test(loadtest),docs: monitoring, cdn checklist and k6 framework
- monitoring.md: 指标→实现映射（1Panel+云监控+结构化日志，不引入额外监控 系统）；pino 字段表如实列自代码（访问日志 req.id/res.statusCode/ responseTime + sync/worker 自定义 msg 字段）；六类告警 jq 检索语句与 阈值建议总表；日志保留（compose json-file 10m×3 已设 + 每日归档建议） - cdn-checklist.md: 全步骤待凭证执行；图片加速三选一架构核对（双层签名 冲突分析，推荐 c 起步=CDN 仅加速 viewer 静态、presigned 直链不变， 图片收益后评估 b=CDN 鉴权+私有桶回源授权）；缓存规则对齐源站 immutable； 失效/预热 CLI 样例与多地域冷缓存测试点模板 - loadtest/: k6 框架——read-feed.js（N 观众四 mix：latest/hot 分页+大图+ 点赞+浏览，ramp 阶段）、upload-burst.js（M 并发上传全链含 complete 与 发布/同步可选轮询，兼作数据种子）、lib/helpers.js（登录/建册/上传全链 +内嵌 JPEG fixture 与 prod-smoke 同源）、README（运行方式/参数表/ P95 输出解读对齐计划 §6 候选指标；D05 未定→阈值占位注释） Co-Authored-By: Claude Code <noreply@anthropic.com>
This commit is contained in:
浅念
@@ -0,0 +1,96 @@
|
||||
# loadtest —— k6 压测框架（P5-prep Task 3）
|
||||
|
||||
对齐项目计划 §6 首轮容量验证建议的 L1/L2 阶梯（L1 基础：相册 1000 照片 /
|
||||
3 人上传 / 100 观众；L2 目标候选：10000 / 10 / 1000；L3 扩展探测不设默认门槛）。
|
||||
**D05（数量并发与网络目标）未定——所有 thresholds 为占位注释**，确认后取消
|
||||
注释启用（候选口径见 §3）。
|
||||
|
||||
## 文件
|
||||
|
||||
| 文件 | 用途 |
|
||||
| --- | --- |
|
||||
| `read-feed.js` | 观众读路径：N 观众（ramp 阶段）持续分页读最新 + 热门切换 + 大图请求 + 点赞批次 + 浏览上报四 mix |
|
||||
| `upload-burst.js` | 工作人员 M 并发小图上传全链（presign → OSS PUT → complete），可选轮询发布可见 / ov 同步；兼作数据种子 |
|
||||
| `lib/helpers.js` | 登录 / 建相册 / 上传全链 / 分页 / 点赞 / 浏览 / 轮询助手 + 内嵌 800x600 JPEG fixture（与 `scripts/prod-smoke.sh` 同源） |
|
||||
|
||||
安装：[k6](https://grafana.com/docs/k6/latest/set-up/install/) ≥ 0.45（Windows/
|
||||
macOS/Linux 均有安装包）。脚本在仓库根 `loadtest/` 下直接 `k6 run`。
|
||||
|
||||
## 运行方式
|
||||
|
||||
```bash
|
||||
# ① 种子：建时间戳相册并上传照片（L1 档：3 人 × 20 张；setup 会打印 ALBUM_TOKEN）
|
||||
k6 run --env BASE_CN=https://photo.example.cn \
|
||||
--env ACCOUNT=admin --env PASSWORD='…' \
|
||||
--env VUS=3 --env ITERATIONS=20 upload-burst.js
|
||||
|
||||
# ② 观众读路径压测（L1 档：100 观众；用 ① 打印的 ALBUM_TOKEN）
|
||||
k6 run --env BASE=https://photo.example.com --env ALBUM_TOKEN=<打印值> \
|
||||
--env VUS=100 --env STEADY=5m read-feed.js
|
||||
|
||||
# ②' L2 档：观众 1000 / 上传 10 人（先种更大相册：10 × 100）
|
||||
k6 run --env BASE=… --env ALBUM_TOKEN=… --env VUS=1000 --env STEADY=10m read-feed.js
|
||||
|
||||
# ③ 上传链含发布观测（传输完成→发布；ov 同步端到端可选）
|
||||
k6 run --env BASE_CN=… --env ACCOUNT=… --env PASSWORD=… \
|
||||
--env ALBUM_ID=<albumId> --env ALBUM_TOKEN=<token> \
|
||||
--env WAIT_PUBLISH=1 --env WAIT_SYNC=1 --env BASE_OV=https://photo.example.com \
|
||||
upload-burst.js
|
||||
|
||||
# 结果归档：--out json=results-$(date +%F-%H%M).json 或 --summary-export=summary.json
|
||||
```
|
||||
|
||||
### 参数总表（--env KEY=VALUE）
|
||||
|
||||
| KEY | 脚本 | 缺省 | 说明 |
|
||||
| --- | --- | --- | --- |
|
||||
| `BASE` | 两者 | 必填（read）/ upload 的 BASE_CN 回退 | 被测 API 基址（read 用 ov 域，upload 用 cn 域），末尾不带 `/v1` |
