Skip to content

Latest commit

 

History

History
58 lines (42 loc) · 2.53 KB

README.md

File metadata and controls

58 lines (42 loc) · 2.53 KB
Raw

Local Docker Kafka Cluster

docker pulls docker stars

Creates a Docker-based Kafka cluster with docker-compose for local development of Kafka-based apps.

Prerequisites

Works for Linux and Mac (with the latest Docker for Mac), and uses docker compose file syntax version 2.

Also, the ports 2181, 9092, 9093, and 9094 (for a default three broker setup) should be available.

Running

Bring up the cluster and specify that you want e.g. five brokers (the default cluster size is three brokers) with

./cluster-up 5

This will create a docker-compose.yml file in the directory that will be used by ./cluster-down do delete the cluster (i.e. this stops and removes all data). If you do not want to delete data, use the docker-compose commands directly.

Connectivity

The cluster will contain one ZooKeeper node only accessible as localhost:2181. The brokers will be reachable under localhost:9092,localhost:9093,... and advertise Docker host IP to Kafka clients. As this is reachable from the host and from the Docker network, it is possible to connect to the cluster both from the host machine as well as from other Docker containers.

The Thing With the Docker Network and Kafka

Kafka Brokers advertise themselves to each other and clients via the Kafka protocol, and announce hostname/IP and port. It is therefore necessary to have fixed (distinct) ports, and have a hostname or IP that is resolvable or routable from both the docker network (for broker-to-broker communication) and from the docker host (for local development).

To achieve this for both Linux and Mac - in the face of some verified existing issues with Mac - the best way to go is to use the docker host IP. Without that issue, it would be more sensible to use the docker container IPs.

Using the docker host IP as advertised broker address makes broker-to-broker traffic pass through the docker host, but this overhead is considered an acceptable price for a solution working both on Linux and Mac.