I am trying to use wurstmeister\kafka-docker image with docker-compose, but I am having real problems with connecting everything.

All the posts or questions that I check, seems not to have any problems, but I am frankly lost. (And there are at least two questions in SO that try to address the problem)

I believe that the problem is my poor understanding of the networking of docker. So the problem:

I can consume and produce from the same container of kafka, but, when I try to create another container (or use my laptop with a python client) I got several errors related to the advertised.host.name parameter (in the image this parameter is KAFKA_ADVERTISED_HOST_NAME)

I already try setting this variable in lot of ways, but it simply don't work.

So I am looking for a authorative answer (i.e. how to set automatically those parameters and what is it meaning) how to set the docker-compose.yml

This is mine:

zookeeper:
  image: wurstmeister/zookeeper
  ports:
    - "2181:2181"

kafka:
  image: wurstmeister/kafka
 # hostname: kafka
  ports:
    - "9092"
  links:
    - zookeeper:zk
  environment:
    KAFKA_ADVERTISED_HOST_NAME: "kafka"
    KAFKA_ADVERTISED_PORT: "9092"
    KAFKA_ZOOKEEPER_CONNECT: "zk:2181"

UPDATE

Following the advise of @dnephin, I modified the start-kafka.sh in the following lines:

...
if [[ -z "$KAFKA_ADVERTISED_PORT" ]]; then
    export KAFKA_ADVERTISED_PORT=$(hostname -i)
fi
...

and remove KAFKA_ADVERTISED_HOST_NAME: "kafka" from the docker-compose.yml

I started the containers in the canonical way:

docker-compose up -d

Both of the containers are running:

$ docker-compose ps
           Name                          Command               State                     Ports                    
-----------------------------------------------------------------------------------------------------------------
infraestructura_kafka_1       start-kafka.sh                   Up      0.0.0.0:32768->9092/tcp                    
infraestructura_zookeeper_1   /opt/zookeeper/bin/zkServe ...   Up      0.0.0.0:2181->2181/tcp, 2888/tcp, 3888/tcp 

Afterwards I did:

docker-compose logs

And everything run smoothly.

For checking the ip addresses:

$ KAFKA_IP=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' infraestructura_kafka_1)                                                                                                            
$ echo $KAFKA_IP
172.17.0.4

and

$ ZK_IP=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' infraestructura_zookeeper_1)                                                                                                           
$ echo $ZK_IP 
172.17.0.3

Then I execute in two differents consoles:

A producer:

$ docker run --rm --interactive wurstmeister/kafka /opt/kafka_2.11-0.9.0.1/bin/kafka-console-producer.sh --topic grillo --broker-list 171.17.0.4:9092  

A consumer:

$ docker run --rm --interactive  wurstmeister/kafka /opt/kafka_2.11-0.9.0.1/bin/kafka-console-consumer.sh --topic grillo --from-beginning --zookeeper 172.17.0.3:2181 

Almost immediately, warnings start flying all over the screen:

[2016-03-11 00:39:17,010] WARN Fetching topic metadata with correlation id 0 for topics [Set(grillo)] from broker [BrokerEndPoint(1001,ba53d4fd7595,9092)] failed (kafka.client.ClientUtils$)
java.nio.channels.ClosedChannelException
        at kafka.network.BlockingChannel.send(BlockingChannel.scala:110)
        at kafka.producer.SyncProducer.liftedTree1$1(SyncProducer.scala:75)
        at kafka.producer.SyncProducer.kafka$producer$SyncProducer$$doSend(SyncProducer.scala:74)
        at kafka.producer.SyncProducer.send(SyncProducer.scala:119)
        at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:59)
        at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:94)
        at kafka.consumer.ConsumerFetcherManager$LeaderFinderThread.doWork(ConsumerFetcherManager.scala:66)
        at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:63)
[2016-03-11 00:39:17,013] WARN [console-consumer-79688_9dd5f575d557-1457656747003-f1ed369d-leader-finder-thread], Failed to find leader for Set([grillo,0]) (kafka.consumer.ConsumerFetcherManager$LeaderFin
derThread)
kafka.common.KafkaException: fetching topic metadata for topics [Set(grillo)] from broker [ArrayBuffer(BrokerEndPoint(1001,ba53d4fd7595,9092))] failed
        at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:73)
        at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:94)
        at kafka.consumer.ConsumerFetcherManager$LeaderFinderThread.doWork(ConsumerFetcherManager.scala:66)
        at kafka.utils.ShutdownableThread.run(ShutdownableThread.scala:63)
Caused by: java.nio.channels.ClosedChannelException
        at kafka.network.BlockingChannel.send(BlockingChannel.scala:110)
        at kafka.producer.SyncProducer.liftedTree1$1(SyncProducer.scala:75)
        at kafka.producer.SyncProducer.kafka$producer$SyncProducer$$doSend(SyncProducer.scala:74)
        at kafka.producer.SyncProducer.send(SyncProducer.scala:119)
        at kafka.client.ClientUtils$.fetchTopicMetadata(ClientUtils.scala:59)
        ... 3 more

