cangku/scripts/start-remote.mjs

import fs from 'node:fs';
import net from 'node:net';
import path from 'node:path';
import process from 'node:process';
import { spawn } from 'node:child_process';
import { fileURLToPath } from 'node:url';
import { loadConnectionConfig } from './remote-access-common.mjs';

const ROOT_DIR = path.resolve(path.dirname(fileURLToPath(import.meta.url)), '..');
const ACCESS_DIR = path.join(ROOT_DIR, '.remote-access');
const CONNECTION_PATH = path.join(ACCESS_DIR, 'connection.json');
const RESTART_MARKER_PATH = path.join(ACCESS_DIR, 'restart-requested.json');

/**
 * 解析后端实际监听端口。
 * 优先级：CANGKU_BACKEND_PORT 环境变量 > backend/.env 的 PORT > 默认 4001。
 * 前端 dev server 的 /api 代理必须指向同一端口，否则请求会被转发到空端口。
 */
function resolveBackendPort() {
  const fromEnv = Number(process.env.CANGKU_BACKEND_PORT);
  if (Number.isInteger(fromEnv) && fromEnv > 0 && fromEnv < 65536) {
    return fromEnv;
  }

  const envPath = path.join(ROOT_DIR, 'backend', '.env');
  if (fs.existsSync(envPath)) {
    const matched = fs.readFileSync(envPath, 'utf8').match(/^PORT=(\d+)\s*$/mu);
    const fromFile = Number(matched?.[1]);
    if (Number.isInteger(fromFile) && fromFile > 0 && fromFile < 65536) {
      return fromFile;
    }
  }

  return 4001;
}

/**
 * 隧道重连退避：1s → 2s → 4s … 封顶 30s。
 * 首次失败也立即重试（等价 autossh 的 AUTOSSH_GATETIME=0），避免网络尚未就绪时白等。
 */
const TUNNEL_RETRY_BASE_MS = 1000;
const TUNNEL_RETRY_MAX_MS = 30000;
const TUNNEL_RETRY_EXPONENT_CAP = 10;

function sleep(ms) {
  return new Promise((resolve) => setTimeout(resolve, ms));
}

export function computeTunnelRetryDelay(attempt) {
  const exponent = Math.min(attempt, TUNNEL_RETRY_EXPONENT_CAP);
  return Math.min(TUNNEL_RETRY_BASE_MS * 2 ** exponent, TUNNEL_RETRY_MAX_MS);
}

export function buildSshArgs(config) {
  return [
    '-N', '-T',
    '-o', 'BatchMode=yes',
    // 远端仍可能主动 RST（中间设备/服务端空闲超时），本地探活只能更快发现，不能阻止；
    // 真正的根治是下面的退避重连循环。
    '-o', 'ExitOnForwardFailure=yes',
    '-o', 'ServerAliveInterval=15',
    '-o', 'ServerAliveCountMax=3',
    '-o', 'TCPKeepAlive=yes',
    '-o', 'ConnectTimeout=10',
    '-o', 'StrictHostKeyChecking=yes',
    '-o', `UserKnownHostsFile=${path.join(ACCESS_DIR, 'known_hosts')}`,
    '-o', `CertificateFile=${path.join(ACCESS_DIR, 'id_ed25519-cert.pub')}`,
    '-i', path.join(ACCESS_DIR, 'id_ed25519'),
    '-p', String(config.ssh.port),
    '-L', `127.0.0.1:${config.localDatabasePort}:${config.ssh.remoteDatabaseHost}:${config.ssh.remoteDatabasePort}`,
    `${config.ssh.user}@${config.ssh.host}`,
  ];
}

function waitForPort(port, timeoutMs = 60000) {
  const started = Date.now();
  return new Promise((resolve, reject) => {
    const attempt = () => {
      const socket = net.createConnection({ host: '127.0.0.1', port });
      socket.once('connect', () => {
        socket.destroy();
        resolve();
      });
      socket.once('error', () => {
        socket.destroy();
        if (Date.now() - started >= timeoutMs) {
          reject(new Error(`SSH 隧道未能在 ${timeoutMs / 1000} 秒内监听端口 ${port}`));
          return;
        }
        setTimeout(attempt, 250);
      });
    };
    attempt();
  });
}

function waitForFile(filePath) {
  return new Promise((resolve) => {
    const timer = setInterval(() => {
      if (fs.existsSync(filePath)) {
        clearInterval(timer);
        resolve();
      }
    }, 500);
  });
}

function waitForExit(child, timeoutMs = 5000) {
  if (child.exitCode !== null || child.signalCode !== null) return Promise.resolve();
  return Promise.race([
    new Promise((resolve) => child.once('exit', resolve)),
    new Promise((resolve) => setTimeout(resolve, timeoutMs)),
  ]);
}

export async function main(mode = process.argv[2] || 'tunnel') {
  if (!['tunnel', 'dev', 'backend'].includes(mode)) {
    throw new Error(`未知启动模式：${mode}`);
  }

  const children = new Set();
  let shuttingDown = false;
  let restartingBackend = false;
  let backend = null;

  const stopChildren = () => {
    shuttingDown = true;
    for (const child of children) {
      if (!child.killed) child.kill('SIGTERM');
    }
  };
  for (const signal of ['SIGINT', 'SIGTERM']) {
    process.once(signal, stopChildren);
  }

  let tunnelRetryAttempt = 0;
  let resolveTunnelReady = null;
  const tunnelReady = new Promise((resolve) => {
    resolveTunnelReady = resolve;
  });

