audio_codebuddy/deploy-package/scripts/inspect-oss-usage.cjs

/**
 * 只读探查 OSS bucket 占用分布。
 * 用法: node deploy-package/scripts/inspect-oss-usage.cjs [bucket]
 *   bucket 可选，默认依次尝试 server/.env 里的两个候选
 *
 * 不删除任何文件。纯统计 + 报告。
 */
// 优先用环境变量；否则从 server/.env.production 文本读
let AK = process.env.OSS_ACCESS_KEY_ID;
let SK = process.env.OSS_ACCESS_KEY_SECRET;
let REGION = process.env.OSS_REGION || 'oss-cn-hangzhou';
let EP = process.env.OSS_INTERNAL_ENDPOINT || process.env.OSS_ENDPOINT;
let envBucket = process.env.OSS_BUCKET_NAME;

if (!AK || !SK) {
  const fs = require('fs');
  const envPath = require('path').join(__dirname, '..', '..', 'server', '.env.production');
  if (fs.existsSync(envPath)) {
    const txt = fs.readFileSync(envPath, 'utf8');
    for (const line of txt.split(/\r?\n/)) {
      const m = line.match(/^\s*([A-Z0-9_]+)\s*=\s*(.*)\s*$/);
      if (!m) continue;
      const k = m[1], v = m[2].replace(/^["']|["']$/g, '');
      if (k === 'OSS_ACCESS_KEY_ID' && !AK) AK = v;
      if (k === 'OSS_ACCESS_KEY_SECRET' && !SK) SK = v;
      if (k === 'OSS_REGION' && !process.env.OSS_REGION) REGION = v;
      if (k === 'OSS_INTERNAL_ENDPOINT' && !EP) EP = v;
      if (k === 'OSS_ENDPOINT' && !EP) EP = v;
      if (k === 'OSS_BUCKET_NAME' && !envBucket) envBucket = v;
    }
  }
}

// 脚本在 deploy-package/scripts 下，ali-oss 在 server/node_modules，从脚本位置往上找
const path = require('path');
const Module = require('module');
const serverNodeModules = path.join(__dirname, '..', '..', 'server', 'node_modules');
if (!Module.globalPaths.includes(serverNodeModules)) Module.globalPaths.push(serverNodeModules);
const OSS = require(path.join(serverNodeModules, 'ali-oss'));
EP = EP || 'oss-cn-hangzhou.aliyuncs.com';

const BUCKETS = process.argv.slice(2);
if (BUCKETS.length === 0) {
  // 顺序: 用户指定的 -> .env 里的
  BUCKETS.push('rrbrr-book-audio');
  const envBucket = process.env.OSS_BUCKET_NAME;
  if (envBucket && !BUCKETS.includes(envBucket)) BUCKETS.push(envBucket);
}

function fmtBytes(b) {
  if (b < 1024) return b + ' B';
  if (b < 1024 ** 2) return (b / 1024).toFixed(1) + ' KB';
  if (b < 1024 ** 3) return (b / 1024 ** 2).toFixed(1) + ' MB';
  return (b / 1024 ** 3).toFixed(2) + ' GB';
}

async function listAll(client, prefix) {
  const all = [];
  let marker = null;
  // 加 max-keys 1000 是 aliyun 单次上限
  do {
    const res = await client.list({ prefix, 'max-keys': 1000, marker });
    all.push(...(res.objects || []));
    marker = res.nextMarker || null;
    // isTruncated 不在 ali-oss 默认返回里，用 nextMarker 兜底
  } while (marker);
  return all;
}

async function probe(bucket) {
  console.log(`\n========== bucket: ${bucket} (region=${REGION}) ==========`);
  // 已知 bucket 实际 endpoint（list 接口对 region 严格，先试 shanghai；失败再 fallback 其他）
  const epCandidates = [
    'oss-cn-shanghai.aliyuncs.com',
    'oss-cn-shanghai-internal.aliyuncs.com',
    EP,
  ];
  let client = null, lastErr = null;
  for (const e of epCandidates) {
    try {
      const c = new OSS({ accessKeyId: AK, accessKeySecret: SK, bucket, endpoint: e, secure: true });
      await c.getBucketInfo(bucket);
      client = c;
      console.log(`  (使用 endpoint: ${e})`);
      break;
    } catch (err) { lastErr = err; }
  }
  if (!client) {
    console.log('  ❌ 所有 endpoint 都连不上:', lastErr?.code, lastErr?.message?.split('\n')[0]);
    return;
  }
  try {
    const info = await client.getBucketInfo(bucket);
    console.log('  ✅ 可访问. creationDate:', info.bucket?.CreationDate);
  } catch (e) {
    console.log('  ❌ 无法访问:', e.code || e.message);
    return;
  }

  // 1) 顶层 prefix 概览
  console.log('\n[1] 顶层一级 prefix 概览:');
  const topRes = await client.list({ 'max-keys': 1000, delimiter: '/' });
  const topPrefixes = (topRes.prefixes || []).sort();
  for (const p of topPrefixes) console.log('  📁', p);

  // 2) 详细: audio/ 目录
  if (topPrefixes.includes('audio/') || topPrefixes.some(p => p.startsWith('audio'))) {
    await drillAudio(client);
  }