and so on

In the console of the producer, I wrote some sentences:

$ docker run --rm --interactive klustera/kafka /opt/kafka_2.11-0.9.0.1/bin/kafka-console-producer.sh --topic grillo --broker-list 171.17.0.4:9092                                                           
Hola
¿Cómo estáń?
¿Todo bien?

And a few moments later, I got this response:

[2016-03-11 00:39:28,955] ERROR Error when sending message to topic grillo with key: null, value: 4 bytes with error: Failed to update metadata after 60000 ms. (org.apache.kafka.clients.producer.internals.ErrorLoggingCallback)
[2016-03-11 00:40:28,956] ERROR Error when sending message to topic grillo with key: null, value: 16 bytes with error: Failed to update metadata after 60000 ms. (org.apache.kafka.clients.producer.internals.ErrorLoggingCallback)
[2016-03-11 00:41:28,956] ERROR Error when sending message to topic grillo with key: null, value: 12 bytes with error: Failed to update metadata after 60000 ms. (org.apache.kafka.clients.producer.internals.ErrorLoggingCallback)

And in the docker-compose logs

...
zookeeper_1 | 2016-03-11 00:39:07,072 [myid:] - INFO  [ProcessThread(sid:0 cport:2181)::PrepRequestProcessor@651] - Got user-level KeeperException when processing sessionid:0x153631368b1000b type:create c
xid:0x2 zxid:0x47 txntype:-1 reqpath:n/a Error Path:/consumers Error:KeeperErrorCode = NodeExists for /consumers
zookeeper_1 | 2016-03-11 00:39:07,243 [myid:] - INFO  [ProcessThread(sid:0 cport:2181)::PrepRequestProcessor@651] - Got user-level KeeperException when processing sessionid:0x153631368b1000b type:create c
xid:0x19 zxid:0x4b txntype:-1 reqpath:n/a Error Path:/consumers/console-consumer-79688/owners/grillo Error:KeeperErrorCode = NoNode for /consumers/console-consumer-79688/owners/grillo
zookeeper_1 | 2016-03-11 00:39:07,247 [myid:] - INFO  [ProcessThread(sid:0 cport:2181)::PrepRequestProcessor@651] - Got user-level KeeperException when processing sessionid:0x153631368b1000b type:create $xid:0x1a zxid:0x4c txntype:-1 reqpath:n/a Error Path:/consumers/console-consumer-79688/owners Error:KeeperErrorCode = NoNode for /consumers/console-consumer-79688/owners
...

UPDATE 2

I made it work, at least, in docker-machine:

First, I defined a variable with the name of the docker-machine:

DOCKER_VM=kafka_test

Then, I modify the docker-compose.yml as follows:

KAFKA_ADVERTISED_HOST_NAME: "${DOCKER_MACHINE_IP}"

Lastly, in the environment of the docker-machine, I execute:

DOCKER_MACHINE_IP=$(docker-machine ip $DOCKER_VM) docker-compose up -d

But in the laptop (I mean, without using a virtual machine, it doesn't work)


My solution to this issue is slightly different. I configure Kafka to advertise on kafka host and, because it's exposed on the host machine on localhost:9092, I add an entry in /etc/hosts for kafka to resolve to localhost. By doing this Kafka can be accessed from both other Docker containers and from localhost.