  /**
   * 启动一条 SSH 隧道并等它监听就绪。
   * 返回 { tunnel, exitPromise }；隧道结束由 exitPromise 上报退出码，交给
   * superviseTunnel 决定重连 —— 这里**不再**牵连后端/前端。
   */
  const startTunnel = async () => {
    const { value: config } = loadConnectionConfig(ROOT_DIR);
    const tunnel = spawn('ssh', buildSshArgs(config), { cwd: ROOT_DIR, stdio: 'inherit' });
    children.add(tunnel);

    const exitPromise = new Promise((resolve) => {
      tunnel.once('exit', (code) => {
        children.delete(tunnel);
        resolve(code);
      });
    });

    // 两个分支都 resolve，避免 Promise.race 落败方变成 unhandled rejection
    const outcome = await Promise.race([
      waitForPort(Number(config.localDatabasePort)).then(() => ({ ready: true })),
      exitPromise.then((code) => ({ ready: false, code })),
    ]);

    if (!outcome.ready) {
      if (!tunnel.killed) tunnel.kill('SIGTERM');
      throw new Error(`SSH 隧道在建立前退出（${outcome.code ?? 'unknown'}）`);
    }

    console.log(`系统数据库隧道已建立：127.0.0.1:${config.localDatabasePort}/${config.databaseName}`);
    return { tunnel, exitPromise };
  };

  /**
   * 隧道自愈循环（Windows 版 autossh 的语义）。
   *
   * 隧道断开只重连隧道本身，后端/前端进程保持不动。这是原先缺失的容错：
   * 旧实现在隧道 exit 时直接 stopChildren()，把整个开发栈一起杀掉，于是每一次
   * 远端 RST 都升级成一次「网站打不开」，还得靠外部守护清残留、整树重启。
   *
   * 建立失败走指数退避（1s→30s）；建立成功后重置退避，断开则固定 1s 后重连。
   */
  const superviseTunnel = async () => {
    for (;;) {
      if (shuttingDown) return;

      let session;
      try {
        session = await startTunnel();
        tunnelRetryAttempt = 0;
        if (resolveTunnelReady) {
          resolveTunnelReady();
          resolveTunnelReady = null;
        }
      } catch (error) {
        if (shuttingDown) return;
        const delay = computeTunnelRetryDelay(tunnelRetryAttempt);
        tunnelRetryAttempt += 1;
        console.error(`SSH 隧道建立失败（${error.message}），${delay}ms 后重试`);
        await sleep(delay);
        continue;
      }

      const code = await session.exitPromise;
      if (shuttingDown) return;
      console.error(`SSH 隧道已退出（${code ?? 'unknown'}），正在重连…（后端/前端不受影响）`);
      await sleep(TUNNEL_RETRY_BASE_MS);
    }
  };

  // 守护循环本身理论上不抛错，兜底仍保留：真异常时按老路径收尾，避免留下半死进程
  const startTunnelSupervisor = () => {
    superviseTunnel().catch((error) => {
      console.error(`隧道守护异常终止：${error.message}`);
      stopChildren();
      process.exitCode = 1;
    });
  };

  const npmCommand = process.platform === 'win32' ? 'npm.cmd' : 'npm';
  const startApplication = (args, kind) => {
    const childEnvironment = {
      ...process.env,
      CANGKU_REMOTE_SUPERVISOR: 'true',
    };
    if (kind === 'frontend' && !childEnvironment.VITE_API_PROXY_TARGET) {
      childEnvironment.VITE_API_PROXY_TARGET = `http://localhost:${resolveBackendPort()}`;
    }
    const child = spawn(npmCommand, args, {
      cwd: ROOT_DIR,
      stdio: 'inherit',
      env: childEnvironment,
      // Windows 下 .cmd 不是可执行文件，必须经 shell 启动，否则 spawn 直接抛 EINVAL
      shell: process.platform === 'win32',
    });
    children.add(child);
    child.once('exit', (code) => {
      children.delete(child);
      if (kind === 'backend' && restartingBackend) return;
      if (!shuttingDown) {
        console.error(`${kind === 'backend' ? '后端' : '前端'}进程已退出（${code ?? 'unknown'}）`);
        stopChildren();
        process.exitCode = code || 1;
      }
    });
    return child;
  };

  const hasConnection = fs.existsSync(CONNECTION_PATH);
  if (hasConnection) {
    if (fs.existsSync(RESTART_MARKER_PATH)) fs.unlinkSync(RESTART_MARKER_PATH);
    startTunnelSupervisor();
    await tunnelReady;
  } else if (mode === 'tunnel') {
    throw new Error('尚未配置远程系统数据库，请先运行应用并在首次启动弹窗中输入请求地址和 API Key');
  }

  if (mode === 'tunnel') return;

  backend = startApplication(
    ['--prefix', 'backend', 'run', mode === 'dev' ? 'dev' : 'start'],
    'backend'
  );
  if (mode === 'dev') {
    startApplication(['--prefix', 'frontend', 'run', 'dev'], 'frontend');
  }

  if (!hasConnection) {
    console.log('首次启动：请在安装页面输入配置请求地址和 API Key。');
    await waitForFile(RESTART_MARKER_PATH);
    restartingBackend = true;
    if (!backend.killed) backend.kill('SIGTERM');
    await waitForExit(backend);
    restartingBackend = false;
    fs.unlinkSync(RESTART_MARKER_PATH);
    startTunnelSupervisor();
    await tunnelReady;
    backend = startApplication(
      ['--prefix', 'backend', 'run', mode === 'dev' ? 'dev' : 'start'],
      'backend'
    );
  }
}

if (process.argv[1] === fileURLToPath(import.meta.url)) {
  main().catch((error) => {
    console.error(`远程运行启动失败：${error.message}`);
    process.exitCode = 1;
  });
}