senmesh-zc/deployment/ow/docker-compose.yml

# =============================================================================
# s-file 信创文件安全传输系统 —— 单台服务器一套（内网机 iw / 外网机 ow 通用同一份）
#
# 两台机器之间只靠光闸摆渡目录互通，彼此不开任何网络端口。
# 随角色变化的取值全在同目录 .env 里，本文件现场不用改。
#
#   ./scripts/install.sh          自检 -> 校验并载入 images/*.tar -> 建目录 -> 起栈 -> 等健康
#
# 数据一律落在宿主机，不建 docker 卷（默认在本目录 volumes/ 下，见 .env 的 SFILE_DATA_ROOT）：
#   volumes/pgdata   PostgreSQL 数据         volumes/redis  Redis AOF
#   volumes/mq       RabbitMQ 数据           volumes/upload 用户上传的文件本体
#   volumes/logs     后端日志(+logs/nginx)   volumes/gate/out|in 光闸发出/收到的传输包
#
# 前端是已构建好的静态包，直接放在 web/ 目录挂进 nginx：
#   改页面 = 替换 web/ 下的文件，docker compose restart web 即可，不用重新构建镜像。
#
# 只发布前端端口；后端 / PostgreSQL / Redis / RabbitMQ 仅在容器网络内可达。
# =============================================================================

name: sfile-${SFILE_ROLE:?缺少 SFILE_ROLE（内网机 iw / 外网机 ow）}

services:
  web:
    image: ${NGINX_IMAGE:?缺少 NGINX_IMAGE}
    container_name: sfile-${SFILE_ROLE}-web
    restart: unless-stopped
    ports:
      - "${WEB_HOST:-0.0.0.0}:${WEB_PORT:-80}:80"
    environment:
      TZ: ${TZ:-Asia/Shanghai}
    volumes:
      - ./nginx/web.conf:/etc/nginx/conf.d/default.conf:ro
      # 前端静态包：宿主机 web/ 目录即站点根，替换文件即换页面
      - ./web:/usr/share/nginx/html:ro
      - ${SFILE_DATA_ROOT:?缺少 SFILE_DATA_ROOT}/logs/nginx:/var/log/nginx
    depends_on:
      app:
        condition: service_healthy
    healthcheck:
      test: ["CMD-SHELL", "curl -sf -o /dev/null http://127.0.0.1:80/ || exit 1"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 20s
    networks: [sfile]

  app:
    image: ${APP_IMAGE:?缺少 APP_IMAGE}
    container_name: sfile-${SFILE_ROLE}-app
    restart: unless-stopped
    stop_grace_period: 30s
    environment:
      TZ: ${TZ:-Asia/Shanghai}
      # 发号器 workerId 按「容器 IP」取键，取不到会回退 99 而 99 > 31 直接启动失败，
      # 因此这里锁死 app 的容器 IP，并把同一个键写进 JVM 参数，两处永远一致
      # java.io.tmpdir 指到数据盘挂载内：Tomcat 接收大文件时先在这里落盘（multipart 上限 500MB），
      # 不指就写进容器 overlay 层，把系统盘吃掉几个 G。老系统那行 java.io.tmpdir 写在
      # application.properties 里其实没生效，这里用 -D 才真正生效
      JAVA_OPTS: "${APP_JAVA_OPTS:--Xms1g -Xmx3g -Dfile.encoding=UTF-8} -Djava.io.tmpdir=/var/s-file/upload/tmpdir -Dcode.snowflake.workid.${APP_NODE_IP:?缺少 APP_NODE_IP}=${APP_WORK_ID:?缺少 APP_WORK_ID}"
      SPRING_CONFIG_ADDITIONAL_LOCATION: file:/config/application.yml

      SPRING_DATASOURCE_URL: jdbc:postgresql://pg:5432/${POSTGRES_DB:?缺少 POSTGRES_DB}
      SPRING_DATASOURCE_USERNAME: ${POSTGRES_USER:?缺少 POSTGRES_USER}
      SPRING_DATASOURCE_PASSWORD: ${POSTGRES_PASSWORD:?缺少 POSTGRES_PASSWORD}

