079project 1.0.0 → 2.0.0

package/GroupStarter.cjs

@@ -6,6 +6,24 @@ const express = require('express');
 const os = require('os');
 const { exec } = require('child_process');
 const axios = require('axios');
+// 新增：日志子进程
+let loggerProc = null;
+function startLogger() {
+  if (loggerProc && !loggerProc.killed) return;
+  const script = path.join(__dirname, 'loggerWorker.cjs');
+  if (!fs.existsSync(script)) {
+    console.warn('[LOGGER] 缺少 loggerWorker.cjs，跳过启动');
+    return;
+  }
+  loggerProc = spawn('node', [script], { cwd: __dirname, stdio: ['inherit', 'inherit', 'inherit', 'ipc'] });
+  loggerProc.on('exit', () => { loggerProc = null; setTimeout(startLogger, 2000); });
+}
+function loggerSend(msg) {
+  try { loggerProc && loggerProc.connected && loggerProc.send(msg); } catch (_) {}
+}
+// ...existing code...
+
+
 
 // 轻量参数解析（不引入依赖）
 function parseArgs(argv) {
@@ -34,12 +52,30 @@ let PUBLIC_DIR_OVERRIDE = args['public-dir'] ? path.resolve(args['public-dir'])
 let SNAPSHOTS_DIR_OVERRIDE = args['snapshots-dir'] ? path.resolve(args['snapshots-dir']) : '';
 const FORWARDER_OFFSET = SERVE_COUNT; // forwarder端口偏移
 
+let USE_PM2 = !!(args['use-pm2'] || process.env.USE_PM2);
 console.log('[CFG] GROUPS_DIR=', GROUPS_DIR);
 console.log('[CFG] SERVE_COUNT=', SERVE_COUNT, ' PORT_BASE=', PORT_BASE, ' MASTER_PORT=', MASTER_PORT);
 if (ROBOTS_DIR_OVERRIDE) console.log('[CFG] ROBOTS_DIR_OVERRIDE=', ROBOTS_DIR_OVERRIDE);
 if (TESTS_DIR_OVERRIDE) console.log('[CFG] TESTS_DIR_OVERRIDE=', TESTS_DIR_OVERRIDE);
 if (PUBLIC_DIR_OVERRIDE) console.log('[CFG] PUBLIC_DIR_OVERRIDE=', PUBLIC_DIR_OVERRIDE);
 if (SNAPSHOTS_DIR_OVERRIDE) console.log('[CFG] SNAPSHOTS_DIR_OVERRIDE=', SNAPSHOTS_DIR_OVERRIDE);
+function hasPm2() {
+  return new Promise((resolve) => {
+    exec('pm2 -v', { windowsHide: true }, (e) => resolve(!e));
+  });
+}
+async function ensurePm2(name, script, argsArr, cwd) {
+  const ok = await hasPm2();
+  if (!ok) return false;
+  return new Promise((resolve) => {
+    const cmd = `pm2 start "${process.execPath}" --name "${name}" -- ${['"' + script + '"', ...argsArr.map(a => `"${a}"`)].join(' ')}`;
+    exec(cmd, { cwd, windowsHide: true }, (e) => resolve(!e));
+  });
+}
+function killPid(pid) { try { process.kill(pid); } catch (_) {} }
+
+// 维护端口->组/角色映射，便于僵尸检测和重启
+const portIndex = new Map(); // port -> { groupId, role: 'serve'|'study'|'forwarder' }
 
 // 读取所有 group_x 文件夹
 const groupFolders = fs.existsSync(GROUPS_DIR)
@@ -198,62 +234,79 @@ function startResourceMonitor() {
   }, 1000);
 }
 
+
 async function startAllGroups() {
+  startLogger();
   for (let i = 0; i < groupFolders.length; i++) {
     const groupDir = groupFolders[i];
     const reg = { id: i, dir: groupDir, servePorts: [], study: null, forwarder: null };
-    const servePorts = [];
 
