Genarrative/scripts/check-production-health-patrol.mjs

#!/usr/bin/env node

import { spawn, spawnSync } from 'node:child_process';
import {
  existsSync,
  mkdirSync,
  mkdtempSync,
  readFileSync,
  rmSync,
  writeFileSync,
} from 'node:fs';
import http from 'node:http';
import { tmpdir } from 'node:os';
import path from 'node:path';

const failures = [];
const tmpRoot = mkdtempSync(
  path.join(tmpdir(), 'genarrative-health-patrol-check-'),
);

try {
  await main();
} finally {
  rmSync(tmpRoot, { recursive: true, force: true });
}

if (failures.length > 0) {
  console.error('[check:production-health-patrol] FAILED');
  for (const failure of failures) {
    console.error(`- ${failure}`);
  }
  process.exit(1);
}

console.log('[check:production-health-patrol] OK');

async function main() {
  assertPublicBaseUrlDefaultsToGatewayEntry();
  await assertNginxModeChecksNginxService();
  await assertPingoraDirectModeChecksPingoraServiceAndPublicHost();
  assertRejectsInvalidGatewayMode();
  assertRejectsInvalidPublicHost();
  assertRejectsInvalidTimeoutAndSlowThreshold();
  assertRejectsInvalidBoolEnv();
}

function assertPublicBaseUrlDefaultsToGatewayEntry() {
  const script = readFileSync(
    'scripts/ops/production-health-patrol.mjs',
    'utf8',
  );
  assertIncludes(
    script,
    "process.env.GENARRATIVE_HEALTH_PATROL_PUBLIC_BASE_URL ||\n      'http://127.0.0.1'",
    'publicBaseUrl 默认必须指向本机网关入口，不能回落到 API 直连端口。',
  );
  if (
    script.includes(
      'process.env.GENARRATIVE_HEALTH_PATROL_PUBLIC_BASE_URL ||\n      process.env.GENARRATIVE_HEALTH_PATROL_API_BASE_URL',
    )
  ) {
    failures.push('publicBaseUrl 默认不应回落到 API base URL。');
  }
}

async function assertNginxModeChecksNginxService() {
  const fixture = await prepareFixture('nginx-mode');
  const result = await runPatrol(fixture, ['--gateway-mode', 'nginx']);

  assertStatus(result, 0, 'nginx gateway mode 巡检应成功。');
  if (result.status !== 0) {
    return;
  }

  const payload = JSON.parse(result.stdout);
  assertEqual(
    payload.gatewayMode,
    'nginx',
    '巡检 JSON 必须记录 gatewayMode=nginx。',
  );

  const commandsLog = readFileSync(fixture.commandsLog, 'utf8');
  assertIncludes(
    commandsLog,
    'systemctl is-active nginx.service',
    'nginx gateway mode 必须检查 nginx.service。',
  );
  if (commandsLog.includes('genarrative-pingora-gateway.service')) {
    failures.push(
      'nginx gateway mode 不应要求 Pingora gateway service active。',
    );
  }
}

async function assertPingoraDirectModeChecksPingoraServiceAndPublicHost() {
  const fixture = await prepareFixture('pingora-direct-mode');
  const result = await runPatrol(fixture, [
    '--gateway-mode',
    'pingora-direct',
    '--public-host',
    'genarrative.example',
  ]);

  assertStatus(result, 0, 'pingora-direct gateway mode 巡检应成功。');
  if (result.status !== 0) {
    return;
  }

  const payload = JSON.parse(result.stdout);
  assertEqual(
    payload.gatewayMode,
    'pingora-direct',
    '巡检 JSON 必须记录 gatewayMode=pingora-direct。',
  );

  const commandsLog = readTextFile(fixture.commandsLog, 'systemctl 命令日志');
  assertIncludes(
    commandsLog,
    'systemctl is-active genarrative-pingora-gateway.service',
    'pingora-direct gateway mode 必须检查 Pingora gateway service。',
  );
  if (commandsLog.includes('systemctl is-active nginx.service')) {
    failures.push(
      'pingora-direct gateway mode 不应要求 nginx.service active。',
    );
  }

  const requestsLog = readTextFile(
    fixture.requestsLog,
    'public probe 请求日志',
  );
  assertIncludes(
    requestsLog,
    'host=genarrative.example path=/',
    'public probe 必须带正式域名 Host header。',
  );
}