docker-compose.yml:

  my-web-service:
    build: ./my-web-service
    ports:
     - "8000:8000"
    links:
     - kafka
  kafka:
    image: "wurstmeister/kafka:0.10.2.0"
    ports:
     - "9092:9092"
    hostname: kafka
    links: 
     - zookeeper
    environment:
     - KAFKA_ADVERTISED_HOST_NAME=kafka
     - KAFKA_ZOOKEEPER_CONNECT=zookeeper:2181
     - KAFKA_ADVERTISED_PORT=9092
  zookeeper:
    image: wurstmeister/zookeeper
    ports:
      - "2181:2181"

Updated hosts file:

more /etc/hosts
127.0.0.1       localhost kafka

For developing app in localhost, there is a solution in the documentation: "HOSTNAME_COMMAND"

kafka:
  image: wurstmeister/kafka
  ports:
    - 9092:9092
environment:
  KAFKA_ZOOKEEPER_CONNECT: zookeeper:2181
  HOSTNAME_COMMAND: "route -n | awk '/UG[ \t]/{print $$2}'"

Hope this help others...


here's an improved version of @radek1st answer.

links is the old docker way, networks is the current method.

imo, making any sort of system change isn't good and should never be needed. it also kind of defeats the purpose of using Docker.

version: '2.1'

networks:
  sb:
    driver: bridge

services:
  zookeeper:
    image: confluentinc/cp-zookeeper:latest
    container_name: zookeeper
    hostname: zookeeper
    networks:
     - sb
    ports:
      - "2181:2181"
    environment:
      ZOOKEEPER_CLIENT_PORT: 2181
      ZOOKEEPER_TICK_TIME: 2000

  kafka:
    image: confluentinc/cp-kafka:latest
    container_name: kafka
    hostname: ${KAFKA_HOSTNAME:-kafka}
    depends_on:
      - zookeeper
    networks:
     - sb
    ports:
      - "9092:9092"
    environment:
      KAFKA_BROKER_ID: 1
      KAFKA_ZOOKEEPER_CONNECT: zookeeper:2181
      KAFKA_ADVERTISED_HOST_NAME: ${KAFKA_HOSTNAME:-kafka}
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://${KAFKA_HOSTNAME:-kafka}:9092
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1

Then I use the follow bash script to start things. This allows me to override the Kafka hostname for local development. ./startup.sh localhost

#!/bin/bash
echo KAFKA_HOSTNAME=${1:-kafka} > .env
docker-compose up -d

Read more -

  • https://docs.docker.com/compose/networking
  • https://docs.docker.com/compose/environment-variables/
  • https://docs.docker.com/compose/compose-file/#variable-substitution

I believe the value you use for KAFKA_ADVERTISED_HOST_NAME will change depending on how the container can be reached.

If you're trying to connect from another container, using kafka should be correct (as long as you use set that name as the link alias).

If you're trying to connect from the host, that name isn't going to work. You'd need to use the container IP address, which you can get using docker inspect. However the container IP address will change, so it might be better to set this from inside the container using $(hostname -i) to retrieve it.


Just try the following & use service discovery for example this one.

zookeeper:
  image: wurstmeister/zookeeper
  ports:
    - "2181:2181"
kafka:
  build: .
  ports:
    - "9092:9092"
  links:
    - zookeeper:zk
  environment:
    KAFKA_ADVERTISED_HOST_NAME: 192.168.59.103
    KAFKA_ADVERTISED_PORT: 9092
    KAFKA_CREATE_TOPICS: "test:1:1"
  volumes:
    - /var/run/docker.sock:/var/run/docker.sock

Or you use this one:

zookeeper:
  image: wurstmeister/zookeeper
  ports: 
    - "2181"
kafka:
  build: .
  ports:
    - "9092"
  links: 
    - zookeeper:zk
  environment:
    KAFKA_ADVERTISED_HOST_NAME: 192.168.59.103
    DOCKER_HOST: 192.168.59.103:2375
  volumes:
    - /var/run/docker.sock:/var/run/docker.sock