clickstream-data-platform/compose.yaml

name: ${COMPOSE_PROJECT_NAME:-clickstream-data-platform}

x-clickhouse-nofile: &clickhouse-nofile
  nofile:
    soft: 262144
    hard: 262144

x-clickhouse-common: &clickhouse-common
  image: ${CLICKHOUSE_IMAGE:-clickhouse/clickhouse-server:26.3.17.56}
  restart: unless-stopped
  environment:
    # Стенд локальный: нодам нужен общий пользователь без пароля для Distributed.
    CLICKHOUSE_SKIP_USER_SETUP: "1"
  depends_on:
    clickhouse-keeper:
      condition: service_healthy
  ulimits: *clickhouse-nofile
  # Гигабайта не хватало: половину съедали страницы самого бинарника (ADR 0004).
  mem_limit: 4g
  healthcheck:
    test: ["CMD-SHELL", "clickhouse-client --host 127.0.0.1 --query 'SELECT 1' >/dev/null 2>&1"]
    interval: 5s
    timeout: 3s
    retries: 30
    start_period: 10s

x-airflow-common: &airflow-common
  image: clickstream-airflow:local
  build:
    context: ./infra/airflow
    args:
      AIRFLOW_BASE_IMAGE: ${AIRFLOW_IMAGE:-apache/airflow:3.3.0}
  environment:
    AIRFLOW__CORE__EXECUTOR: LocalExecutor
    AIRFLOW__CORE__PARALLELISM: "4"
    AIRFLOW__CORE__LOAD_EXAMPLES: "false"
    AIRFLOW__CORE__AUTH_MANAGER: airflow.api_fastapi.auth.managers.simple.simple_auth_manager.SimpleAuthManager
    AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_USERS: "${AIRFLOW_ADMIN_USER:-admin}:admin"
    AIRFLOW__CORE__SIMPLE_AUTH_MANAGER_PASSWORDS_FILE: /opt/airflow/auth/simple_auth_manager_passwords.json
    AIRFLOW__CORE__EXECUTION_API_SERVER_URL: http://airflow-apiserver:8080/execution/
    AIRFLOW__CORE__FERNET_KEY: ${AIRFLOW_FERNET_KEY:-AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=}
    AIRFLOW__API_AUTH__JWT_SECRET: ${AIRFLOW_JWT_SECRET:-local-only-airflow-jwt-secret-do-not-use}
    AIRFLOW__API_AUTH__JWT_ISSUER: airflow
    AIRFLOW__API__SECRET_KEY: ${AIRFLOW_API_SECRET_KEY:-local-only-airflow-api-secret-do-not-use}
    AIRFLOW__DATABASE__SQL_ALCHEMY_CONN: "postgresql+psycopg2://${AIRFLOW_METADATA_USER:-airflow}:${AIRFLOW_METADATA_PASSWORD:-airflow-local}@postgres-metadata:5432/airflow"
    AIRFLOW__SCHEDULER__ENABLE_HEALTH_CHECK: "true"
    AIRFLOW_ADMIN_USER: ${AIRFLOW_ADMIN_USER:-admin}
    AIRFLOW_ADMIN_PASSWORD: ${AIRFLOW_ADMIN_PASSWORD:-airflow}
  volumes:
    - ./dags:/opt/airflow/dags:ro
    - ./infra/airflow/init.sh:/opt/airflow/init.sh:ro
    - airflow_logs:/opt/airflow/logs
    - airflow_auth:/opt/airflow/auth

x-generator-common: &generator-common
  image: clickstream-generator:local
  build:
    context: .
    dockerfile: generator/Dockerfile
  restart: "no"
  depends_on:
    kafka-init:
      condition: service_completed_successfully
    # Читать топик движок Kafka начинает не при своём создании, а когда над ним
    # появляется матвью приёма, и она создаётся последней по номеру файла
    # (sql/ddl/40-stg-views.sql). Отправь генератор события раньше — они лягут в
    # топик и молча минуют хранилище.
    clickhouse-init:
      condition: service_completed_successfully
  environment:
    # Адрес брокера и имя топика — факты стенда, и называет их стенд.
    # Остальное (день, зерно, число дней, предел пачки) приходит аргументами
    # от того, кто запускает: у службы нет позиции на оси мира.
    KAFKA_BOOTSTRAP_SERVERS: ${KAFKA_BOOTSTRAP_SERVERS:-kafka:9092}
    KAFKA_TOPIC: ${KAFKA_TOPIC:-hits}

