compose.observability.yaml

# Copyright The OpenTelemetry Authors
# SPDX-License-Identifier: Apache-2.0

# Observability stack layer: adds Jaeger, Prometheus, OpenSearch, and Grafana,
# and patches the otel-collector to export to them.
#
# Usage:
#   docker compose -f compose.yaml -f compose.observability.yaml up
#   docker compose -f compose.yaml -f compose.full.yaml -f compose.observability.yaml up

services:
  # Jaeger
  jaeger:
    image: ${JAEGERTRACING_IMAGE}
    container_name: jaeger
    command:
      - "--config=file:/etc/jaeger/config.yml"
    deploy:
      resources:
        limits:
          memory: 1200M
    restart: unless-stopped
    ports:
      - "${JAEGER_UI_PORT}"
      - "${JAEGER_GRPC_PORT}"
    environment:
      - JAEGER_HOST
      - JAEGER_GRPC_PORT
      - PROMETHEUS_ADDR
      - OTEL_COLLECTOR_HOST
      - OTEL_COLLECTOR_PORT_HTTP
      - MEMORY_MAX_TRACES=25000
    volumes:
      - ./src/jaeger/config.yml:/etc/jaeger/config.yml
    logging:
      driver: "json-file"
      options:
        max-size: "5m"
        max-file: "2"

  # Grafana
  grafana:
    image: ${GRAFANA_IMAGE}
    container_name: grafana
    deploy:
      resources:
        limits:
          memory: 175M
    restart: unless-stopped
    environment:
      - "GF_INSTALL_PLUGINS=grafana-opensearch-datasource"
    volumes:
      - ./src/grafana/grafana.ini:/etc/grafana/grafana.ini
      - ./src/grafana/provisioning/:/etc/grafana/provisioning/
    ports:
      - "${GRAFANA_PORT}"
    logging:
      driver: "json-file"
      options:
        max-size: "5m"
        max-file: "2"

  # Prometheus
  prometheus:
    image: ${PROMETHEUS_IMAGE}
    container_name: prometheus
    command:
      - --web.console.templates=/etc/prometheus/consoles
      - --web.console.libraries=/etc/prometheus/console_libraries
      - --storage.tsdb.retention.time=7d
      - --config.file=/etc/prometheus/prometheus-config.yaml
      - --storage.tsdb.path=/prometheus
      - --web.enable-lifecycle
      - --web.route-prefix=/
      - --web.enable-otlp-receiver
      - --enable-feature=exemplar-storage
    volumes:
      - ./src/prometheus/prometheus-config.yaml:/etc/prometheus/prometheus-config.yaml
    deploy:
      resources:
        limits:
          memory: 200M
    restart: unless-stopped
    ports:
      - "${PROMETHEUS_PORT}:${PROMETHEUS_PORT}"
    logging:
      driver: "json-file"
      options:
        max-size: "5m"
        max-file: "2"

  # OpenSearch
  opensearch:
    image: ${IMAGE_NAME}:${DEMO_VERSION}-opensearch
    container_name: opensearch
    build:
      context: ./
      dockerfile: ${OPENSEARCH_DOCKERFILE}
      cache_from:
        - ${IMAGE_NAME}:${IMAGE_VERSION}-opensearch
    deploy:
      resources:
        limits:
          memory: 1G
    restart: unless-stopped
    environment:
      - cluster.name=demo-cluster
      - node.name=demo-node
      - bootstrap.memory_lock=true
      - discovery.type=single-node
      - OPENSEARCH_JAVA_OPTS=-Xms400m -Xmx400m
      - DISABLE_INSTALL_DEMO_CONFIG=true
      - DISABLE_SECURITY_PLUGIN=true
      # Workaround on OSX for https://bugs.openjdk.org/browse/JDK-8345296
      - _JAVA_OPTIONS
    ulimits:
      memlock:
        soft: -1
        hard: -1
      nofile:
        soft: 65536
        hard: 65536
    ports:
      - "9200"
    healthcheck:
      test: curl -s http://localhost:9200/_cluster/health | grep -E '"status":"(green|yellow)"'
      start_period: 10s
      interval: 5s
      timeout: 10s
      retries: 10
    logging:
      driver: "json-file"
      options:
        max-size: "5m"
        max-file: "2"

  # Patch frontend-proxy to wait for observability UIs
  frontend-proxy:
    depends_on:
      jaeger:
        condition: service_started
      grafana:
        condition: service_started

  # Patch otel-collector to add observability exporters and wait for backends.
  # Note: otelcol-config-full.yml is included here to maintain correct config ordering
  # (core → full → observability → extras). When running minimal+observability without
  # compose.full.yaml, the Kafka/PostgreSQL receivers in the full config are harmless —
  # they will simply fail to scrape since those services are not running.
  otel-collector:
    command:
      - "--config=/etc/otelcol-config.yml"
      - "--config=/etc/otelcol-config-full.yml"
      - "--config=/etc/otelcol-config-observability.yml"
      - "--config=/etc/otelcol-config-extras.yml"
      - "--feature-gates=service.profilesSupport"
    volumes:
      - ${OTEL_COLLECTOR_CONFIG}:/etc/otelcol-config.yml
      - ${OTEL_COLLECTOR_CONFIG_FULL}:/etc/otelcol-config-full.yml
      - ${OTEL_COLLECTOR_CONFIG_OBSERVABILITY}:/etc/otelcol-config-observability.yml
      - ${OTEL_COLLECTOR_CONFIG_EXTRAS}:/etc/otelcol-config-extras.yml
    environment:
      - KAFKA_ADDR
    depends_on:
      jaeger:
        condition: service_started
      opensearch:
        condition: service_healthy