How To Set Up Kafka Without Zookeeper using Docker Compose?

The image is a featured image for the Kafka Without Zookeeper using Docker Compose article and presents the Apache Kafka logo in the foreground and leds similar to clusters in the blurred background.

1. Introduction

In this last of the 3 articles about setting up Apache Kafka, I would like to show you how to set up Apache Kafka Without Zookeeper using Docker Compose.

After reading it, you will know precisely:

  • what does the Docker Compose configuration look like to set up 3 Kafka Brokers,
  • what environment variables we have to set,
  • and what workaround we have to do in order to make it run.

If after finishing it, you will still be eager to broaden your knowledge, then you can find the previous two articles right here:

In my articles, I focus on the practice and getting things done.

However, if you would like to get a strong understading of DevOps concepts, which are more and more in demand nowadays, then I highly recommend to check out KodeKloud courses, like this Docker and Kubernetes learning paths.

2. Apache Kafka Without Zookeeper

Before we dive into the practice part, let’s take a second to understand the motivation behind the Zookeeper removal (I’ve already written about Zookeeper responsibilities in the first part, so I’ll skip it here).

To put it simply, the motivation behind it can be summarized with these points:

  • simplifying the deployment and configuration process (Zookeeper is a separate system, with its own syntax, configs, and deployment patterns),
  • scalability improvement,
  • enabling support for more partitions.

The whole change has been introduced for the first time as a KIP-500 (Kafka Improvement Proposal) and has been offered as an early access version with the 2.8.0 release.

If you would like to understand deeply the reasons and arguments behind the Zookeeper removal, I highly encourage you to visit the KIP-500 ticket page linked above.

Make a real progress thanks to practical examples, exercises, and quizzes.

Image presents a Kotlin Course box mockup for "Kotlin Handbook. Learn Through Practice"

3. Kafka Without Zookeeper Docker Compose File

With that being said, let’s take a look at the docker-compose.yml file:

version: '3'
    image: confluentinc/cp-kafka:7.2.1
    container_name: kafka1
      KAFKA_NODE_ID: 1
      KAFKA_LISTENERS: PLAINTEXT://kafka1:9092,CONTROLLER://kafka1:9093
      KAFKA_CONTROLLER_QUORUM_VOTERS: '1@kafka1:9093,2@kafka2:9093,3@kafka3:9093'
      KAFKA_PROCESS_ROLES: 'broker,controller'
      - ./
    command: "bash -c '/tmp/ && /etc/confluent/docker/run'"
    image: confluentinc/cp-kafka:7.2.1
    container_name: kafka2
      KAFKA_NODE_ID: 2
      KAFKA_LISTENERS: PLAINTEXT://kafka2:9092,CONTROLLER://kafka2:9093
      KAFKA_CONTROLLER_QUORUM_VOTERS: '1@kafka1:9093,2@kafka2:9093,3@kafka3:9093'
      KAFKA_PROCESS_ROLES: 'broker,controller'
      - ./
    command: "bash -c '/tmp/ && /etc/confluent/docker/run'"
    image: confluentinc/cp-kafka:7.2.1
    container_name: kafka3
      KAFKA_NODE_ID: 3
      KAFKA_LISTENERS: PLAINTEXT://kafka3:9092,CONTROLLER://kafka3:9093
      KAFKA_CONTROLLER_QUORUM_VOTERS: '1@kafka1:9093,2@kafka2:9093,3@kafka3:9093'
      KAFKA_PROCESS_ROLES: 'broker,controller'
      - ./
    command: "bash -c '/tmp/ && /etc/confluent/docker/run'"

Don’t worry if you feel overwhelmed at this point, I will walk you step by step through the configuration values in the following paragraphs.

Nevertheless, don’t run the docker compose up command yet, we will need one more file to run it ๐Ÿ™‚

4. Add Necessary Workarounds

As the next step, we need to run a few commands when containers start.

In my example, I will call this file And although the name is not important, it has to match values specified when mounting volumes and running the command from within the containers:


sed -i '/KAFKA_ZOOKEEPER_CONNECT/d' /etc/confluent/docker/configure

sed -i 's/cub zk-ready/echo ignore zk-ready/' /etc/confluent/docker/ensure

echo "kafka-storage format --ignore-formatted -t NqnEdODVKkiLTfJvqd1uqQ== -c /etc/kafka/" >> /etc/confluent/docker/ensure

As we can see, the above script contains exactly 3 commands.

The first one is responsible for turning off the check for the KAFKA_ZOOKEEPER_CONNECT parameter. Without this line, the error will be thrown: Command [/usr/local/bin/dub ensure KAFKA_ZOOKEEPER_CONNECT] FAILED !

The second one ignores the cub zk-ready. Without it, everything will end up with another exception: zk-ready: error: too few arguments.