|
||||
| `ALBUM_TOKEN` | read 必填 / upload 可选 | — | 公开相册 token（种子模式首次运行打印） |
|
||||
| `ACCOUNT` / `PASSWORD` | upload | 必填 | 工作人员（或 admin）账号 |
|
||||
| `ALBUM_ID` | upload | 空=建新册 | 复用相册；配合 `ALBUM_TOKEN` 才有发布轮询 |
|
||||
| `VUS` | 两者 | 100 / 3 | 观众数 / 上传并发人数 |
|
||||
| `ITERATIONS` | upload | 20 | 每人上传张数（L1 合计 = VUS×ITERATIONS） |
|
||||
| `STEADY` / `RAMP_UP` / `RAMP_DOWN` | read | 5m / 1m / 1m | ramp 阶段时长 |
|
||||
| `PAGES` / `PAGE_LIMIT` | read | 3 / 30 | 翻页深度 / 页大小（契约上限 60） |
|
||||
| `HOT_SHARE` / `BIGIMG_SHARE` / `LIKE_SHARE` | read | 0.5 / 0.3 / 0.4 | 迭代内行为混合概率 |
|
||||
| `THINK_MIN` / `THINK_MAX` | read | 0.5 / 3 | 观众思考时间秒 |
|
||||
| `WAIT_PUBLISH` / `WAIT_SYNC` / `BASE_OV` | upload | 0 / 0 / 空 | 公开可见轮询开关（cn 发布 / ov 同步） |
|
||||
| `POLL_TIMEOUT` / `POLL_INTERVAL` | upload | 60 / 2 | 轮询上限/间隔秒 |
|
||||
|
||||
## 输出解读（P95 关注点 ←→ 项目计划 §6 候选指标）
|
||||
|
||||
| 候选指标（计划原文） | k6 指标名 | 说明 |
|
||||
| --- | --- | --- |
|
||||
| 普通列表 API P95 < 500ms | `feed_list_duration` / `feed_hot_duration` | 按 tag `list_public_photos(_hot)` 分组的 Trend；「首批照片 P95<3s」是浏览器端口径（含渲染/图载），k6 以首页列表请求近似 + §冷缓存另测 |
|
||||
| 传输完成到发布 P95 < 10s | `publish_wait_duration` | `WAIT_PUBLISH=1` 时 complete → cn 公开列表可见的等待；上传耗时（`upload_oss_put_duration`）单独统计，不计入处理时延（计划注记） |
|
||||
| （同上，跨境） | `ov_sync_wait_duration` | `WAIT_SYNC=1`：complete → ov 公开可见（含推送 2s 间隔） |
|
||||
| 点赞确认及广播 P95 < 2s | `like_confirm_duration` | API 确认口径；WS 广播时延属前端观测面，k6 不覆盖 |
|
||||
| — | `upload_presign_duration` / `upload_complete_duration` | 上传链分步；`upload_chain_duration` 为三步端到端 |
|
||||
| — | `like_rejected_total` / `publish_timeout_total` | 非目标信号：前者多为限速触发（见下），后者=发布超时需查 worker |
|
||||
|
||||
通用看板项：`http_req_duration`（按 name tag 细分）、`http_req_failed`、
|
||||
`checks` 通过率、VU 数与迭代吞吐（`iteration_duration` 倒数）。**冷缓存/地域
|
||||
时延不在 k6 范围**——用 `infra/deploy/cdn-checklist.md` §6 的多地域 curl 记录表。
|
||||
|
||||
## 阈值占位（D05 确认后启用）
|
||||
|
||||
两脚本 `options.thresholds` 均为注释占位（候选值即上表）。确认 D05 后：
|
||||
1. 取消注释并按定案值调整；
|
||||
2. L2 档跑 `k6 run … --tag testid=l2-$(date +%F)` 便于结果对比归档；
|
||||
3. 阈值失败会让 k6 以非零码退出——可直接作为 P5 容量报告的机器判定。
|
||||
|
||||
## 操作注意（如实阅读再跑）
|
||||
|
||||
- **限速会污染指标**：likes 有 Redis 滑窗限速（`LIKE_IP_WINDOW_DELTA=2000` /
|
||||
24h / IP；`LIKE_SESSION_WINDOW_DELTA=500` / 会话）——高并发点赞从单一压测机
|
||||
出发很快触发 429（计入 `like_rejected_total`，且账面不是真实服务容量）。
|
||||
staging 压测调大对应 env 或 `LIKE_SHARE=0`；需要真实验证限速本身时反向利用。
|
||||
visits 的键含 visitorId（每 VU 唯一），`VISIT_WINDOW_REQUESTS=60`/窗 通常够用。
|
||||
- **反代后 IP 归并**：经 OpenResty 反代时服务端 req.ip 是网关地址，trust proxy
|
||||
未配置前所有限速按网关 IP 聚合——压测读数要意识到这一点（登记为 P5 项，
|
||||
likes.controller.ts 文件头同注）。
|
||||
- **对象存储是外部依赖**：`oss_put` / `oss_image_get` 打到 MinIO/OSS/CDN 直链，
|
||||
其时延与被测 API 分开解读（tag 已区分）；生产桶上跑会产生真实流量费用。
|
||||
- **不自动清理**：种子照片留在相册内（时间戳标题可辨认），清理=后台软删或
