在Docker Compose中部署Jaeger（不使用all-in-one镜像）

错误信息 #

traces export: context deadline exceeded: rpc error: code = Unavailable desc = connection error: desc = "transport: Error while dialing: dial tcp 172.18.0.10:4317: connect: connection refused"

我的Docker Compose配置 #

version: '3'
services:
  service-a:
    build:
      context: .
      dockerfile: service-a/Dockerfile
    networks:
      - service-jaeger
    ports:
      - "8081:8081"
    environment:
      - OUTBOUND_HOST_PORT=service-b:8082
      - OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger-collector:4317
      - OTEL_EXPORTER_OTLP_INSECURE=true

  service-b:
    build:
      context: .
      dockerfile: service-b/Dockerfile
    networks:
      - service-jaeger
    environment:
      - OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger-collector:4317
      - OTEL_EXPORTER_OTLP_INSECURE=true

  # ElasticSearch 作为追踪和日志存储
  elasticsearch:
    image: docker.elastic.co/elasticsearch/elasticsearch:8.11.3
    networks:
      - service-jaeger
    environment:
      - "ES_JAVA_OPTS=-Xms2g -Xmx2g"
      - "bootstrap.memory_lock=true"
      - "discovery.type=single-node"
      - "xpack.security.enabled=false"
      - "xpack.security.enrollment.enabled=false"
    ports:
      - "127.0.0.1:9200:9200"
    restart: on-failure
    volumes:
      - esdata:/usr/share/elasticsearch/data
    ulimits:
      memlock:
        soft: -1
        hard: -1
      nofile:
        soft: 65536
        hard: 65536

  # Apache Kafka 作为临时存储和追踪数据流处理系统
  kafka:
    image: confluentinc/cp-kafka:5.2.1
    networks:
      - service-jaeger
    ports:
      - "9092:9092"
    environment:
      KAFKA_ZOOKEEPER_CONNECT: "zookeeper:2181"
      KAFKA_ADVERTISED_LISTENERS: "PLAINTEXT://kafka:9092"
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: 1
      KAFKA_BROKER_ID: 1
    restart: on-failure

  # Apache Zookeeper 用于协调Kafka服务
  zookeeper:
    image: confluentinc/cp-zookeeper:5.2.1
    networks:
      - service-jaeger
    ports:
      - "2181:2181"
    environment:
      ZOOKEEPER_CLIENT_PORT: "2181"

  # Jaeger Collector 接收Agent的追踪数据并发送到Kafka
  jaeger-collector:
    image: jaegertracing/jaeger-collector:1.11.0
    hostname: jaeger-collector
    ports:
      - "14269:14269"
      - "14268:14268"
      - "14267:14267"
      - "9411:9411"
    networks:
      - service-jaeger
    restart: on-failure
    environment:
      LOG_LEVEL: "info"
      SPAN_STORAGE_TYPE: "kafka"
      KAFKA_BROKERS: "kafka:9092"
      METRICS_BACKEND: "prometheus"
    depends_on:
      - elasticsearch

  # Jaeger Ingester 从Kafka接收追踪数据并写入ElasticSearch
  jaeger-ingester:
    image: jaegertracing/jaeger-ingester:1.11.0
    networks:
      - service-jaeger
    ports:
      - "14270:14270"
      - "14271:14271"
    restart: on-failure
    environment:
      LOG_LEVEL: "info"
      INGESTER_PARALLELISM: "1"
      INGESTER_DEADLOCKINTERVAL: "0ms"
      SPAN_STORAGE_TYPE: "elasticsearch"
      ES_SERVER_URLS: "http://elasticsearch:9200"
      KAFKA_BROKERS: "kafka:9092"
      METRICS_BACKEND: "prometheus"
    depends_on:
      - kafka