And finally, the last line is a KRaft required step, which formats the storage directory with a new cluster identifier. Please keep in mind that the specified UUID value (NqnEdODVKkiLTfJvqd1uqQ== in my case) has to be 16 bytes of a base64-encoded UUID.

5. Docker Compose Instructions

With all of the above being said, we can get back to the docker-compose.yml file.

5.1. Volumes and Commands

Although I will skip the things like the services, images, or container names (I’ve covered them in the previous articles), let’s take a look at these 3 lines:

  - ./
command: "bash -c '/tmp/ && /etc/confluent/docker/run'"

Volumes are the preferred mechanism for persisting data generated by and used by Docker containers. With the above setting, we inform Docker that we want the file to be mounted inside the tmp directory inside the container with the same name. Of course, we could change the name of this file inside, but then, the change should also be reflected in the command.

When it comes to the command, it simply overrides the default command. With the above config, we instruct Docker to run our container with our custom script.

6. Apache Kafka Brokers Environment Variables

Finally, let’s go through all the environments we had to set in order to make everything work.

6.1. KAFKA_NODE_ID Environment Variable

As the name suggest, the KAFKA_NODE_ID variable is used to specify a unique identifier of our node in the cluster.

6.2. Kafka Listeners and Protocol Mapping

Nextly, let’s see the settings responsible for appropriate hosts and ports mapping:

  • KAFKA_LISTENER_SECURITY_PROTOCOL_MAP – with this one, we simply define the key-value pairs of listeners’ names with security protocols used with them
  • KAFKA_LISTENERS – with this variable, we define all exposed listeners
  • KAFKA_ADVERTISED_LISTENERS – this one, on the other hand, contains all listeners used by clients

It’s worth mentioning here that when working with Docker Compose, the container name becomes a hostname- like kafka1, kafka2, and kafka3 in our examples.


As the next step, let’s see the KAFKA_ADVERTISED_LISTENER. This one, simply allows us to specify a list of coma-separated listeners’ names used by the controller. It’s necessary when running in a KRaft mode.

Additionally, none of the values specified here can be put inside the advertised listeners covered in the previous paragraph. Otherwise, the following error will be thrown:

The advertised.listeners config must not contain KRaft controller listeners from controller.listener.names when process.roles contains the broker role because Kafka clients that send requests via advertised listeners do not send requests to KRaft controllers — they only send requests to KRaft brokers.


As the next one comes the KAFKA_CONTROLLER_QUORUM_VOTERS environment variable. And this one allows us to specify the set of voters in our node. When it comes to the format, values have to be put in the following order: {id}@{host}:{port} .

So, in our case, we do specify 9093 of our nodes, because this is the port exposed as a controller.


And the last one- the KAFKA_PROCESS_ROLES variable lets us specify the role of the particular broker. The value can be set to either broker, controller, or both of them (just like above).

7. Testing

And finally, we can proceed to the part where we will verify if everything is working, as expected. In my example, I will show you how to use Console Producer/Consumer, but feel free to connect with anything you like.

Firstly, let’s start our services with the docker compose up:

The broker has been unfenced. Transitioning from RECOVERY to RUNNING.

After some time, we should see the above for each of our containers.

Additionally, we can validate with a docker ps command:

CONTAINER ID IMAGE                       COMMAND                CREATED            STATUS            PORTS    NAMES
dc3966ba58a4 confluentinc/cp-kafka:7.2.1 "bash -c '/tmp/run_wโ€ฆ" About a minute ago Up About a minute 9092/tcp kafka3
10242d9d1e6a confluentinc/cp-kafka:7.2.1 "bash -c '/tmp/run_wโ€ฆ" About a minute ago Up About a minute 9092/tcp kafka2
c404c866cde6 confluentinc/cp-kafka:7.2.1 "bash -c '/tmp/run_wโ€ฆ" About a minute ago Up About a minute 9092/tcp kafka1

As we can clearly see, every container is up and running. Moreover, the desired 9092 port is exposed for each of them.

When it comes to the testing strategy, I would like to connect to each broker separately and perform a different action:

  • firstly, connect to the kafka1 and create a new topic called exampleTopic
  • secondly, to kafka2 and produce some messages
  • and finally, read all of the messages from within the kafka3

To do so, let’s start with the following:

docker exec -it kafka1 /bin/bash
kafka-topics --bootstrap-server kafka1:9092 --create --topic exampleTopic

# Output:
Created topic exampleTopic.

# To exit the container, please specify exit

As we can see, the exampleTopic was created successfully.

So as the next thing, let’s produce some messages from kafka2 broker:

docker exec -it kafka2 /bin/bash
kafka-console-producer --bootstrap-server kafka2:9092 --topic exampleTopic


# To exit the producer, please hit Ctrl + D. Then to exit the container please type 'exit'.

No errors have been thrown, which let us assume that everything is good so far.