|
||||
`DELETE /v1/photos/{id}`（prod-smoke 同款）；相册无删除端点。
|
||||
- 对生产环境执行前必须获得明确许可——上传/点赞/浏览写路径全部落真实数据。
|
||||
@@ -0,0 +1,220 @@
|
||||
// loadtest/lib/helpers.js —— k6 公共助手（P5-prep Task 3）。
|
||||
// 被 read-feed.js / upload-burst.js 复用：登录 / 建相册 / 上传全链 / 公开读
|
||||
// 分页 / 点赞批次 / 浏览上报 / 公开可见轮询。端点与请求体形态对齐
|
||||
// scripts/prod-smoke.sh（契约权威 = packages/contracts/openapi.yaml）。
|
||||
//
|
||||
// 内置测试图与 prod-smoke.sh 同一张：800x600 JPEG（quality 55，3118 字节），
|
||||
// base64 内嵌（下方 JPEG_B64 由 scripts/prod-smoke.sh 同步生成，勿手改），
|
||||
// k6 侧用 k6/encoding 解码为二进制。刻意不用 1x1——避免变体流水线踩最小
|
||||
// 尺寸边界。
|
||||
import http from 'k6/http';
|
||||
import { sleep } from 'k6';
|
||||
import encoding from 'k6/encoding';
|
||||
|
||||
export const JPEG_B64 = '/9j/2wBDAA4KCw0LCQ4NDA0QDw4RFiQXFhQUFiwgIRokNC43NjMuMjI6QVNGOj1OPjIySGJJTlZYXV5dOEVmbWVabFNbXVn/2wBDAQ8QEBYTFioXFypZOzI7WVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVlZWVn/wAARCAJYAyADASIAAhEBAxEB/8QAFQABAQAAAAAAAAAAAAAAAAAAAAX/xAAUEAEAAAAAAAAAAAAAAAAAAAAA/8QAFQEBAQAAAAAAAAAAAAAAAAAAAAX/xAAUEQEAAAAAAAAAAAAAAAAAAAAA/9oADAMBAAIRAxEAPwCsAkKIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD//2Q==';
|
||||
|
||||
/** 解码一次，全脚本复用（Uint8Array 可直接作 http 请求体）。 */
|
||||
export const JPEG_BYTES = new Uint8Array(encoding.b64decode(JPEG_B64, 'binary'));
|
||||
|
||||
// ── env 参数读取（--env KEY=VALUE → __ENV.KEY；缺省回退）──
|
||||
|
||||
export function envStr(name, fallback) {
|
||||
const v = __ENV[name];
|
||||
return v === undefined || v === '' ? fallback : v;
|
||||
}
|
||||
|
||||
export function envInt(name, fallback) {
|
||||
const raw = __ENV[name];
|
||||
if (raw === undefined || raw === '') {
|
||||
return fallback;
|
||||
}
|
||||
const n = Number.parseInt(raw, 10);
|
||||
return Number.isFinite(n) && n > 0 ? n : fallback;
|
||||
}
|
||||
|
||||
export function envFloat(name, fallback) {
|
||||
const raw = __ENV[name];
|
||||
if (raw === undefined || raw === '') {
|
||||
return fallback;
|
||||
}
|
||||
const n = Number.parseFloat(raw);
|
||||
return Number.isFinite(n) && n >= 0 ? n : fallback;
|
||||
}
|
||||
|
||||
/** 压测内唯一键（clientTaskId/batchId/requestId 等幂等键；时间戳+随机足够）。 */
|
||||
export function uniqId(prefix) {
|
||||
return `${prefix}-${Date.now().toString(36)}-${Math.floor(Math.random() * 1e9).toString(36)}`;
|
||||
}
|
||||
|
||||
// ── 认证与相册 ──
|
||||
|
||||
export function authHeaders(token) {
|
||||
return { Authorization: `Bearer ${token}`, 'Content-Type': 'application/json' };
|
||||
}
|
||||
|
||||
/** 密码登录 → sessionToken（响应体 {sessionToken,...}，同 prod-smoke ②）。 */
|
||||
export function login(base, account, password) {
|
||||
const res = http.post(
|
||||
`${base}/v1/auth/password`,
|
||||
JSON.stringify({ account, password }),
|
||||
