Docker环境下Airflow连接Kafka broker:29092失败的解决方法

问题场景 #

用户部署了包含Zookeeper、Kafka Broker、Schema Registry及Control Center的Docker服务，使用Docker中的Airflow尝试连接Kafka时出现DNS解析失败及元数据更新超时错误。

部署的docker-compose.yml配置 #

version: '3'

services:
  zookeeper:
      image: confluentinc/cp-zookeeper:7.5.0
      hostname: zookeeper
      container_name: zookeeper
      ports:
        - "2181:2181"
      environment:
        ZOOKEEPER_CLIENT_PORT: 2181
        ZOOKEEPER_TICK_TIME: 2000
      healthcheck:
        test: ['CMD', 'bash', '-c', "echo 'ruok' | nc localhost 2181"]
        interval: 10s
        timeout: 5s
        retries: 5
      # networks:
      #   - confluent

  broker:
    image: confluentinc/cp-server:7.5.0
    hostname: broker
    container_name: broker
    depends_on:
      zookeeper:
        condition: service_healthy
    ports:
      - "9092:9092"
      - "9101:9101"
      - "29092:29092"
    environment:
      KAFKA_BROKER_ID: 1
      KAFKA_ZOOKEEPER_CONNECT: 'zookeeper:2181'
      KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://broker:29092,CONNECTIONS_FROM_HOST://localhost:9092
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONNECTIONS_FROM_HOST:PLAINTEXT
      KAFKA_METRIC_REPORTERS: io.confluent.metrics.reporter.ConfluentMetricsReporter
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_GROUP_INITIAL_REBALANCE_DELAY_MS: 0
      KAFKA_CONFLUENT_LICENSE_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_CONFLUENT_BALANCER_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: 1
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: 1
      KAFKA_JMX_PORT: 9101
      KAFKA_JMX_HOSTNAME: localhost
      KAFKA_CONFLUENT_SCHEMA_REGISTRY_URL: http://schema-registry:8081
      CONFLUENT_METRICS_REPORTER_BOOTSTRAP_SERVERS: broker:29092
      CONFLUENT_METRICS_REPORTER_TOPIC_REPLICAS: 1
      CONFLUENT_METRICS_ENABLE: 'true'
      CONFLUENT_SUPPORT_CUSTOMER_ID: 'anonymous'
    # networks:
    #   - confluent
    healthcheck:
      test: [ "CMD", "bash", "-c", 'nc -z localhost 9092' ]
      interval: 10s
      timeout: 5s
      retries: 5

  schema-registry:
    image: confluentinc/cp-schema-registry:7.5.0
    hostname: schema-registry
    container_name: schema-registry
    depends_on:
      broker:
        condition: service_healthy
    ports:
      - "8081:8081"
    environment:
      SCHEMA_REGISTRY_HOST_NAME: schema-registry
      SCHEMA_REGISTRY_KAFKASTORE_BOOTSTRAP_SERVERS: 'broker:29092'
      SCHEMA_REGISTRY_LISTENERS: http://0.0.0.0:8081
    # networks:
    #   - confluent
    healthcheck:
      test: [ "CMD", "curl", "-f", "http://localhost:8081/" ]
      interval: 30s
      timeout: 10s
      retries: 5

  control-center:
    image: confluentinc/cp-enterprise-control-center:7.5.0
    hostname: control-center
    container_name: control-center
    depends_on:
      broker:
        condition: service_healthy
      schema-registry:
        condition: service_healthy
    ports:
      - "9021:9021"
    environment:
      CONTROL_CENTER_BOOTSTRAP_SERVERS: 'broker:29092'
      CONTROL_CENTER_SCHEMA_REGISTRY_URL: "http://schema-registry:8081"
      CONTROL_CENTER_REPLICATION_FACTOR: 1
      CONTROL_CENTER_INTERNAL_TOPICS_PARTITIONS: 1
      CONTROL_CENTER_MONITORING_INTERCEPTOR_TOPIC_PARTITIONS: 1
      CONFLUENT_METRICS_TOPIC_REPLICATION: 1
      # CONFLIENT_METRICS_ENABLE: 'false'
      PORT: 9021
    # networks:
    #   - confluent
    healthcheck:
      test: [ "CMD", "curl", "-f", "http://localhost:90921/health" ]
      interval: 30s
      timeout: 10s
      retries: 5

Airflow连接Kafka的代码 #

producer = KafkaProducer(bootstrap_servers=['broker:29092'], max_block_ms=5000, api_version=(0, 10, 2))

报错日志 #

ERROR - An error occured: KafkaTimeoutError: Failed to update metadata after 5.0 secs.

[2024-01-30, 16:26:38 UTC] {conn.py:1527} WARNING - DNS lookup failed for broker:29092, exception was [Errno -3] Temporary failure in name resolution. Is your advertised.listeners (called advertised.host.name before Kafka 9) correct and resolvable?

[2024-01-30, 16:26:38 UTC] {conn.py:315} ERROR - DNS lookup failed for broker:29092 (0)

解决方法 #

1. 让Airflow与Kafka容器处于同一Docker网络 #

DNS解析失败的核心原因是Airflow容器无法识别broker主机名，两者不在同一Docker网络中。

• 步骤1：修复Kafka的docker-compose.yml网络配置
  取消所有服务中# networks: - confluent的注释，并在文件顶部添加自定义网络：

  version: '3'
  
  networks:
    confluent:
      name: confluent-network
  
  services:
    zookeeper:
      # ... 其他配置不变
      networks:
        - confluent
    broker:
      # ... 其他配置不变
      networks:
        - confluent
    schema-registry:
      # ... 其他配置不变
      networks:
        - confluent
    control-center:
      # ... 其他配置不变
      networks:
        - confluent
  

  重新部署Kafka服务：docker-compose down && docker-compose up -d

• 步骤2：将Airflow容器加入同一网络
  执行命令将Airflow容器连接到confluent-network：

  docker network connect confluent-network <你的Airflow容器名称>
  

  如果Airflow也是通过docker-compose部署的，直接在其docker-compose.yml中添加同一个网络配置即可。

2. 备选方案：通过宿主机IP访问Kafka #

如果不想调整网络，可以修改Kafka的监听配置，让Airflow通过宿主机IP连接：

• 修改broker服务的环境变量：

  KAFKA_ADVERTISED_LISTENERS: PLAINTEXT://broker:29092,CONNECTIONS_FROM_HOST://localhost:9092,CONNECTIONS_FROM_AIRFLOW://<你的宿主机IP>:9092
  KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: PLAINTEXT:PLAINTEXT,CONNECTIONS_FROM_HOST:PLAINTEXT,CONNECTIONS_FROM_AIRFLOW:PLAINTEXT
  

• 更新Airflow连接代码，替换<你的宿主机IP>为实际IP：

  producer = KafkaProducer(bootstrap_servers=['<你的宿主机IP>:9092'], max_block_ms=5000, api_version=(0, 10, 2))
  

3. 修复Control Center的健康检查错误 #

注意到docker-compose.yml中Control Center的健康检查URL写错了，http://localhost:90921/health应改为http://localhost:9021/health，否则Control Center的健康检查会持续失败，建议同步修正。

内容的提问来源于stack exchange，提问作者Phát Minh Nguyễn