  # Jaeger Agent 本地接收客户端追踪数据并转发到Collector
  jaeger-agent:
    image: jaegertracing/jaeger-agent:1.11.0
    hostname: jaeger-agent
    command: ["--collector.host-port=jaeger-collector:14267"]
    ports:
      - "5775:5775/udp"
      - "6831:6831/udp"
      - "6832:6832/udp"
      - "5778:5778"
    networks:
      - service-jaeger
    restart: on-failure
    environment:
      SPAN_STORAGE_TYPE: "elasticsearch"
      METRICS_BACKEND: "prometheus"
    depends_on:
      - jaeger-collector

  # Jaeger Query 用于查询追踪数据的UI服务
  jaeger-query:
    image: jaegertracing/jaeger-query:1.11.0
    environment:
      SPAN_STORAGE_TYPE: "elasticsearch"
      no_proxy: "localhost"
      METRICS_BACKEND: "prometheus"
    ports:
      - "16686:16686"
      - "16687:16687"
    networks:
      - service-jaeger
    restart: on-failure
    command: [
      "--es.server-urls=http://elasticsearch:9200",
      "--span-storage.type=elasticsearch",
      "--log-level=debug",
      "--query.ui-config=/usr/share/jaeger-query/jaeger-query-config.json"
    ]
    volumes:
      - ./jaeger-query-config.json:/usr/share/jaeger-query/jaeger-query-config.json:ro
    depends_on:
      - jaeger-agent

volumes:
  esdata:

networks:
  service-jaeger:
    driver: bridge

核心问题 #

你使用的Jaeger Collector版本为1.11.0，该版本过于老旧，不支持OpenTelemetry标准的OTLP协议4317端口（Jaeger从1.20+版本才开始默认启用OTLP的gRPC端口4317）。你的服务配置了通过该端口发送追踪数据，但Collector未监听此端口，因此出现连接拒绝错误。

修复步骤 #

1. 升级Jaeger组件版本（推荐方案） #

将所有Jaeger相关服务（collector、ingester、agent、query）的镜像版本从1.11.0升级到**1.40+**的稳定版本（例如1.50.0）。新版本默认支持OTLP的gRPC端口4317和HTTP端口4318，无需额外配置即可兼容你的服务的OTLP exporter设置。

修改后的Jaeger Collector配置示例：

jaeger-collector:
  image: jaegertracing/jaeger-collector:1.50.0
  hostname: jaeger-collector
  ports:
    - "14269:14269"
    - "14268:14268"
    - "14267:14267"
    - "9411:9411"
    - "4317:4317"  # 显式映射OTLP gRPC端口，新版本默认已开启
    - "4318:4318"  # 可选：OTLP HTTP端口
  networks:
    - service-jaeger
  restart: on-failure
  environment:
    LOG_LEVEL: "info"
    SPAN_STORAGE_TYPE: "kafka"
    KAFKA_BROKERS: "kafka:9092"
    METRICS_BACKEND: "prometheus"
  depends_on:
    - kafka  # 修正依赖：Collector使用Kafka存储，应依赖Kafka而非ElasticSearch

2. 不升级版本的替代方案 #

若暂时无法升级，需修改服务的OTLP配置，改为使用Jaeger 1.11.0支持的协议端口：

• 将服务的OTEL_EXPORTER_OTLP_ENDPOINT改为http://jaeger-collector:14268（Jaeger Thrift HTTP端口）
• 添加协议配置OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf（适配OpenTelemetry SDK的协议转换）

修改后的service-a环境变量示例：

environment:
  - OUTBOUND_HOST_PORT=service-b:8082
  - OTEL_EXPORTER_OTLP_ENDPOINT=http://jaeger-collector:14268
  - OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
  - OTEL_EXPORTER_OTLP_INSECURE=true

3. 修正依赖关系错误 #

当前jaeger-collector的depends_on设置为elasticsearch，但Collector配置的存储类型是Kafka，这会导致Collector启动时Kafka可能未就绪，需改为依赖kafka。

内容的提问来源于stack exchange，提问作者bdparrish