{ headers: { 'Content-Type': 'application/json' }, tags: { name: 'auth_password' } },
|
||||
);
|
||||
if (res.status !== 200) {
|
||||
throw new Error(`登录失败 HTTP ${res.status}: ${res.body}`);
|
||||
}
|
||||
const token = res.json('sessionToken');
|
||||
if (!token) {
|
||||
throw new Error(`登录响应缺 sessionToken: ${res.body}`);
|
||||
}
|
||||
return token;
|
||||
}
|
||||
|
||||
/** 建时间戳相册 → { albumId, publicToken }（同 prod-smoke ③）。 */
|
||||
export function createAlbum(base, token, title, description) {
|
||||
const res = http.post(
|
||||
`${base}/v1/albums`,
|
||||
JSON.stringify({
|
||||
translations: [
|
||||
{ locale: 'zh-CN', title, description: description || 'k6 loadtest fixture' },
|
||||
],
|
||||
defaultLocale: 'zh-CN',
|
||||
startAt: new Date().toISOString(),
|
||||
}),
|
||||
{ headers: authHeaders(token), tags: { name: 'albums_create' } },
|
||||
);
|
||||
if (res.status !== 201) {
|
||||
throw new Error(`建相册失败 HTTP ${res.status}: ${res.body}`);
|
||||
}
|
||||
return { albumId: res.json('id'), publicToken: res.json('publicToken') };
|
||||
}
|
||||
|
||||
// ── 上传全链（presign → OSS PUT → complete）──
|
||||
|
||||
/**
|
||||
* 单照片全链上传。返回逐步耗时（毫秒，k6 timings）与 id，供脚本各自打点：
|
||||
* { photoId, uploadId, timings: { presign, put, complete, chainMs } }
|
||||
* 任一步非预期状态码即抛错（k6 计一次 error 并终止单迭代）。
|
||||
*/
|
||||
export function uploadPhoto(base, token, albumId, clientTaskId) {
|
||||
const started = Date.now();
|
||||
const presign = http.post(
|
||||
`${base}/v1/uploads`,
|
||||
JSON.stringify({
|
||||
albumId,
|
||||
clientTaskId,
|
||||
filename: 'k6.jpg',
|
||||
bytes: JPEG_BYTES.byteLength,
|
||||
contentType: 'image/jpeg',
|
||||
}),
|
||||
{ headers: authHeaders(token), tags: { name: 'uploads_presign' } },
|
||||
);
|
||||
if (presign.status !== 201) {
|
||||
throw new Error(`presign 失败 HTTP ${presign.status}: ${presign.body}`);
|
||||
}
|
||||
const uploadId = presign.json('uploadId');
|
||||
const photoId = presign.json('photoId');
|
||||
const url = presign.json('url');
|
||||
if (!uploadId || !url) {
|
||||
throw new Error(`presign 响应缺 uploadId/url: ${presign.body}`);
|
||||
}
|
||||
// PUT 目标是 presigned 对象存储 URL（staging=MinIO / 生产=OSS），非被测 API 域。
|
||||
const put = http.put(url, JPEG_BYTES, {
|
||||
headers: { 'Content-Type': 'image/jpeg' },
|
||||
tags: { name: 'oss_put' },
|
||||
});
|
||||
if (put.status < 200 || put.status > 299) {
|
||||
throw new Error(`OSS PUT 失败 HTTP ${put.status}`);
|
||||
}
|
||||
const complete = http.post(
|
||||
`${base}/v1/uploads/${uploadId}/complete`,
|
||||
'',
|
||||
{ headers: authHeaders(token), tags: { name: 'uploads_complete' } },
|
||||
);
|
||||
if (complete.status !== 200) {
|
||||
throw new Error(`complete 失败 HTTP ${complete.status}: ${complete.body}`);
