cgswong/kafka repository overview

⁠Dockerized Apache Kafka

This repository holds the build definition and supporting files for building a Docker⁠ image to run Apache Kafka⁠ in containers. The image is available from DockerHub⁠

Configuration is parameterized, enabling a Kafka cluster to be run from multiple container instances.

⁠How to use this image

⁠ZooKeeper

Kafka requires a running ZooKeeper⁠ ensemble in order to manage broker and consumer coordination for its Topics⁠. Consumer offsets also get committed to ZooKeeper⁠ as a way for all consumers in a single 'group' to synchronize with each other.

You should use the predefined ZooKeeper image⁠ to start a ZooKeeper⁠ ensemble prior to attempting to run a Kafka broker. You'll need to start an ensemble that can obtain a fail-over quorum, preferably 3 or 5 nodes.

The image provides a cluster-able Kafka broker. As a minimum the following Docker run environment variables (i.e. docker run -e) must be set:

• kafka_broker_id - Defaults to 1.
• kafka_advertised_host_name - This is given to a consumer by ZooKeeper when connecting to a topic.
• kafka_zookeeper_connect - Comma separated ZooKeeper connection string in the form [hostname/IP]:[port] where [port] defaults to 2181 if none is given. In a shared ZooKeeper environment use a chroot syntax where the directory to store the specific Kafka environment is appended to the last ZooKeeper entry, for example, [hostname1/IP1]:[port],[hostname2/IP2]:[port]/kafka1, where Kafka entries would then go under /kafka1.

So, assuming your Docker host is 172.17.8.101, has ZooKeeper⁠ running and should now run Kafka as well, execute the following:

docker run -d --name kafka1 -e kafka_broker_id=1 -e kafka_advertised_host_name=172.17.8.101 -e kafka_zookeeper_connect=172.17.8.101/kafka cgswong/kafka:latest

⁠Additional configuration

Environment variables are accepted as a means to provide further configuration by reading those starting with kafka_. Any matching variables will get added to Kafka's server.properties file by

1. removing the kafka_ prefix
2. replacing any occurrences of _ with .

For example, an environment variable kafka_num_partitions=3 will result in num.partitions=3 within server.properties. Similarly, to auto create topics when a broker publishes to a non-existent topic you can use an environment variable kafka_auto_create_topics_enable=true which will result in auto.create.topics.enable=true within server.properties with the default replication factor and number of partitions which can also be set similarly.

You can also download a configuration file by setting the special environment variable kafka_cfg_url to the download file location. Within this file you can also take advantage of variable substitution.

Other Kafka settings can also be enabled via Docker run environment variables, for example, -e KAFKA_JMX_OPTS=-Dcom.sun.management.jmxremote=false. A few environment variables of interest:

• GC_LOG_ENABLED: Set to true to enable garbage collection logging.
• JMX_PORT: JMX port for remote monitoring. Defaults to exposed port 19092 in this image.
• KAFKA_GC_LOG_OPTS: Garbage collection logging options. GC_LOG_ENABLED must be enabled.
• KAFKA_HEAP_OPTS: JVM memory settings or heap size, defaults to -Xmx1G -Xms1G (1GB).
• KAFKA_JMX_OPTS: JMX options
• KAFKA_JVM_PERFORMANCE_OPTS: JVM performance options. Defaults to -server -XX:+UseG1GC -XX:MaxGCPauseMillis=20 -XX:InitiatingHeapOccupancyPercent=35 -XX:+DisableExplicitGC -Djava.awt.headless=true
• KAFKA_LOG4J_OPTS: Log4j options
• KAFKA_OPTS: Generic JVM settings that you may want to set.

⁠Issues

If you have any problems with or questions about this image, please contact me through a GitHub issue⁠.