 3
      KAFKA_PROCESS_ROLES: controller
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@controller-1:9093,2@controller-2:9093,3@controller-3:9093"
      KAFKA_LISTENERS: CONTROLLER://0.0.0.0:9093
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: CONTROLLER:PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
    volumes:
      - controller_3_data:/var/lib/kafka/data
    networks:
      - kafka-connector
    restart: always

  kafka-1:
    image: confluentinc/cp-kafka:latest
    hostname: kafka-1
    container_name: kafka-1
    ports:
      - 29092:29092
      - 9092:9092
    environment:
      KAFKA_BROKER_ID: 11
      KAFKA_NODE_ID: 11
      KAFKA_PROCESS_ROLES: broker
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://192.168.18.2:9092
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONTROLLER:PLAINTEXT
      KAFKA_CONTROLLER_QUORUM_VOTERS: "2@controller-2:9093,3@controller-3:9093"
      KAFKA_INTER_BROKER_LISTENER_NAME: PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 3
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: 2
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: 3
    volumes:
      - kafka_1_data:/var/lib/kafka/data
    networks:
      - kafka-connector
    restart: always

  kafka-2:
    image: confluentinc/cp-kafka:latest
    hostname: kafka-2
    container_name: kafka-2
    ports:
      - 29093:29092
      - 9094:9092
    environment:
      KAFKA_BROKER_ID: 12
      KAFKA_NODE_ID: 12
      KAFKA_PROCESS_ROLES: broker
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://192.168.18.2:9094
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONTROLLER:PLAINTEXT
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@controller-1:9093,3@controller-3:9093"
      KAFKA_INTER_BROKER_LISTENER_NAME: PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 3
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: 2
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: 3
    volumes:
      - kafka_2_data:/var/lib/kafka/data
    networks:
      - kafka-connector
    restart: always

  kafka-3:
    image: confluentinc/cp-kafka:latest
    hostname: kafka-3
    container_name: kafka-3
    ports:
      - 29094:29092
      - 9096:9092
    environment:
      KAFKA_BROKER_ID: 13
      KAFKA_NODE_ID: 13
      KAFKA_PROCESS_ROLES: broker
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://192.168.18.2:9096
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONTROLLER:PLAINTEXT
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@controller-1:9093,2@controller-2:9093"
      KAFKA_INTER_BROKER_LISTENER_NAME: PLAINTEXT
      KAFKA_LOG_DIRS: /var/lib/kafka/data
      CLUSTER_ID: KixvqJ76Qn-xLPPrSfBQSw
      KAFKA_CONTROLLER_LISTENER_NAMES: CONTROLLER
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 3
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: 2
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: 3
    volumes:
      - kafka_3_data:/var/lib/kafka/data
    networks:
      - kafka-connector
    restart: always

  debezium:
    image: debezium/connect:2.4
    depends_on:
      - kafka-1
      - kafka-2
      - kafka-3
    environment:
      - BOOTSTRAP_SERVERS=kafka-1:29092,kafka-2:29093,kafka-3:29094
      - GROUP_ID=1
      - CONFIG_STORAGE_TOPIC=my_connect_configs
      - OFFSET_STORAGE_TOPIC=my_connect_offsets
      - STATUS_STORAGE_TOPIC=my_connect_statuses
    ports:
      - 8083:8083
    networks:
      - kafka-connector
    restart: always
    logging:
      driver: "json-file"
      options:
        max-size: "1g"
        max-file: "5"

  kafdrop:
    image: obsidiandynamics/kafdrop
    depends_on:
      - kafka-1
      - kafka-2
      - kafka-3
    ports:
      - 9000:9000
    environment:
      KAFKA_BROKER_CONNECT: kafka-1:9092,kafka-2:9094,kafka-3:9096
    networks:
      - kafka-connector
    restart: always
    logging:
      driver: "json-file"
      options:
        max-size: "1g"
        max-file: "5"
volumes:
  controller_1_data: ~
  controller_2_data: ~
  controller_3_data: ~
  kafka_1_data: ~
  kafka_2_data: ~
  kafka_3_data: ~

networks:
  kafka-connector:
    driver: bridge

Don't mind Debezium atm. I'm only trying to create the cluster and produce some messages.

The error I'm getting from broker-3 is this:

[2024-10-18 11:39:12,347] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:12,420] INFO [RaftManager id=13] Registered the listener org.apache.kafka.image.loader.MetadataLoader@1574972196 (org.apache.kafka.raft.KafkaRaftClient)
[2024-10-18 11:39:12,455] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:12,559] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:12,695] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:12,801] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:12,901] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:13,002] INFO [MetadataLoader id=13] initializeNewPublishers: the loader is still catching up because we still don't know the high water mark yet. (org.apache.kafka.image.loader.MetadataLoader)
[2024-10-18 11:39:13,063] ERROR Encountered fatal fault: Unexpected error in raft IO thread (org.apache.kafka.server.fault.ProcessTerminatingFaultHandler)
java.lang.IllegalStateException: Cannot transition to Follower with leaderId=3 and epoch=80 since it is not one of the voters [1, 2]
    at org.apache.kafka.raft.QuorumState.transitionToFollower(QuorumState.java:381)
    at org.apache.kafka.raft.KafkaRaftClient.transitionToFollower(KafkaRaftClient.java:518)
    at org.apache.kafka.raft.KafkaRaftClient.maybeTransition(KafkaRaftClient.java:1523)
    at org.apache.kafka.raft.KafkaRaftClient.maybeHandleCommonResponse(KafkaRaftClient.java:1473)
    at org.apache.kafka.raft.KafkaRaftClient.handleFetchResponse(KafkaRaftClient.java:1071)
    at org.apache.kafka.raft.KafkaRaftClient.handleResponse(KafkaRaftClient.java:1550)
    at org.apache.kafka.raft.KafkaRaftClient.handleInboundMessage(KafkaRaftClient.java:1676)
    at org.apache.kafka.raft.KafkaRaftClient.poll(KafkaRaftClient.java:2251)
    at kafka.raft.KafkaRaftManager$RaftIoThread.doWork(RaftManager.scala:64)
    at org.apache.kafka.server.util.ShutdownableThread.run(ShutdownableThread.java:127)

If broker-2 is down then the only difference is in this part:

[2024-10-18 11:43:54,404] ERROR Encountered fatal fault: Unexpected error in raft IO thread (org.apache.kafka.server.fault.ProcessTerminatingFaultHandler)
java.lang.IllegalStateException: Cannot transition to Follower with leaderId=2 and epoch=123 since it is not one of the voters [1, 3]

Solution

Each of your Quorum voters should be the same. Let Kafka decide which it chooses. Don't break them into [1, 2], [1, 3], [2, 3]

The error comes from the one missing...

leaderId=3 and epoch=80 since it is not one of the voters [1, 2]

Cannot transition to Follower with leaderId=2 and epoch=123 since it is not one of the voters [1, 3]

You can easily do this in YAML by

KAFKA_CONTROLLER_QUORUM_VOTERS: &kraft-voters "1@controller-1:9093,2@controller-2:9093,3@controller-3:9093"

...

KAFKA_CONTROLLER_QUORUM_VOTERS: *kraft-voters

Also unclear what port 29092 is used for. You can remove those