|
||||
}
|
||||
return {
|
||||
photoId,
|
||||
uploadId,
|
||||
timings: {
|
||||
presign: presign.timings.duration,
|
||||
put: put.timings.duration,
|
||||
complete: complete.timings.duration,
|
||||
chainMs: Date.now() - started,
|
||||
},
|
||||
};
|
||||
}
|
||||
|
||||
// ── 观众端四类操作（ov 域公开端点；亦可打 cn 做对照）──
|
||||
|
||||
/**
|
||||
* 公开照片列表一页（契约：limit 1..60 缺省 30；cursor 不透明需 encode；
|
||||
* sort=latest|hot 仅 ov 实现非法值 400）。tag 参数让脚本能按 name 分组统计。
|
||||
*/
|
||||
export function fetchFeedPage(base, albumToken, options) {
|
||||
const opts = options || {};
|
||||
let qs = `limit=${opts.limit || 30}`;
|
||||
if (opts.cursor) {
|
||||
qs += `&cursor=${encodeURIComponent(opts.cursor)}`;
|
||||
}
|
||||
if (opts.sort) {
|
||||
qs += `&sort=${opts.sort}`;
|
||||
}
|
||||
return http.get(`${base}/v1/public/albums/${albumToken}/photos?${qs}`, {
|
||||
tags: { name: opts.tag || 'list_public_photos' },
|
||||
});
|
||||
}
|
||||
|
||||
/** 浏览上报（匿名；visitorId 客户端持久 8..128 字符，requestId 幂等键）。 */
|
||||
export function reportVisit(base, albumToken, visitorId, requestId) {
|
||||
return http.post(
|
||||
`${base}/v1/public/albums/${albumToken}/visits`,
|
||||
JSON.stringify({ visitorId, requestId }),
|
||||
{
|
||||
headers: { 'Content-Type': 'application/json' },
|
||||
tags: { name: 'visit_report' },
|
||||
},
|
||||
);
|
||||
}
|
||||
|
||||
/** 点赞批次（匿名；sessionId 限速键 / batchId 幂等键 / delta 1..20 整数）。 */
|
||||
export function likeBatch(base, albumToken, photoId, sessionId, batchId, delta) {
|
||||
return http.post(
|
||||
`${base}/v1/public/albums/${albumToken}/photos/${photoId}/likes`,
|
||||
JSON.stringify({ sessionId, batchId, delta }),
|
||||
{
|
||||
headers: { 'Content-Type': 'application/json' },
|
||||
tags: { name: 'like_batch' },
|
||||
},
|
||||
);
|
||||
}
|
||||
|
||||
/**
|
||||
* 轮询公开列表直到 photoId 可见（worker 发布事务异步 + cn→ov 推送 2s 间隔；
|
||||
* 同 prod-smoke ④ 口径）。返回 { found, waitedMs, polls }。
|
||||
*/
|
||||
export function pollUntilVisible(
|
||||
base,
|
||||
albumToken,
|
||||
photoId,
|
||||
timeoutSec,
|
||||
intervalSec,
|
||||
tag,
|
||||
) {
|
||||
const started = Date.now();
|
||||
let polls = 0;
|
||||
for (;;) {
|
||||
polls += 1;
|
||||
const res = fetchFeedPage(base, albumToken, { limit: 30, tag: tag || 'public_read_poll' });
|
||||
if (res.status === 200) {
|
||||
const items = res.json('items');
|
||||
if (Array.isArray(items) && items.some((it) => it.photoId === photoId)) {
|
||||
return { found: true, waitedMs: Date.now() - started, polls };
|
||||
}
|
||||
}
|
||||
if (Date.now() - started > timeoutSec * 1000) {
|
||||
return { found: false, waitedMs: Date.now() - started, polls };
|
||||
}
|
||||
sleep(intervalSec);
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,172 @@