+    // 预计算本组端口
+    const groupServePorts = Array.from({ length: SERVE_COUNT }, (_, j) => PORT_BASE + i * (SERVE_COUNT + 2) + j);
+    const studyPort = PORT_BASE + i * (SERVE_COUNT + 2) + SERVE_COUNT;
+    const forwarderPort = PORT_BASE + i * (SERVE_COUNT + 2) + FORWARDER_OFFSET + 1;
+
+    const servePorts = [];
     for (let j = 0; j < SERVE_COUNT; j++) {
-      const port = PORT_BASE + i * (SERVE_COUNT + 2) + j;
+      const port = groupServePorts[j];
       servePorts.push(port);
       const serveScript = path.join(groupDir, 'main_Serve.cjs');
       if (fs.existsSync(serveScript)) {
         let attempts = 0;
+        const peers = groupServePorts.filter(p => p !== port).join(',');
         while (attempts < 3) {
           try {
-            const proc = spawn('node', [serveScript, port.toString(),'--expose-gc'], {
+            const proc = spawn('node', [
+              serveScript,
+              port.toString(),
+              '--expose-gc',
+              '--group-id', String(i),
+              '--forwarder-port', String(forwarderPort),
+              '--study-port', String(studyPort),
+              '--peers', peers
+            ], {
               cwd: groupDir,
               stdio: 'inherit',
             });
+            // 注册到日志子进程
+            portIndex.set(port, { groupId: i, role: 'serve' });
+            loggerSend({ type: 'register', processes: [{ groupId: i, role: 'serve', port, pid: null }] });
             groupProcesses.push(proc);
             reg.servePorts.push({ pid: null, port });
             proc.on('spawn', () => {
               const idx = reg.servePorts.findIndex((p) => p.port === port);
               if (idx >= 0) reg.servePorts[idx].pid = proc.pid;
+              loggerSend({ type: 'update', port, pid: proc.pid, role: 'serve', groupId: i });
             });
             break;
           } catch (error) {
             attempts++;
-            console.warn(
-              `[WARN] 启动服务 ${serveScript} 失败 (尝试 ${attempts}/3): ${error.message}`
-            );
-            if (attempts === 3)
-              console.error(`[ERROR] 无法启动服务 ${serveScript}`);
+            console.warn(`[WARN] 启动服务 ${serveScript} 失败 (尝试 ${attempts}/3): ${error.message}`);
+            if (attempts === 3) console.error(`[ERROR] 无法启动服务 ${serveScript}`);
             await new Promise((resolve) => setTimeout(resolve, 3000));
           }
         }
       }
     }
 
-    // study端口
-    const studyPort = PORT_BASE + i * (SERVE_COUNT + 2) + SERVE_COUNT;
-    const studyScript = path.join(groupDir, 'main_Study.cjs');
-    if (fs.existsSync(studyScript)) {
+    // study 进程（保持不变）
+    if (fs.existsSync(path.join(groupDir, 'main_Study.cjs'))) {
       const proc = spawn(
         'node',
-        [studyScript, studyPort.toString(), '--max-old-space-size=12288', '--expose-gc'],
+        [path.join(groupDir, 'main_Study.cjs'), studyPort.toString(), '--max-old-space-size=16384', '--expose-gc'],
         { cwd: groupDir, stdio: 'inherit' }
       );
+      portIndex.set(studyPort, { groupId: i, role: 'study' });
+      loggerSend({ type: 'register', processes: [{ groupId: i, role: 'study', port: studyPort, pid: null }] });
       groupProcesses.push(proc);
       reg.study = { pid: null, port: studyPort };
       proc.on('spawn', () => {
         reg.study.pid = proc.pid;
+        loggerSend({ type: 'update', port: studyPort, pid: proc.pid, role: 'study', groupId: i });
       });
     }
 