x-superset-common: &superset-common
  image: clickstream-superset:local
  build:
    context: ./infra/superset
    args:
      SUPERSET_BASE_IMAGE: ${SUPERSET_IMAGE:-apache/superset:6.1.0}
  environment:
    SUPERSET_CONFIG_PATH: /app/pythonpath/superset_config.py
    SUPERSET_SECRET_KEY: ${SUPERSET_SECRET_KEY:-local-only-superset-secret-do-not-use-outside-this-stand}
    SUPERSET__SQLALCHEMY_DATABASE_URI: "postgresql+psycopg2://${SUPERSET_METADATA_USER:-superset}:${SUPERSET_METADATA_PASSWORD:-superset-local}@postgres-metadata:5432/superset"
    SUPERSET_ADMIN_USER: ${SUPERSET_ADMIN_USER:-admin}
    SUPERSET_ADMIN_PASSWORD: ${SUPERSET_ADMIN_PASSWORD:-superset}
  volumes:
    - ./infra/superset/superset_config.py:/app/pythonpath/superset_config.py:ro
    - ./infra/superset/init.sh:/app/infra/init.sh:ro
    - ./infra/superset/import:/app/import:ro
    - superset_home:/app/superset_home

services:
  clickhouse-keeper:
    image: ${CLICKHOUSE_IMAGE:-clickhouse/clickhouse-server:26.3.17.56}
    restart: unless-stopped
    user: clickhouse
    ulimits: *clickhouse-nofile
    entrypoint:
      - clickhouse
      - keeper
      - --config-file=/etc/clickhouse-keeper/keeper_config.xml
    volumes:
      - ./infra/clickhouse/keeper/keeper_config.xml:/etc/clickhouse-keeper/keeper_config.xml:ro
      - clickhouse_keeper_data:/var/lib/clickhouse
    mem_limit: 512m
    healthcheck:
      test: ["CMD-SHELL", "clickhouse keeper-client --host 127.0.0.1 --port 9181 --query ruok | grep -q imok"]
      interval: 5s
      timeout: 3s
      retries: 30
      start_period: 10s

  # Инициатор DDL и будущее подключение Airflow.
  clickhouse-01:
    <<: *clickhouse-common
    hostname: clickhouse-01
    ports:
      - "127.0.0.1:${CLICKHOUSE_01_HTTP_PORT:-28123}:8123"
      - "127.0.0.1:${CLICKHOUSE_01_TCP_PORT:-29000}:9000"
    volumes:
      - ./infra/clickhouse/config.d/cluster.xml:/etc/clickhouse-server/config.d/cluster.xml:ro
      - ./infra/clickhouse/config.d/prometheus.xml:/etc/clickhouse-server/config.d/prometheus.xml:ro
      - ./infra/clickhouse/config.d/macros-01.xml:/etc/clickhouse-server/config.d/macros.xml:ro
      - clickhouse_01_data:/var/lib/clickhouse

  # Будущее подключение Superset.
  clickhouse-02:
    <<: *clickhouse-common
    hostname: clickhouse-02
    ports:
      - "127.0.0.1:${CLICKHOUSE_02_HTTP_PORT:-28124}:8123"
      - "127.0.0.1:${CLICKHOUSE_02_TCP_PORT:-29001}:9000"
    volumes:
      - ./infra/clickhouse/config.d/cluster.xml:/etc/clickhouse-server/config.d/cluster.xml:ro
      - ./infra/clickhouse/config.d/prometheus.xml:/etc/clickhouse-server/config.d/prometheus.xml:ro
      - ./infra/clickhouse/config.d/macros-02.xml:/etc/clickhouse-server/config.d/macros.xml:ro
      - clickhouse_02_data:/var/lib/clickhouse