      SPRING_REDIS_HOST: redis
      SPRING_REDIS_PORT: 6379

      SPRING_RABBITMQ_HOST: mq
      SPRING_RABBITMQ_PORT: 5672
      SPRING_RABBITMQ_USERNAME: ${MQ_USER:?缺少 MQ_USER}
      SPRING_RABBITMQ_PASSWORD: ${MQ_PASSWORD:?缺少 MQ_PASSWORD}
      # 两侧交换机/广播 ID 必须不同，防止有人把两套栈并到一台机器时串消息
      RABBITMQ_CUSTOM_TOPIC_EXCHANGE_NAME: s-file-${SFILE_ROLE}-topic
      RABBITMQ_CUSTOM_DELAYED_EXCHANGE_NAME: s-file-${SFILE_ROLE}-delayed
      RABBITMQ_CUSTOM_BROADCAST_ID: s-file-${SFILE_ROLE}-broadcast

      SECURITY_JWT_SECRET: ${JWT_SECRET:?缺少 JWT_SECRET}
      USER_DEFAULT_PASSWORD: ${USER_DEFAULT_PASSWORD:?缺少 USER_DEFAULT_PASSWORD}

      # 内网机 1 / 外网机 2，写反了摆渡方向就反了
      FILE_TRANSFER_LOCATION: "${SFILE_TRANSFER_LOCATION:?缺少 SFILE_TRANSFER_LOCATION}"
      # 摆渡目录不进 upload 树（老系统是 ${file.upload-dir}/transfer/send）：光闸要单独挂这两个目录，
      # 嵌在 upload 里就得把整个文件库暴露给光闸。这里只是容器内路径，宿主机侧看 .env 的 GATE_*；
      # 传输包是过路文件、路径不入库，所以改它不影响历史数据
      FILE_TRANSFER_SEND_DIR: /data/transfer/send
      FILE_TRANSFER_RECEIVE_DIR: /data/transfer/receive
      FILE_TRANSFER_MONITOR_CRON: "${FILE_TRANSFER_MONITOR_CRON:-0/5 * * * * ?}"

      MESHSAFE_ENABLED: ${MESHSAFE_ENABLED:-true}
      MESHSAFE_BASE_URL: ${MESHSAFE_BASE_URL:-}
      MESHSAFE_CLIENT_ID: ${MESHSAFE_CLIENT_ID:-}
      MESHSAFE_SECRET: ${MESHSAFE_SECRET:-}
      MESHSAFE_CHANNEL: ${MESHSAFE_CHANNEL:-sfile-${SFILE_ROLE}}
      MESHSAFE_FAIL_OPEN: ${MESHSAFE_FAIL_OPEN:-true}
      MESHSAFE_TIMEOUT_SECONDS: ${MESHSAFE_TIMEOUT_SECONDS:-120}
      MESHSAFE_POLL_TIMEOUT_MINUTES: ${MESHSAFE_POLL_TIMEOUT_MINUTES:-60}
      MESHSAFE_POLL_CRON: ${MESHSAFE_POLL_CRON:-0/10 * * * * ?}
    # 容器标准输出留一份即可（应用日志已写 /usr/local/logs/s-file.log 并按天轮转），
    # 不限速的话 docker 的 json-file 默认不封顶，跑半年能把系统盘吃满
    logging:
      driver: json-file
      options:
        max-size: "20m"
        max-file: "5"
    volumes:
      - ./config/application.yml:/config/application.yml:ro
      # 容器内路径与老系统逐字一致（库里存的是绝对路径前缀，不能改），宿主机侧由 .env 决定
      - ${SFILE_DATA_ROOT}/upload:/var/s-file/upload
      - ${SFILE_DATA_ROOT}/logs:/usr/local/logs
      # 摆渡目录：容器内固定路径，宿主机侧由光闸设备负责单向搬运
      - ${GATE_SEND_DIR:?缺少 GATE_SEND_DIR}:/data/transfer/send
      - ${GATE_RECV_DIR:?缺少 GATE_RECV_DIR}:/data/transfer/receive
    # MeshSafe 与应用同机部署时靠它从容器内访问宿主机端口；
    # Docker < 20.10 不支持 host-gateway，或 MeshSafe 在独立服务器时，删掉这两行即可
    extra_hosts:
      - "host.docker.internal:host-gateway"
    depends_on:
      pg:
        condition: service_healthy
      redis:
        condition: service_healthy
      mq:
        condition: service_healthy
    healthcheck:
      test: ["CMD-SHELL", "curl -s --max-time 5 -o /dev/null http://127.0.0.1:8080/ || exit 1"]
      interval: 30s
      timeout: 8s
      retries: 5
      # 首次建库要跑全量 DDL，冷启动给足时间
      start_period: 180s
    networks:
      sfile:
        ipv4_address: ${APP_NODE_IP}

  pg:
    image: ${PG_IMAGE:?缺少 PG_IMAGE}
    container_name: sfile-${SFILE_ROLE}-pg
    restart: unless-stopped
    stop_grace_period: 60s
    environment:
      TZ: ${TZ:-Asia/Shanghai}
      POSTGRES_DB: ${POSTGRES_DB:?缺少 POSTGRES_DB}
      POSTGRES_USER: ${POSTGRES_USER:?缺少 POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:?缺少 POSTGRES_PASSWORD}
      # 空库首次启动才执行 initdb.d，已有数据卷时会被忽略
      POSTGRES_INITDB_ARGS: "--encoding=UTF8 --lc-collate=C --lc-ctype=C"
    volumes:
      # postgres 18+ 数据目录挂载点是 /var/lib/postgresql，15- 是 .../data，见 .env 的 PG_DATA_DIR
      - ${SFILE_DATA_ROOT}/pgdata:${PG_DATA_DIR:-/var/lib/postgresql}
      - ./sql/01-init-table.sql:/docker-entrypoint-initdb.d/01-init-table.sql:ro
      - ./sql/02-init-data.sql:/docker-entrypoint-initdb.d/02-init-data.sql:ro
    healthcheck:
      # 必须强制走 TCP(-h 127.0.0.1)：首次 initdb 期间镜像会临时起一个只听 unix socket
      # 的实例，此时用默认 socket 探测也会返回成功，导致 app 抢跑连库失败
      test: ["CMD-SHELL", "pg_isready -h 127.0.0.1 -U ${POSTGRES_USER} -d ${POSTGRES_DB}"]
      interval: 10s
      timeout: 5s
      retries: 10
      start_period: 60s
    networks: [sfile]

  redis:
    image: ${REDIS_IMAGE:?缺少 REDIS_IMAGE}
    container_name: sfile-${SFILE_ROLE}-redis
    restart: unless-stopped
    command: ["redis-server", "--appendonly", "yes", "--maxmemory", "${REDIS_MAXMEMORY:-512mb}", "--maxmemory-policy", "allkeys-lru"]
    environment:
      TZ: ${TZ:-Asia/Shanghai}
    volumes:
      - ${SFILE_DATA_ROOT}/redis:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 10s
      timeout: 3s
      retries: 5
    networks: [sfile]

  mq:
    image: ${MQ_IMAGE:?缺少 MQ_IMAGE}
    container_name: sfile-${SFILE_ROLE}-mq
    restart: unless-stopped
    stop_grace_period: 30s
    environment:
      TZ: ${TZ:-Asia/Shanghai}
      RABBITMQ_DEFAULT_USER: ${MQ_USER:?缺少 MQ_USER}
      RABBITMQ_DEFAULT_PASS: ${MQ_PASSWORD:?缺少 MQ_PASSWORD}
    volumes:
      - ${SFILE_DATA_ROOT}/mq:/var/lib/rabbitmq
    healthcheck:
      test: ["CMD-SHELL", "rabbitmq-diagnostics -q ping"]
      interval: 20s
      timeout: 10s
      retries: 10
      start_period: 60s
    networks: [sfile]

networks:
  sfile:
    driver: bridge
    ipam:
      config:
        # 锁 app 容器 IP 是为了发号器 workerId 取键；subnet/网关/app IP 三个值必须同网段。
        # 若与客户机器上已有 docker 网段冲突，整段换掉（如 172.31.88.0/24）即可
        - subnet: ${APP_SUBNET:?缺少 APP_SUBNET}
          gateway: ${APP_GATEWAY:?缺少 APP_GATEWAY}