  // 3) 其他目录快速占位
  for (const p of topPrefixes) {
    if (p.startsWith('audio')) continue;
    if (['cover/', 'video/', 'uploads/'].includes(p)) {
      const objs = await listAll(client, p);
      const total = objs.reduce((s, o) => s + (o.size || 0), 0);
      console.log(`  📦 ${p}  文件数=${objs.length}  占用=${fmtBytes(total)}`);
    }
  }
}

async function drillAudio(client) {
  console.log('\n[2] audio/ 目录深度扫描 (可能慢):');
  const t0 = Date.now();
  const objs = await listAll(client, 'audio/');
  const dt = ((Date.now() - t0) / 1000).toFixed(1);
  console.log(`  扫描完成,耗时 ${dt}s,文件数 ${objs.length}`);

  // 按二级目录聚合 (audio/{segment}/...)
  const bySecond = new Map();
  let totalBytes = 0;
  for (const o of objs) {
    const key = o.name;
    const parts = key.split('/');
    const seg = parts[1] || '(root)';
    const cur = bySecond.get(seg) || { count: 0, bytes: 0 };
    cur.count++;
    cur.bytes += o.size || 0;
    bySecond.set(seg, cur);
    totalBytes += o.size || 0;
  }
  console.log(`  总计: ${fmtBytes(totalBytes)} across ${bySecond.size} 二级目录`);

  // 找最大 20 个文件
  const top = [...objs].sort((a, b) => (b.size || 0) - (a.size || 0)).slice(0, 20);
  console.log('\n  Top 20 最大文件:');
  for (const o of top) {
    console.log(`    ${fmtBytes(o.size || 0).padStart(10)}  ${o.name}`);
  }

  // 按二级目录分桶统计
  const secondArr = [...bySecond.entries()]
    .map(([seg, v]) => ({ seg, ...v }))
    .sort((a, b) => b.bytes - a.bytes);

  // 看下分布: 前 10 大的二级目录 + 它们的"段类型" (uuid / preview-uuid / merged/chapter/...)
  console.log('\n  按二级目录 (前 30 大占用):');
  for (const x of secondArr.slice(0, 30)) {
    let kind = x.seg;
    if (x.seg === 'merged') kind = 'merged/ (章节合并)';
    else if (x.seg.startsWith('preview-')) kind = `preview-uuid (试听预览)  ${x.count} 文件`;
    else if (x.seg.startsWith('sync-')) kind = `sync-uuid (同步合成)  ${x.count} 文件`;
    else if (x.seg.length === 36 && x.seg.includes('-')) kind = `uuid  ${x.count} 文件`;
    else kind = `${x.seg}  ${x.count} 文件`;
    console.log(`    ${fmtBytes(x.bytes).padStart(10)}  ${kind}`);
  }

  // 统计 uuid 数量
  const uuidLike = secondArr.filter(x => /^[0-9a-f-]{36}$/.test(x.seg));
  const previewLike = secondArr.filter(x => x.seg.startsWith('preview-'));
  const syncLike = secondArr.filter(x => x.seg.startsWith('sync-'));
  const mergedTotal = secondArr.filter(x => x.seg === 'merged').reduce((s, x) => s + x.bytes, 0);

  console.log('\n[3] 分类汇总:');
  console.log(`  uuid 形态 (TTS 正常合成):       ${uuidLike.length} 个目录,  ${fmtBytes(uuidLike.reduce((s, x) => s + x.bytes, 0))}`);
  console.log(`  preview-uuid 形态 (试听预览):   ${previewLike.length} 个目录, ${fmtBytes(previewLike.reduce((s, x) => s + x.bytes, 0))}`);
  console.log(`  sync-uuid 形态 (同步合成):       ${syncLike.length} 个目录,   ${fmtBytes(syncLike.reduce((s, x) => s + x.bytes, 0))}`);
  console.log(`  merged/ 章节合并:               ${fmtBytes(mergedTotal)}`);

  // 大小分布 (找规律: 是不是大多数都是同样大小的 preview 24MB 之类的)
  const sizeBuckets = new Map();
  for (const x of secondArr) {
    const mb = x.bytes / 1024 / 1024;
    let bucket;
    if (mb < 0.1) bucket = '<0.1MB';
    else if (mb < 1) bucket = '0.1-1MB';
    else if (mb < 5) bucket = '1-5MB';
    else if (mb < 20) bucket = '5-20MB';
    else if (mb < 100) bucket = '20-100MB';
    else bucket = '>100MB';
    sizeBuckets.set(bucket, (sizeBuckets.get(bucket) || 0) + 1);
  }
  console.log('\n[4] 单目录大小分布:');
  for (const [k, v] of sizeBuckets) console.log(`  ${k.padEnd(10)} : ${v} 个目录`);
}

(async () => {
  if (!AK || !SK) {
    console.error('❌ 缺少 OSS_ACCESS_KEY_ID / OSS_ACCESS_KEY_SECRET，请确认 server/.env.production');
    process.exit(1);
  }
  for (const b of BUCKETS) await probe(b);
  console.log('\n--- 完成（只读，未删除任何文件）---\n');
})().catch(e => {
  console.error('脚本异常:', e);
  process.exit(1);
});