  kafka:
    image: ${KAFKA_IMAGE:-apache/kafka:4.3.1}
    hostname: kafka
    restart: unless-stopped
    ports:
      - "127.0.0.1:${KAFKA_EXTERNAL_PORT:-29092}:29092"
    environment:
      KAFKA_NODE_ID: "1"
      KAFKA_PROCESS_ROLES: "broker,controller"
      KAFKA_CONTROLLER_QUORUM_VOTERS: "1@kafka:9093"
      KAFKA_LISTENERS: "INTERNAL://:9092,CONTROLLER://:9093,EXTERNAL://:29092"
      KAFKA_ADVERTISED_LISTENERS: "INTERNAL://kafka:9092,EXTERNAL://localhost:${KAFKA_EXTERNAL_PORT:-29092}"
      KAFKA_LISTENER_SECURITY_PROTOCOL_MAP: "INTERNAL:PLAINTEXT,CONTROLLER:PLAINTEXT,EXTERNAL:PLAINTEXT"
      KAFKA_INTER_BROKER_LISTENER_NAME: "INTERNAL"
      KAFKA_CONTROLLER_LISTENER_NAMES: "CONTROLLER"
      KAFKA_OFFSETS_TOPIC_REPLICATION_FACTOR: "1"
      KAFKA_TRANSACTION_STATE_LOG_REPLICATION_FACTOR: "1"
      KAFKA_TRANSACTION_STATE_LOG_MIN_ISR: "1"
      KAFKA_SHARE_COORDINATOR_STATE_TOPIC_REPLICATION_FACTOR: "1"
      KAFKA_SHARE_COORDINATOR_STATE_TOPIC_MIN_ISR: "1"
      KAFKA_GROUP_INITIAL_REBALANCE_DELAY_MS: "0"
      KAFKA_NUM_PARTITIONS: "1"
      KAFKA_LOG_DIRS: "/var/lib/kafka/data"
      KAFKA_HEAP_OPTS: "-Xmx512m -Xms256m"
      CLUSTER_ID: "4L6g3nShT-eMCtK--X86sw"
    volumes:
      - kafka_data:/var/lib/kafka/data
    mem_limit: 1g
    healthcheck:
      test: ["CMD-SHELL", "/opt/kafka/bin/kafka-topics.sh --bootstrap-server localhost:9092 --list >/dev/null 2>&1"]
      interval: 15s
      timeout: 10s
      retries: 20
      start_period: 30s

  kafka-init:
    image: ${KAFKA_IMAGE:-apache/kafka:4.3.1}
    restart: "no"
    depends_on:
      kafka:
        condition: service_healthy
    entrypoint: ["/bin/bash", "-ec"]
    command:
      - |
        topic_partition_count() {
          /opt/kafka/bin/kafka-topics.sh \
            --bootstrap-server kafka:9092 \
            --describe \
            --topic hits | \
            sed -n '1s/.*PartitionCount: \([0-9][0-9]*\).*/\1/p'
        }

        /opt/kafka/bin/kafka-topics.sh \
          --bootstrap-server kafka:9092 \
          --create \
          --if-not-exists \
          --topic hits \
          --partitions 2 \
          --replication-factor 1

        partition_count="$$(topic_partition_count)"
        # Брокер разрешает автосоздание, хотя потребитель ClickHouse его не запрашивает.
        # На сохранённом томе другой клиент мог уже создать hits с одним разделом.
        if [[ "$$partition_count" == "1" ]]; then
          /opt/kafka/bin/kafka-topics.sh \
            --bootstrap-server kafka:9092 \
            --alter \
            --topic hits \
            --partitions 2
          partition_count="$$(topic_partition_count)"
        fi
        if [[ "$$partition_count" != "2" ]]; then
          printf 'ОШИБКА: у топика hits разделов: %s, ожидалось: 2.\n' \
            "$${partition_count:-неизвестно}" >&2
          exit 1
        fi