function assertRejectsInvalidGatewayMode() {
  const result = spawnSync(
    'node',
    [
      'scripts/ops/production-health-patrol.mjs',
      '--gateway-mode',
      'direct',
      '--skip-journal',
      '--json',
    ],
    {
      cwd: process.cwd(),
      encoding: 'utf8',
    },
  );

  if ((result.status ?? 0) === 0) {
    failures.push('非法 gateway mode 必须失败。');
  }
  assertIncludes(
    `${result.stdout}\n${result.stderr}`,
    '--gateway-mode 只支持 nginx 或 pingora-direct',
    '非法 gateway mode 必须给出明确错误。',
  );
}

function assertRejectsInvalidPublicHost() {
  const result = spawnSync(
    'node',
    [
      'scripts/ops/production-health-patrol.mjs',
      '--public-host',
      'https://genarrative.example',
      '--skip-journal',
      '--json',
    ],
    {
      cwd: process.cwd(),
      encoding: 'utf8',
    },
  );

  if ((result.status ?? 0) === 0) {
    failures.push('非法 public host 必须失败。');
  }
  assertIncludes(
    `${result.stdout}\n${result.stderr}`,
    '--public-host 只接受域名或 host:port',
    '非法 public host 必须给出明确错误。',
  );
}

function assertRejectsInvalidTimeoutAndSlowThreshold() {
  const cases = [
    {
      args: ['--timeout-ms', '0'],
      env: {},
      expected: '--timeout-ms 必须是正整数',
      reason: '非法 CLI timeout 必须失败。',
    },
    {
      args: ['--slow-ms', 'abc'],
      env: {},
      expected: '--slow-ms 必须是正整数',
      reason: '非法 CLI slow threshold 必须失败。',
    },
    {
      args: [],
      env: {
        GENARRATIVE_HEALTH_PATROL_TIMEOUT_MS: 'abc',
      },
      expected: 'GENARRATIVE_HEALTH_PATROL_TIMEOUT_MS 必须是正整数',
      reason: '非法 env timeout 必须失败。',
    },
    {
      args: [],
      env: {
        GENARRATIVE_HEALTH_PATROL_SLOW_MS: '0',
      },
      expected: 'GENARRATIVE_HEALTH_PATROL_SLOW_MS 必须是正整数',
      reason: '非法 env slow threshold 必须失败。',
    },
  ];

  for (const testCase of cases) {
    const result = spawnSync(
      'node',
      [
        'scripts/ops/production-health-patrol.mjs',
        '--skip-journal',
        '--json',
        ...testCase.args,
      ],
      {
        cwd: process.cwd(),
        encoding: 'utf8',
        env: {
          ...process.env,
          ...testCase.env,
        },
      },
    );

    if ((result.status ?? 0) === 0) {
      failures.push(testCase.reason);
    }
    assertIncludes(
      `${result.stdout}\n${result.stderr}`,
      testCase.expected,
      `${testCase.reason} 必须给出明确错误。`,
    );
  }
}

function assertRejectsInvalidBoolEnv() {
  const cases = [
    {
      env: {
        GENARRATIVE_HEALTH_PATROL_FAIL_ON_WARNING: 'ture',
      },
      expected: 'GENARRATIVE_HEALTH_PATROL_FAIL_ON_WARNING 必须是布尔值',
      reason: '生产健康巡检必须拒绝拼写错误的 fail-on-warning env。',
    },
    {
      env: {
        GENARRATIVE_HEALTH_PATROL_SKIP_JOURNAL: 'maybe',
      },
      expected: 'GENARRATIVE_HEALTH_PATROL_SKIP_JOURNAL 必须是布尔值',
      reason: '生产健康巡检必须拒绝非法 skip-journal env。',
    },
  ];

  for (const testCase of cases) {
    const result = spawnSync(
      'node',
      ['scripts/ops/production-health-patrol.mjs', '--json'],
      {
        cwd: process.cwd(),
        encoding: 'utf8',
        env: {
          ...process.env,
          ...testCase.env,
        },
      },
    );

    if ((result.status ?? 0) === 0) {
      failures.push(testCase.reason);
    }
    assertIncludes(
      `${result.stdout}\n${result.stderr}`,
      testCase.expected,
      `${testCase.reason} 必须给出明确错误。`,
    );
  }
}

async function prepareFixture(name) {
  const root = path.join(tmpRoot, name);
  const fakeBin = path.join(root, 'bin');
  const commandsLog = path.join(root, 'commands.log');
  const requestsLog = path.join(root, 'requests.log');

  mkdirSync(fakeBin, { recursive: true });