Nevertheless, to fully make sure, let’s consume these messages:

docker exec -it kafka3 /bin/bash
kafka-console-consumer --bootstrap-server kafka3:9092 --topic exampleTopic --from-beginning

# Output:

# To exit the consumer, please hit Ctrl + C. As the output, we should see:
^CProcessed a total of 2 messages

# And again, to exit the container please type 'exit'.

As we can see, everything worked as expected and our brokers are correctly communicating within the cluster.

8. Kafka Without Zookeeper Using Docker Compose Summary

And that’s all for today’s article on how to set up Kafka Without Zookeeper using Docker Compose.

I really hope that you enjoyed this and my previous articles in an Apache Kafka series and I will be forever thankful if you would like to share your feedback with me in the comments section below (positive and negative, as well).

Take care and have a great week! ๐Ÿ™‚

Share this:

Picture of Hi there! ๐Ÿ‘‹

Hi there! ๐Ÿ‘‹

My name is Piotr and I've created Codersee to share my knowledge about Kotlin, Spring Framework, and other related topics through practical, step-by-step guides. Always eager to chat and exchange knowledge.

Related content

7 Responses

  1. Getting this error while running the docker-compose.yaml.

    kafka3 | [2023-01-19 06:52:27,859] ERROR [RaftManager nodeId=3] Unexpected error INCONSISTENT_CLUSTER_ID in VOTE response: InboundResponse(correlationId=2849, data=VoteResponseData(errorCode=104, topics=[]), sourceId=2) (org.apache.kafka.raft.KafkaRaftClient)
    kafka1 | [2023-01-19 06:52:27,859] ERROR [RaftManager nodeId=3] Unexpected error INCONSISTENT_CLUSTER_ID in VOTE response: InboundResponse(correlationId=2849, data=VoteResponseData(errorCode=104, topics=[]), sourceId=2) (org.apache.kafka.raft.KafkaRaftClient)
    kafka2 | [2023-01-19 06:52:27,859] ERROR [RaftManager nodeId=3] Unexpected error INCONSISTENT_CLUSTER_ID in VOTE response: InboundResponse(correlationId=2849, data=VoteResponseData(errorCode=104, topics=[]), sourceId=2) (org.apache.kafka.raft.KafkaRaftClient)

    Please help in resolving it.

    1. Hi!

      If this is environment you don’t care about, I would personally go with docker compose down and prune all the containers, networking and volumes.

      – docker-compose down
      – backup kafka-logs directory using cp kafka-logs -r kafka-logs-bak
      – delete the kafka-logs/ file
      – docker-compose up -d

      If delete does not help, then please check out if this is not related to old containers or something like that

  2. Hi,

    Thanks Piotr for the solution. I tired the running docker-compose after pruning all the containers. But still getting the same issue.

    One change I have made in ur
    echo “kafka-storage format –ignore-formatted -t NqnEdODVKkiLTfJvqd1uqQ== -c /etc/kafka/” >> /etc/confluent/docker/ensure


    echo “kafka-storage format –ignore-formatted -t $(kafka-storage random-uuid) -c /etc/kafka/” >> /etc/confluent/docker/ensure

    I am not sure whether it is fine or not?

    Please help.

  3. Hi, I’m using this tutorial in conjunction with
    So I am following the SpringBoot Kotlin tutorial and I’m failing to connect to the broker, as the application fails to start with the following:
    2023-05-11T19:36:04.909+01:00 WARN 47848 — [ntainer#0-0-C-1] org.apache.kafka.clients.NetworkClient : [Consumer clientId=consumer-0-1, groupId=0] Connection to node -1 (localhost/ could not be established. Broker may not be available.
    2023-05-11T19:36:04.909+01:00 WARN 47848 — [ntainer#0-0-C-1] org.apache.kafka.clients.NetworkClient : [Consumer clientId=consumer-0-1, groupId=0] Bootstrap broker localhost:9001 (id: -1 rack: null) disconnected
    2023-05-11T19:36:05.837+01:00 INFO 47848 — [ntainer#0-0-C-1] org.apache.kafka.clients.NetworkClient : [Consumer clientId=consumer-0-1, groupId=0] Node -1 disconnected.

Leave a Reply

Your email address will not be published. Required fields are marked *


Image presents 3 ebooks with Java, Spring and Kotlin interview questions.

Never miss any important updates from the Kotlin world and get 3 ebooks!

You may opt out any time. Terms of Use and Privacy Policy

Free tutorials and courses on Kotlin & backend

To make Codersee work, we log user data. By using our site, you agree to our Privacy Policy and Terms of Use.