  # DDL применяется с ноды 1 по порядку имён файлов после готовности всего кластера.
  clickhouse-init:
    image: ${CLICKHOUSE_IMAGE:-clickhouse/clickhouse-server:26.3.17.56}
    restart: "no"
    depends_on:
      kafka-init:
        condition: service_completed_successfully
      clickhouse-01:
        condition: service_healthy
      clickhouse-02:
        condition: service_healthy
    environment:
      LC_ALL: C
    entrypoint: ["/bin/bash", "-ec"]
    command:
      - |
        set -- /ddl/*.sql
        if [[ ! -e "$$1" ]]; then
          printf 'ОШИБКА: в /ddl нет файлов SQL.\n' >&2
          exit 1
        fi
        for ddl_file do
          printf 'Применяем %s.\n' "$${ddl_file##*/}"
          clickhouse-client \
            --host clickhouse-01 \
            --multiquery \
            --queries-file "$$ddl_file"
        done
    volumes:
      - ./sql/ddl:/ddl:ro

  # Стартовый мир: восемь модельных дней уезжают в топик при каждом подъёме.
  # Что это за дни и каким мир обязан выйти — data/world-inventory.json.
  #
  # Число дней стоит здесь числом: YAML не читает Python, и одно из двух мест
  # (второе — `STARTING_DAYS` в inventory.py) лишнее по построению. Разъедутся
  # они — покраснеют счётчики `make check-clickhouse`.
  #
  # Повторный `make up` заливает мир заново, и это не оплошность: `WatchID` у
  # событий те же, ReplacingMergeTree схлопнет повтор в ODS. Сырьё в STG при
  # этом честно удвоится — свойство слоя, описанное в storage.md.
  world-init:
    <<: *generator-common
    command: ["batch", "--day", "0", "--days", "8"]

  # Тот же образ для ручных прогонов: `make generate-batch`, `make generate-live`.
  # Под профилем — чтобы обычный подъём стенда её не трогал.
  generator:
    <<: *generator-common
    profiles: ["generator"]
    command: ["batch"]

  postgres-metadata:
    image: ${POSTGRES_IMAGE:-postgres:16-alpine}
    restart: unless-stopped
    command:
      - postgres
      - -c
      - shared_buffers=64MB
      - -c
      - max_connections=50
    environment:
      POSTGRES_USER: ${POSTGRES_ADMIN_USER:-postgres}
      POSTGRES_PASSWORD: ${POSTGRES_ADMIN_PASSWORD:-postgres-local}
      POSTGRES_DB: postgres
      AIRFLOW_METADATA_USER: ${AIRFLOW_METADATA_USER:-airflow}
      AIRFLOW_METADATA_PASSWORD: ${AIRFLOW_METADATA_PASSWORD:-airflow-local}
      SUPERSET_METADATA_USER: ${SUPERSET_METADATA_USER:-superset}
      SUPERSET_METADATA_PASSWORD: ${SUPERSET_METADATA_PASSWORD:-superset-local}
    volumes:
      - ./infra/postgres/init/10-metadata-databases.sh:/docker-entrypoint-initdb.d/10-metadata-databases.sh:ro
      - postgres_metadata_data:/var/lib/postgresql/data
    mem_limit: 256m
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U \"$${POSTGRES_USER}\" -d postgres >/dev/null"]
      interval: 5s
      timeout: 5s
      retries: 20
      start_period: 10s

  airflow-init:
    <<: *airflow-common
    user: "0:0"
    restart: "no"
    depends_on:
      postgres-metadata:
        condition: service_healthy
      clickhouse-01:
        condition: service_healthy
      # Не убирать: без зависимого успешный clickhouse-init считается
      # упавшим для --wait.
      clickhouse-init:
        condition: service_completed_successfully
      # По той же причине — и ещё по одной. Заливка стартового мира идёт
      # последней в цепи разовых служб, и зависимого ей взять больше негде.
      # Заодно это правда про порядок: даг `next_day` этапа 5 продолжает ось с
      # того дня, на котором заливка остановилась, — Airflow приходит в мир,
      # который уже есть.
      world-init:
        condition: service_completed_successfully
    entrypoint: ["/bin/bash"]
    command: ["/opt/airflow/init.sh"]

  airflow-apiserver:
    <<: *airflow-common
    restart: unless-stopped
    depends_on:
      airflow-init:
        condition: service_completed_successfully
    command: api-server
    ports:
      - "127.0.0.1:${AIRFLOW_PORT:-28080}:8080"
    mem_limit: 512m
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://127.0.0.1:8080/api/v2/monitor/health | python -c 'import json,sys; sys.exit(0 if json.load(sys.stdin)[\"metadatabase\"][\"status\"] == \"healthy\" else 1)'"]
      interval: 10s
      timeout: 5s
      retries: 30
      start_period: 30s