-    // forwarder端口
-    const forwarderPort = PORT_BASE + i * (SERVE_COUNT + 2) + FORWARDER_OFFSET + 1;
+    // forwarder 进程（保持不变）
     groupForwarderPorts.push(forwarderPort);
     const forwarderScript = path.join(groupDir, 'forwarder.js');
     if (fs.existsSync(forwarderScript)) {
@@ -268,22 +321,20 @@ async function startAllGroups() {
         ],
         { cwd: groupDir, stdio: 'inherit' }
       );
+      portIndex.set(forwarderPort, { groupId: i, role: 'forwarder' });
+      loggerSend({ type: 'register', processes: [{ groupId: i, role: 'forwarder', port: forwarderPort, pid: null }] });
       groupProcesses.push(proc);
       reg.forwarder = { pid: null, port: forwarderPort };
       proc.on('spawn', () => {
         reg.forwarder.pid = proc.pid;
+        loggerSend({ type: 'update', port: forwarderPort, pid: proc.pid, role: 'forwarder', groupId: i });
       });
     }
 
     // 写入组配置文件
     const config = { servePorts, studyPort, forwarderPort };
-    fs.writeFileSync(
-      path.join(groupDir, 'group_ports.json'),
-      JSON.stringify(config, null, 2)
-    );
-    console.log(
-      `[START] group_${i} 端口分配: serve=${servePorts.join(',')}, study=${studyPort}, forwarder=${forwarderPort}`
-    );
+    fs.writeFileSync(path.join(groupDir, 'group_ports.json'), JSON.stringify(config, null, 2));
+    console.log(`[START] group_${i} 端口分配: serve=${servePorts.join(',')}, study=${studyPort}, forwarder=${forwarderPort}`);
     groupRegistry.push(reg);
     await new Promise((r) => setTimeout(r, GROUP_START_DELAY));
   }
@@ -378,18 +429,26 @@ function startMasterForwarder() {
         return res.status(400).json({ error: 'invalid groupId' });
       }
       const reg = groupRegistry[groupId];
-      const spawnServe = () => {
-        const serveScript = path.join(reg.dir, 'main_Serve.cjs');
-        for (const sp of reg.servePorts) {
-          const p = spawn('node', [serveScript, sp.port.toString(), '--expose-gc','--max-old-space-size=16384'], {
-            cwd: reg.dir,
-            stdio: 'inherit',
-          });
-          p.on('spawn', () => {
-            sp.pid = p.pid;
-          });
-        }
-      };
+    const spawnServe = () => {
+      const serveScript = path.join(reg.dir, 'main_Serve.cjs');
+      const groupServePorts = reg.servePorts.map(s => s.port);
+      for (const sp of reg.servePorts) {
+        const peers = groupServePorts.filter(p => p !== sp.port).join(',');
+        const p = spawn('node', [
+          serveScript,
+          sp.port.toString(),
+          '--expose-gc',
+          '--group-id', String(groupId),
+          '--forwarder-port', String(reg.forwarder.port),
+          '--study-port', String(reg.study.port),
+          '--peers', peers
+        ], {
+          cwd: reg.dir,
+          stdio: 'inherit',
+        });
+        p.on('spawn', () => { sp.pid = p.pid; });
+      }
+    };
       const spawnStudy = () => {
         const studyScript = path.join(reg.dir, 'main_Study.cjs');
         const p = spawn(
@@ -474,7 +533,9 @@ function startMasterForwarder() {
     if (!message || typeof message !== 'string') {
       return res.status(400).json({ error: '消息不能为空' });
     }
-
+      const reqId = `${Date.now()}_${Math.floor(Math.random() * 1e6)}`;
+  // 记录输入
+  loggerSend({ type: 'io', ts: Date.now(), reqId, phase: 'in', input: message });
     // 轻微节流：高负载时延迟处理，避免CPU尖峰
     const curCpu = metrics.cpu[metrics.cpu.length - 1] || 0;
     const curMem = metrics.mem[metrics.mem.length - 1] || 0;
@@ -604,6 +665,17 @@ function startMasterForwarder() {
       mem: curMem,
       throttledMs: throttleMs,
     });
+      try {
+    loggerSend({
+      type: 'io',
+      ts: Date.now(),
+      reqId,
+      phase: 'out',
+      sampled: SELECTED,
+      top: unique[0]?.text || '',
+      responses: unique.slice(0, 10).map(u => ({ port: u.port, len: (u.text || '').length, score: u.score }))
+    });
+  } catch (_) {}
   });
 
   // 健康检查
@@ -628,13 +700,117 @@ function startMasterForwarder() {
     console.log(`[MASTER] 对外API: POST /api/chat（前端不会调用，可留作他用）`);
   });
 }
+// 僵尸端口检测 + 自动重启
+async function probePort(port) {
+  // 先 HTTP 探活
+  try {
+    const url = `http://localhost:${port}/health`;
+    const r = await axios.get(url, { timeout: 1500 });
+    if (r && r.status === 200) return true;
+  } catch (_) {}
+  // 回退：netstat 检查 LISTEN
+  return new Promise((resolve) => {
+    exec('netstat -ano -p TCP', { windowsHide: true }, (e, stdout) => {
+      if (e || !stdout) return resolve(false);
+      const re = new RegExp(`:${port}\\s+.*LISTENING\\s+(\\d+)`, 'i');
+      resolve(re.test(stdout));
+    });
+  });
+}
 
 // 启动所有组和总控 forwarder
 (async () => {
   await startAllGroups();
   startMasterForwarder();
 })();
+async function restartByPort(groupId, role, port) {
+  try {
+    const reg = groupRegistry[groupId];
+    if (!reg) return false;
+    const groupDir = reg.dir;
+
+    if (USE_PM2 && await hasPm2()) {
+      const name = `g${groupId}-${role}-${port}`;
+      const script = path.join(groupDir, role === 'study' ? 'main_Study.cjs' :
+        (role === 'forwarder' ? 'forwarder.js' : 'main_Serve.cjs'));
+      const args = role === 'forwarder'
+        ? [port.toString(), ...reg.servePorts.map(s => s.port.toString()), reg.study.port.toString(), '--expose-gc']
+        : [
+            port.toString(),
+            '--expose-gc',
+            '--group-id', String(groupId),
+            '--forwarder-port', String(reg.forwarder.port),
+            '--study-port', String(reg.study.port),
+            '--peers', reg.servePorts.map(s => s.port).filter(p => p !== port).join(',')
+          ];
+      await ensurePm2(name, script, args, groupDir);
+      console.log(`[PM2] 已重启 ${name}`);
+      return true;
+    }
 
+    if (role === 'serve') {
+      const found = reg.servePorts.find(s => s.port === port);
+      if (found?.pid) killPid(found.pid);
+      const serveScript = path.join(groupDir, 'main_Serve.cjs');
+      const peers = reg.servePorts.map(s => s.port).filter(p => p !== port).join(',');
+      const p = spawn('node', [
+        serveScript,
+        port.toString(),
+        '--expose-gc',
+        '--group-id', String(groupId),
+        '--forwarder-port', String(reg.forwarder.port),
+        '--study-port', String(reg.study.port),
+        '--peers', peers
+      ], { cwd: groupDir, stdio: 'inherit' });
+      p.on('spawn', () => {
+        if (found) found.pid = p.pid;
+        loggerSend({ type: 'update', port, pid: p.pid, role, groupId });
+      });
+      return true;
+    } else if (role === 'study') {
+      if (reg.study?.pid) killPid(reg.study.pid);
+      const studyScript = path.join(groupDir, 'main_Study.cjs');
+      const p = spawn('node', [studyScript, port.toString(), '--max-old-space-size=16384', '--expose-gc'], { cwd: groupDir, stdio: 'inherit' });
+      p.on('spawn', () => {
+        reg.study.pid = p.pid;
+        loggerSend({ type: 'update', port, pid: p.pid, role, groupId });
+      });
+      return true;
+    } else if (role === 'forwarder') {
+      if (reg.forwarder?.pid) killPid(reg.forwarder.pid);
+      const forwarderScript = path.join(groupDir, 'forwarder.js');
+      const args = [port.toString(), ...reg.servePorts.map(s => s.port.toString()), reg.study.port.toString(), '--expose-gc'];
+      const p = spawn('node', [forwarderScript, ...args], { cwd: groupDir, stdio: 'inherit' });
+      p.on('spawn', () => {
+        reg.forwarder.pid = p.pid;
+        loggerSend({ type: 'update', port, pid: p.pid, role, groupId });
+      });
+      return true;
+    }
+  } catch (e) {
+    console.warn('[RESTART] 重启失败:', e.message);
+  }
+  return false;
+}
+// 反触发/僵尸扫描：定期检测端口状态，异常则报警并重启
+setInterval(async () => {
+  const checks = [];
+  for (const [port, meta] of portIndex.entries()) {
+    checks.push({ port, meta, p: probePort(port) });
+  }
+  const results = await Promise.all(checks.map(c => c.p.catch(() => false)));
+  for (let i = 0; i < checks.length; i++) {
+    const ok = results[i];
+    const { port, meta } = checks[i];
+    const alive = ok ? 'OK' : 'ZOMBIE';
+    if (!ok) {
+      console.warn(`[ANTI-TRIGGER] 发现僵尸端口: ${port} (${meta.role} @ group ${meta.groupId})，自动重启`);
+      // 记录事件到 io 日志（子进程）
+      loggerSend({ type: 'io', ts: Date.now(), phase: 'event', event: 'zombie', port, role: meta.role, groupId: meta.groupId });
+      await restartByPort(meta.groupId, meta.role, port);
+    }
+  }
+}, 15000);
 // 进程退出处理
 process.on('SIGINT', () => {
   console.log('收到退出信号，正在关闭所有工作组进程...');

package/README.md

@@ -2,7 +2,7 @@
 
 ## 项目简介
 Phoenix AI 是一个基于多模态架构的人工智能系统，支持文本、图片、语音三种数据类型的处理。系统通过分布式工作组架构实现高效的任务分片和并行处理，适用于大规模数据集和复杂任务场景。
-
+目前，大多数模块仍然可能不那么稳定，如果有问题或者提交bug，请联系3873636760@qq.com
 ---
 
 ## 文件结构
@@ -62,8 +62,8 @@ Phoenix AI 是一个基于多模态架构的人工智能系统，支持文本、
 
 3. **访问服务**  
    - 文本组入口: http://localhost:9100
-   - 图片组入口: http://localhost:9200
-   - 语音组入口: http://localhost:9300
+   - 图片组入口: http://localhost:9200（未测试，谨慎使用）
+   - 语音组入口: http://localhost:9300（未测试，谨慎使用）
    - 多模态管理器: http://localhost:9050
 *** 注意，目前python的部分-前端序列化部分尚未完成，所以python可能没有用处 ***
 ---

package/loggerworker.cjs

@@ -0,0 +1,202 @@
+/* 日志子进程：接收主进程发来的 IO 记录与进程注册信息，按周期采样资源并输出文本表格 */
+const fs = require('fs');
+const path = require('path');
+const os = require('os');
+const { exec } = require('child_process');
+
+const LOG_DIR = path.join(__dirname, 'logs');
+fs.mkdirSync(LOG_DIR, { recursive: true });
+
+let watch = new Map(); // key: port -> { pid, groupId, role, port }
+let prevSample = new Map(); // pid -> { cpuSec, readBytes, writeBytes, ts }
+let lastMetricsWriteHour = '';
+let lastIoWriteHour = '';
+let sampleIntervalMs = 5000;
+let cores = os.cpus().length;
+
+// 工具：获取当前小时文件名
+function hourSuffix() {
+  const d = new Date();
+  const y = d.getFullYear();
+  const m = String(d.getMonth() + 1).padStart(2, '0');
+  const dd = String(d.getDate()).padStart(2, '0');
+  const hh = String(d.getHours()).padStart(2, '0');
+  return `${y}${m}${dd}_${hh}`;
+}
+function ioLogPath() { return path.join(LOG_DIR, `io_${hourSuffix()}.log`); }
+function metricsLogPath() { return path.join(LOG_DIR, `metrics_${hourSuffix()}.txt`); }
+
+function appendLine(file, line) {
+  fs.appendFile(file, line + '\n', () => {});
+}
+
+function writeIo(record) {
+  const nowHour = hourSuffix();
+  if (lastIoWriteHour !== nowHour) lastIoWriteHour = nowHour;
+  const file = ioLogPath();
+  appendLine(file, JSON.stringify(record));
+}
+function getProcInfoPowershell(pids) {
+  return new Promise((resolve) => {
+    if (!pids.length) return resolve([]);
+    const pidList = pids.join(',');
+    const cmd = [
+      'powershell.exe -NoProfile -Command',
+      `"Get-Process -Id ${pidList} | Select-Object Id,CPU,WorkingSet64,Name,`,
+      'IOReadBytes,IOWriteBytes | ConvertTo-Json -Compress"'
+    ].join(' ');
+    exec(cmd, { windowsHide: true, maxBuffer: 10 * 1024 * 1024 }, (err, stdout) => {
+      if (err || !stdout) return resolve([]);
+      try {
+        const data = JSON.parse(stdout);
+        resolve(Array.isArray(data) ? data : [data]);
+      } catch {
+        resolve([]);
+      }
+    });
+  });
+}
+
+function getTcpConnCounts() {
+  return new Promise((resolve) => {
+    // 使用 netstat -ano 统计 Owning PID 的 TCP 连接数量
+    exec('netstat -ano', { windowsHide: true, maxBuffer: 10 * 1024 * 1024 }, (err, stdout) => {
+      if (err || !stdout) return resolve(new Map());
+      const lines = stdout.split(/\r?\n/).filter(Boolean);
+      const map = new Map(); // pid -> count
+      for (const line of lines) {
+        // Proto  Local Address  Foreign Address  State  PID
+        const parts = line.trim().split(/\s+/);
+        if (parts.length < 5) continue;
+        const pid = Number(parts[parts.length - 1]);
+        if (!Number.isFinite(pid)) continue;
+        map.set(pid, (map.get(pid) || 0) + 1);
+      }
+      resolve(map);
+    });
+  });
+}
+
+function getListeningPortsByPid() {
+  return new Promise((resolve) => {
+    exec('netstat -ano -p TCP', { windowsHide: true, maxBuffer: 10 * 1024 * 1024 }, (err, stdout) => {
+      const map = new Map(); // pid -> Set(ports)
+      if (err || !stdout) return resolve(map);
+      const lines = stdout.split(/\r?\n/).filter(Boolean);
+      for (const line of lines) {
+        const parts = line.trim().split(/\s+/);
+        if (parts.length < 5) continue;
+        const state = parts[3] || parts[4] || '';
+        const pid = Number(parts[parts.length - 1]);
+        if (!Number.isFinite(pid)) continue;
+        const local = parts[1] || '';
+        const port = Number(local.split(':').pop());
+        if (!Number.isFinite(port)) continue;
+        if (!map.has(pid)) map.set(pid, new Set());
+        if (state.toUpperCase().includes('LISTEN')) {
+          map.get(pid).add(port);
+        }
+      }
+      resolve(map);
+    });
+  });
+}
+
+async function sampleAndWrite() {
+  try {
+    const ports = Array.from(watch.keys());
+    const procs = Array.from(watch.values());
+    const pids = Array.from(new Set(procs.map(p => p.pid))).filter(Boolean);
+
+    const [procInfo, connCounts, listenMap] = await Promise.all([
+      getProcInfoPowershell(pids),
+      getTcpConnCounts(),
+      getListeningPortsByPid()
+    ]);
+
+    const infoByPid = new Map();
+    for (const it of procInfo) {
+      infoByPid.set(Number(it.Id), {
+        cpuSec: Number(it.CPU || 0),
+        rss: Number(it.WorkingSet64 || 0),
+        ioRead: Number(it.IOReadBytes || 0),
+        ioWrite: Number(it.IOWriteBytes || 0),
+        name: it.Name || ''
+      });
+    }
+
+    const nowTs = Date.now();
+    const rows = [];
+    let aliveCount = 0;
+
+    for (const port of ports) {
+      const { pid, role, groupId } = watch.get(port);
+      const info = infoByPid.get(pid) || { cpuSec: 0, rss: 0, ioRead: 0, ioWrite: 0, name: '' };
+      const prev = prevSample.get(pid) || { cpuSec: info.cpuSec, readBytes: info.ioRead, writeBytes: info.ioWrite, ts: nowTs };
+      const dt = Math.max(1, (nowTs - prev.ts) / 1000); // sec
+      const dCpu = Math.max(0, info.cpuSec - prev.cpuSec); // sec
+      const cpuPct = Math.min(100, (dCpu / dt) * (100 / Math.max(1, 1 / cores))); // 近似
+      const dRead = Math.max(0, info.ioRead - prev.readBytes);
+      const dWrite = Math.max(0, info.ioWrite - prev.writeBytes);
+      const conn = connCounts.get(pid) || 0;
+      const rssMb = (info.rss / 1024 / 1024).toFixed(1);
+      const rps = (dRead / dt) | 0;
+      const wps = (dWrite / dt) | 0;
+      const listeningPorts = listenMap.get(pid) || new Set();
+      const up = listeningPorts.has(port) ? 'Y' : 'N';
+      if (up === 'Y') aliveCount++;
+
+      rows.push({
+        port,
+        pid,
+        role,
+        groupId,
+        cpu: cpuPct.toFixed(1),
+        mem: rssMb,
+        conn,
+        rps,
+        wps,
+        up
+      });
+
+      prevSample.set(pid, { cpuSec: info.cpuSec, readBytes: info.ioRead, writeBytes: info.ioWrite, ts: nowTs });
+    }
+
+    // 输出文本表
+    const nowHour = hourSuffix();
+    if (lastMetricsWriteHour !== nowHour) lastMetricsWriteHour = nowHour;
+    const file = metricsLogPath();
+    const header = `# ${new Date().toISOString()}  AlivePorts=${aliveCount}/${ports.length}\n` +
+      'PORT   PID      ROLE       GID  CPU%  MEM(MB)  CONN  DiskR/s  DiskW/s  UP\n' +
+      '-----  -------  ---------  ---  ----  -------  ----  -------  -------  --';
+    const lines = rows
+      .sort((a, b) => a.port - b.port)
+      .map(r =>
+        `${String(r.port).padEnd(5)}  ${String(r.pid).padEnd(7)}  ${String(r.role).padEnd(9)}  ${String(r.groupId).padEnd(3)}  ${String(r.cpu).padStart(4)}  ${String(r.mem).padStart(7)}  ${String(r.conn).padStart(4)}  ${String(r.rps).padStart(7)}  ${String(r.wps).padStart(7)}  ${r.up}`
+      );
+    appendLine(file, [header, ...lines, ''].join('\n'));
+  } catch (e) {
+    // 忽略单次采样错误
+  }
+}
+
+process.on('message', (msg) => {
+  if (!msg || typeof msg !== 'object') return;
+  if (msg.type === 'register' && Array.isArray(msg.processes)) {
+    for (const p of msg.processes) {
+      if (p && p.port) {
+        watch.set(p.port, { pid: p.pid, role: p.role, groupId: p.groupId, port: p.port });
+      }
+    }
+  } else if (msg.type === 'update' && msg.port) {
+    const old = watch.get(msg.port) || {};
+    watch.set(msg.port, { ...old, ...msg });
+  } else if (msg.type === 'io') {
+    // { ts, reqId, input, selectedPorts, responses[] }
+    writeIo(msg);
+  } else if (msg.type === 'config' && msg.sampleIntervalMs) {
+    sampleIntervalMs = Math.max(1000, Number(msg.sampleIntervalMs) || sampleIntervalMs);
+  }
+});
+
+setInterval(sampleAndWrite, sampleIntervalMs);