  writeFileSync(
    path.join(fakeBin, 'systemctl'),
    [
      '#!/usr/bin/env bash',
      `printf 'systemctl %s\\n' "$*" >> ${shellQuote(commandsLog)}`,
      'if [[ "$1" == "list-units" && "$2" == "genarrative-external-generation-worker@*.service" ]]; then',
      '  echo "genarrative-external-generation-worker@1.service loaded active running Genarrative external generation worker 1"',
      '  exit 0',
      'fi',
      'echo active',
      'exit 0',
      '',
    ].join('\n'),
    'utf8',
  );
  writeFileSync(
    path.join(fakeBin, 'journalctl'),
    [
      '#!/usr/bin/env bash',
      `printf 'journalctl %s\\n' "$*" >> ${shellQuote(commandsLog)}`,
      'echo "-- No entries --"',
      'exit 0',
      '',
    ].join('\n'),
    'utf8',
  );
  chmodExecutable(path.join(fakeBin, 'systemctl'));
  chmodExecutable(path.join(fakeBin, 'journalctl'));

  const server = http.createServer((request, response) => {
    writeFileSync(
      requestsLog,
      `host=${request.headers.host || ''} path=${request.url || ''}\n`,
      {
        encoding: 'utf8',
        flag: 'a',
      },
    );
    response.writeHead(200, {
      'Content-Type': 'application/json',
    });
    response.end(JSON.stringify({ ok: true }));
  });
  await listen(server, '127.0.0.1', 0);

  const address = server.address();
  if (!address || typeof address === 'string') {
    throw new Error('测试 HTTP server 未返回 TCP 端口。');
  }

  return {
    root,
    fakeBin,
    commandsLog,
    requestsLog,
    server,
    baseUrl: `http://127.0.0.1:${address.port}`,
  };
}

async function runPatrol(fixture, args) {
  try {
    return await spawnProcess(
      'node',
      [
        'scripts/ops/production-health-patrol.mjs',
        '--api-base-url',
        fixture.baseUrl,
        '--spacetime-base-url',
        fixture.baseUrl,
        '--public-base-url',
        fixture.baseUrl,
        '--skip-journal',
        '--json',
        ...args,
      ],
      {
        cwd: process.cwd(),
        env: {
          ...process.env,
          PATH: `${fixture.fakeBin}:${process.env.PATH || ''}`,
        },
      },
    );
  } finally {
    await closeServer(fixture.server);
  }
}

function spawnProcess(command, args, options) {
  return new Promise((resolve) => {
    const child = spawn(command, args, {
      ...options,
      stdio: ['ignore', 'pipe', 'pipe'],
    });
    let stdout = '';
    let stderr = '';
    child.stdout.setEncoding('utf8');
    child.stderr.setEncoding('utf8');
    child.stdout.on('data', (chunk) => {
      stdout += chunk;
    });
    child.stderr.on('data', (chunk) => {
      stderr += chunk;
    });
    child.on('error', (error) => {
      resolve({
        status: 1,
        stdout,
        stderr: `${stderr}${error.message}`,
      });
    });
    child.on('close', (code, signal) => {
      resolve({
        status: code ?? 1,
        signal,
        stdout,
        stderr,
      });
    });
  });
}

function listen(server, host, port) {
  return new Promise((resolve, reject) => {
    server.once('error', reject);
    server.listen(port, host, resolve);
  });
}

function closeServer(server) {
  return new Promise((resolve) => {
    server.close(() => resolve());
  });
}

function readTextFile(filePath, label) {
  if (!existsSync(filePath)) {
    failures.push(`${label} 未生成：${filePath}`);
    return '';
  }
  return readFileSync(filePath, 'utf8');
}

function chmodExecutable(filePath) {
  spawnSync('chmod', ['0755', filePath], {
    cwd: process.cwd(),
    encoding: 'utf8',
  });
}

function shellQuote(value) {
  return `'${String(value).replace(/'/g, "'\\''")}'`;
}

function assertStatus(result, expected, reason) {
  const actual = result.status ?? 0;
  if (actual !== expected) {
    failures.push(
      `${reason} 预期退出码 ${expected}，实际 ${actual}。\nstdout:\n${result.stdout}\nstderr:\n${result.stderr}`,
    );
  }
}

function assertIncludes(content, needle, reason) {
  if (!content.includes(needle)) {
    failures.push(`${reason} 缺少: ${needle}`);
  }
}

function assertEqual(actual, expected, reason) {
  if (actual !== expected) {
    failures.push(`${reason} 实际 ${actual}，预期 ${expected}。`);
  }
}