  airflow-scheduler:
    <<: *airflow-common
    restart: unless-stopped
    depends_on:
      airflow-init:
        condition: service_completed_successfully
    command: scheduler
    mem_limit: 640m
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://127.0.0.1:8974/health >/dev/null"]
      interval: 10s
      timeout: 5s
      retries: 30
      start_period: 30s

  airflow-dag-processor:
    <<: *airflow-common
    restart: unless-stopped
    depends_on:
      airflow-init:
        condition: service_completed_successfully
      airflow-apiserver:
        condition: service_healthy
    command: dag-processor
    mem_limit: 384m
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://airflow-apiserver:8080/api/v2/monitor/health | python -c 'import json,sys; sys.exit(0 if json.load(sys.stdin)[\"dag_processor\"][\"status\"] == \"healthy\" else 1)'"]
      interval: 10s
      timeout: 5s
      retries: 30
      start_period: 30s

  superset-init:
    <<: *superset-common
    restart: "no"
    healthcheck:
      disable: true
    depends_on:
      postgres-metadata:
        condition: service_healthy
      clickhouse-02:
        condition: service_healthy
    entrypoint: ["/bin/bash"]
    command: ["/app/infra/init.sh"]

  superset:
    <<: *superset-common
    restart: unless-stopped
    depends_on:
      superset-init:
        condition: service_completed_successfully
    ports:
      - "127.0.0.1:${SUPERSET_PORT:-28088}:8088"
    mem_limit: 1g
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://127.0.0.1:8088/health >/dev/null"]
      interval: 10s
      timeout: 5s
      retries: 30
      start_period: 30s

  prometheus:
    image: ${PROMETHEUS_IMAGE:-prom/prometheus:v3.13.2}
    restart: unless-stopped
    depends_on:
      clickhouse-keeper:
        condition: service_healthy
      clickhouse-01:
        condition: service_healthy
      clickhouse-02:
        condition: service_healthy
    command:
      - --config.file=/etc/prometheus/prometheus.yml
      - --storage.tsdb.path=/prometheus
      - --storage.tsdb.retention.time=7d
      - --storage.tsdb.retention.size=512MB
    volumes:
      - ./infra/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - prometheus_data:/prometheus
    ports:
      - "127.0.0.1:${PROMETHEUS_PORT:-29090}:9090"
    mem_limit: 512m
    healthcheck:
      test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:9090/-/ready >/dev/null 2>&1"]
      interval: 10s
      timeout: 5s
      retries: 15
      start_period: 10s

  grafana:
    image: ${GRAFANA_IMAGE:-grafana/grafana:13.1.1}
    restart: unless-stopped
    depends_on:
      prometheus:
        condition: service_healthy
    environment:
      GF_SECURITY_ADMIN_USER: ${GRAFANA_ADMIN_USER:-admin}
      GF_SECURITY_ADMIN_PASSWORD: ${GRAFANA_ADMIN_PASSWORD:-admin}
      GF_USERS_ALLOW_SIGN_UP: "false"
      GF_NEWS_NEWS_FEED_ENABLED: "false"
      GF_ANALYTICS_REPORTING_ENABLED: "false"
      GF_ANALYTICS_CHECK_FOR_UPDATES: "false"
    volumes:
      - ./infra/grafana/provisioning/datasources:/etc/grafana/provisioning/datasources:ro
      - grafana_data:/var/lib/grafana
    ports:
      - "127.0.0.1:${GRAFANA_PORT:-23000}:3000"
    mem_limit: 512m
    healthcheck:
      test: ["CMD-SHELL", "curl -sf http://127.0.0.1:3000/api/health >/dev/null"]
      interval: 10s
      timeout: 5s
      retries: 15
      start_period: 15s

volumes:
  clickhouse_keeper_data:
  clickhouse_01_data:
  clickhouse_02_data:
  kafka_data:
  postgres_metadata_data:
  airflow_logs:
  airflow_auth:
  superset_home:
  prometheus_data:
  grafana_data: