HyperNode-AI/deploy/manage.sh

#!/usr/bin/env bash
# HyperNode-AI 部署管理器 — 一键部署 / 卸载 / 更新
#
# 新服务器:
#   curl -fsSL https://git.bz121.com/dekun/HyperNode-AI/raw/branch/main/deploy/manage.sh | bash
#
# 已安装:
#   bash /opt/hypernode/deploy/manage.sh
#
set -e
if [[ -n "${BASH_VERSION:-}" ]]; then
  set -o pipefail
fi

INSTALL_ROOT="${INSTALL_ROOT:-/opt/hypernode}"
DATA_DIR="${DATA_DIR:-/var/lib/hypernode}"
GIT_URL="${GIT_URL:-https://git.bz121.com/dekun/HyperNode-AI.git}"
GIT_BRANCH="${GIT_BRANCH:-main}"
APP_USER="hypernode"
SERVICE_NAME="hypernode"
BIND_HOST="0.0.0.0"
BIND_PORT="33333"
ENV_FILE="${INSTALL_ROOT}/.env"
GENERATED_ADMIN_PASSWORD=""

_script_src="${BASH_SOURCE[0]:-}"
if [[ -n "${_script_src}" && -f "${_script_src}" ]]; then
  REPO_ROOT="$(cd "$(dirname "${_script_src}")/.." && pwd)"
else
  REPO_ROOT=""
fi
unset _script_src

set -u

log() {
  printf '\n==> %s\n' "$*"
}

ask() {
  local value=""
  if [[ ! -r /dev/tty ]]; then
    echo "请在终端中运行。通过 curl 管道执行时，也需要当前会话有终端。" >&2
    exit 1
  fi
  if ! read -r -p "$1" value </dev/tty; then
    exit 0
  fi
  printf '%s' "${value}"
}

repo_ready() {
  [[ -f "${1}/app/main.py" && -f "${1}/requirements.txt" && -f "${1}/deploy/manage.sh" ]]
}

git_in_install() {
  git -c safe.directory="${INSTALL_ROOT}" -C "${INSTALL_ROOT}" "$@"
}

require_root() {
  if [[ "$(id -u)" -ne 0 ]]; then
    echo "请使用 root 执行。可以先 sudo -i，再运行安装命令。" >&2
    echo "或：curl -fsSL https://git.bz121.com/dekun/HyperNode-AI/raw/branch/main/deploy/manage.sh | sudo bash" >&2
    exit 1
  fi
}

ensure_git() {
  if command -v git >/dev/null 2>&1; then
    return 0
  fi
  if ! command -v apt-get >/dev/null 2>&1; then
    echo "未找到 git，且当前系统没有 apt-get。" >&2
    exit 1
  fi
  export DEBIAN_FRONTEND=noninteractive
  apt-get update -qq
  apt-get install -y git ca-certificates curl
}

bootstrap_repo() {
  if repo_ready "${INSTALL_ROOT}"; then
    REPO_ROOT="${INSTALL_ROOT}"
    return 0
  fi
  if [[ -n "${REPO_ROOT}" ]] && repo_ready "${REPO_ROOT}"; then
    return 0
  fi

  echo "HyperNode-AI 部署管理器 — 首次自举"
  echo "将克隆到: ${INSTALL_ROOT}"
  require_root
  ensure_git
  if [[ -e "${INSTALL_ROOT}" ]]; then
    echo "错误: ${INSTALL_ROOT} 已存在，但不是有效的 HyperNode-AI 目录。" >&2
    echo "请先处理该目录，或设置 INSTALL_ROOT 指向其他路径。" >&2
    exit 1
  fi
  git clone --branch "${GIT_BRANCH}" "${GIT_URL}" "${INSTALL_ROOT}"
  exec bash "${INSTALL_ROOT}/deploy/manage.sh" "$@" </dev/tty
}

install_packages() {
  export DEBIAN_FRONTEND=noninteractive
  apt-get update
  apt-get install -y python3 python3-venv python3-pip git ca-certificates curl rsync
  python3 -c 'import sys; raise SystemExit(0 if sys.version_info >= (3, 10) else 1)'
}

ensure_user() {
  if ! id -u "${APP_USER}" >/dev/null 2>&1; then
    useradd --system --home-dir "${INSTALL_ROOT}" --shell /usr/sbin/nologin "${APP_USER}"
  fi
}

ensure_install_tree() {
  if repo_ready "${INSTALL_ROOT}"; then
    return 0
  fi
  if [[ -n "${REPO_ROOT}" && "${REPO_ROOT}" != "${INSTALL_ROOT}" ]] && repo_ready "${REPO_ROOT}"; then
    mkdir -p "${INSTALL_ROOT}"
    rsync -a --delete \
      --exclude '.venv/' \
      --exclude '.env' \
      --exclude '__pycache__/' \
      --exclude '*.db' \
      "${REPO_ROOT}/" "${INSTALL_ROOT}/"
    return 0
  fi
  ensure_git
  git clone --branch "${GIT_BRANCH}" "${GIT_URL}" "${INSTALL_ROOT}"
}

write_env() {
  mkdir -p "${DATA_DIR}"
  if [[ -f "${ENV_FILE}" ]]; then
    log "保留已有配置 ${ENV_FILE}"
    return 0
  fi
  local jwt_secret admin_password
  jwt_secret="$(openssl rand -hex 32)"
  admin_password="$(openssl rand -hex 12)"
  GENERATED_ADMIN_PASSWORD="${admin_password}"
  log "写入 ${ENV_FILE}"
  cat >"${ENV_FILE}" <<EOF
DATABASE_URL=sqlite:////${DATA_DIR}/hypernode.db
YIELD_TIMEZONE=Asia/Shanghai
YIELD_HOUR=0
YIELD_MINUTE=5
DISABLE_SCHEDULER=false
JWT_SECRET=${jwt_secret}
ADMIN_USERNAME=admin
ADMIN_PASSWORD=${admin_password}
ADMIN_WALLET=admin-root
EOF
}

install_python_deps() {
  if [[ ! -x "${INSTALL_ROOT}/.venv/bin/python" ]]; then
    python3 -m venv "${INSTALL_ROOT}/.venv"
  fi
  "${INSTALL_ROOT}/.venv/bin/pip" install --upgrade pip
  "${INSTALL_ROOT}/.venv/bin/pip" install -r "${INSTALL_ROOT}/requirements.txt"
}

write_service() {
  cat >"/etc/systemd/system/${SERVICE_NAME}.service" <<EOF
[Unit]
Description=HyperNode-AI control plane
After=network-online.target
Wants=network-online.target

[Service]
Type=simple
User=${APP_USER}
Group=${APP_USER}
WorkingDirectory=${INSTALL_ROOT}
EnvironmentFile=${ENV_FILE}
Environment=PYTHONUNBUFFERED=1
Environment=PYTHONDONTWRITEBYTECODE=1
ExecStart=${INSTALL_ROOT}/.venv/bin/uvicorn app.main:app --host ${BIND_HOST} --port ${BIND_PORT} --workers 1
Restart=always
RestartSec=3
TimeoutStopSec=20
KillSignal=SIGINT
NoNewPrivileges=true
PrivateTmp=true

[Install]
WantedBy=multi-user.target
EOF
  systemctl daemon-reload
}

fix_permissions() {
  mkdir -p "${DATA_DIR}"
  chown -R "${APP_USER}:${APP_USER}" "${INSTALL_ROOT}" "${DATA_DIR}"
  chmod 750 "${INSTALL_ROOT}" "${DATA_DIR}"
  if [[ -f "${ENV_FILE}" ]]; then
    chmod 640 "${ENV_FILE}"
  fi
}

wait_until_ready() {
  local ready=0
  local _i
  echo "等待服务监听 ${BIND_PORT} ..."
  for ((_i = 1; _i <= 40; _i++)); do
    if curl -fsS "http://127.0.0.1:${BIND_PORT}/health" >/dev/null 2>&1; then
      ready=1
      break
    fi
    sleep 1
  done
  if [[ "${ready}" -ne 1 ]]; then
    echo "服务未在 40 秒内就绪。状态和最近日志：" >&2
    systemctl status "${SERVICE_NAME}.service" --no-pager >&2 || true
    journalctl -u "${SERVICE_NAME}.service" -n 80 --no-pager >&2 || true
    return 1
  fi
  echo "健康检查通过：http://127.0.0.1:${BIND_PORT}/health"
}

verify_app() {
  log "检查程序能否启动"
  if ! /usr/sbin/runuser -u "${APP_USER}" -- bash -c "cd '${INSTALL_ROOT}' && exec '${INSTALL_ROOT}/.venv/bin/python' -c 'import app.main'"; then
    echo "程序导入失败，没有重启服务。请把上面的报错发出来。" >&2
    return 1
  fi
}

cmd_deploy() {
  log "一键部署"
  install_packages
  ensure_user
  ensure_install_tree
  write_env
  install_python_deps
  fix_permissions
  write_service
  verify_app || return 1
  systemctl enable "${SERVICE_NAME}.service"
  systemctl restart "${SERVICE_NAME}.service"
  wait_until_ready || return 1
  log "部署完成"
  echo "监听地址：${BIND_HOST}:${BIND_PORT}"
  echo "健康检查：curl http://127.0.0.1:${BIND_PORT}/health"
  echo "控制台：http://<服务器 IP>:${BIND_PORT}/"
  echo "接口文档：http://<服务器 IP>:${BIND_PORT}/docs"
  if [[ -n "${GENERATED_ADMIN_PASSWORD}" ]]; then
    echo "管理员账号：admin"
    echo "管理员密码：${GENERATED_ADMIN_PASSWORD}"
    echo "密码写在 ${ENV_FILE}，请立刻记下。"
  fi
  echo "程序目录：${INSTALL_ROOT}"
  echo "数据库：${DATA_DIR}/hypernode.db"
  echo "日志：journalctl -u ${SERVICE_NAME}.service -f"
  echo "反代请自行配置，上游地址使用 127.0.0.1:${BIND_PORT}"
}

cmd_update() {
  if [[ ! -f "/etc/systemd/system/${SERVICE_NAME}.service" ]]; then
    echo "尚未部署。请先选择「一键部署」。" >&2
    return 1
  fi
  log "更新程序"
  install_packages
  if [[ -d "${INSTALL_ROOT}/.git" ]]; then
    if ! git_in_install diff --quiet -- web; then
      log "控制台文件与仓库不一致，按仓库版本覆盖 web/"
      git_in_install checkout -- web
    fi
    git_in_install pull --ff-only origin "${GIT_BRANCH}"
  else
    ensure_install_tree
  fi
  install_python_deps
  fix_permissions
  write_service
  verify_app || return 1
  systemctl restart "${SERVICE_NAME}.service"
  wait_until_ready || return 1
  log "更新完成，配置和数据库已保留"
}

cmd_uninstall() {
  local answer
  echo "即将停止服务并删除 ${INSTALL_ROOT} 与 systemd 单元。"
  answer="$(ask "确认卸载请输入 yes：")"
  if [[ "${answer}" != "yes" ]]; then
    echo "已取消卸载。"
    return 0
  fi

  log "卸载服务"
  systemctl stop "${SERVICE_NAME}.service" >/dev/null 2>&1 || true
  systemctl disable "${SERVICE_NAME}.service" >/dev/null 2>&1 || true
  rm -f "/etc/systemd/system/${SERVICE_NAME}.service"
  systemctl daemon-reload
  rm -rf "${INSTALL_ROOT}"

  answer="$(ask "同时删除数据库 ${DATA_DIR} ？输入 yes 删除，其他键保留：")"
  if [[ "${answer}" == "yes" ]]; then
    rm -rf "${DATA_DIR}"
    echo "数据库已删除。"
  else
    echo "数据库已保留在 ${DATA_DIR}"
  fi

  if id -u "${APP_USER}" >/dev/null 2>&1; then
    userdel "${APP_USER}" || echo "系统用户 ${APP_USER} 未删除，可稍后执行 userdel ${APP_USER}"
  fi
  log "卸载完成"
}

detect_ip() {
  hostname -I 2>/dev/null | awk '{print $1}'
}

show_banner() {
  local ip path
  ip="$(detect_ip)"
  path="${INSTALL_ROOT}"
  [[ -n "${ip}" ]] || ip="?"
  echo ""
  echo "========================================"
  echo " HyperNode-AI 部署管理器"
  echo " 仓库: git.bz121.com/dekun/HyperNode-AI"
  echo " 路径: ${path}"
  echo " 端口: ${BIND_PORT}"
  echo " IP:   ${ip}"
  echo "========================================"
}

show_menu() {
  echo " 1) 一键部署"
  echo " 2) 一键卸载"
  echo " 3) 更新"
  echo " 0) 退出"
  echo ""
}

main_menu() {
  local choice
  bootstrap_repo "$@"
  require_root
  if [[ -r /etc/os-release ]]; then
    # shellcheck disable=SC1091
    source /etc/os-release
    if [[ "${ID:-}" != "ubuntu" ]]; then
      echo "当前系统是 ${ID:-unknown}，这个脚本用于 Ubuntu 22.04。" >&2
      exit 1
    fi
    if [[ "${VERSION_ID:-}" != "22.04" ]]; then
      echo "警告：当前是 Ubuntu ${VERSION_ID:-unknown}，脚本按 22.04 编写，将继续。" >&2
    fi
  fi

  while true; do
    show_banner
    show_menu
    choice="$(ask "请选择 [0-3]: ")"
    case "${choice}" in
      1) cmd_deploy || echo "部署失败。" ;;
      2) cmd_uninstall || echo "卸载失败。" ;;
      3) cmd_update || echo "更新失败。" ;;
      0)
        echo "已退出。"
        exit 0
        ;;
      *) echo "无效选项：${choice}" ;;
    esac
    ask "按 Enter 返回菜单..." >/dev/null
  done
}

main_menu "$@"