|
||||
// loadtest/read-feed.js —— L1/L2 观众读路径压测场景（P5-prep Task 3）。
|
||||
//
|
||||
// 模拟 N 名观众（VU=观众，ramp 阶段爬坡-稳态-退坡）在 ov 域公开相册上的四类
|
||||
// 行为混合：持续分页读最新（latest）/ 热门（hot）切换 / 大图请求（presigned
|
||||
// 直链打到对象存储，单独 tag 区分）/ 点赞批次 / 浏览上报（每 VU 一次，
|
||||
// 30min 去重窗口内重复上报 counted=false 属正常语义）。
|
||||
//
|
||||
// 用法（README.md 有完整参数表）：
|
||||
// k6 run --env BASE=https://photo.example.com --env ALBUM_TOKEN=xxx \
|
||||
// --env VUS=100 --env STEADY=5m read-feed.js
|
||||
//
|
||||
// 阈值（thresholds）为占位注释：D05（数量并发与网络目标）未定，候选口径
|
||||
// 对齐项目计划 §6 候选指标，确认后在 options.thresholds 取消注释启用。
|
||||
import http from 'k6/http';
|
||||
import exec from 'k6/execution';
|
||||
import { check, sleep } from 'k6';
|
||||
import { Counter, Trend } from 'k6/metrics';
|
||||
import {
|
||||
envFloat,
|
||||
envInt,
|
||||
envStr,
|
||||
fetchFeedPage,
|
||||
likeBatch,
|
||||
reportVisit,
|
||||
uniqId,
|
||||
} from './lib/helpers.js';
|
||||
|
||||
// ── 参数（全部可 --env 覆盖；缺省 = L1 基础档）──
|
||||
const BASE = envStr('BASE', '');
|
||||
const ALBUM_TOKEN = envStr('ALBUM_TOKEN', '');
|
||||
const VUS = envInt('VUS', 100); // L1=100 / L2=1000（在线观众）
|
||||
const RAMP_UP = envStr('RAMP_UP', '1m');
|
||||
const STEADY = envStr('STEADY', '5m');
|
||||
const RAMP_DOWN = envStr('RAMP_DOWN', '1m');
|
||||
const PAGES = envInt('PAGES', 3); // 每迭代翻页深度
|
||||
const PAGE_LIMIT = envInt('PAGE_LIMIT', 30); // 契约上限 60
|
||||
const HOT_SHARE = envFloat('HOT_SHARE', 0.5); // 迭代内切热门排序概率
|
||||
const BIGIMG_SHARE = envFloat('BIGIMG_SHARE', 0.3); // 迭代内取大图概率
|
||||
const LIKE_SHARE = envFloat('LIKE_SHARE', 0.4); // 迭代内点赞批次概率
|
||||
const THINK_MIN = envFloat('THINK_MIN', 0.5); // 观众思考时间（秒）
|
||||
const THINK_MAX = envFloat('THINK_MAX', 3);
|
||||
|
||||
// ── 指标（按操作分组，输出解读见 README）──
|
||||
const listLatency = new Trend('feed_list_duration'); // latest 分页请求
|
||||
const hotLatency = new Trend('feed_hot_duration'); // hot 排序页
|
||||
const imageLatency = new Trend('image_get_duration'); // 对象存储直链大图（外部依赖）
|
||||
const likeLatency = new Trend('like_confirm_duration'); // 点赞批次确认
|
||||
const visitLatency = new Trend('visit_report_duration'); // 浏览上报
|
||||
const likeAccepted = new Counter('like_accepted_total');
|
||||
const likeRejected = new Counter('like_rejected_total'); // 429 限速 / batchAccepted=false
|
||||
const listBroken = new Counter('feed_list_error_total'); // 非 200 / 解析失败
|
||||
|
||||
export const options = {
|
||||
scenarios: {
|
||||
viewers: {
|
||||
executor: 'ramping-vus',
|
||||
startVUs: 0,
|
||||
stages: [
|
||||
{ duration: RAMP_UP, target: VUS },
|
||||
{ duration: STEADY, target: VUS },
|
||||
{ duration: RAMP_DOWN, target: 0 },
|
||||
],
|
||||
gracefulRampDown: '30s',
|
||||
},
|
||||
},
|
||||
// thresholds 占位 —— D05 确认后取消注释（候选口径 = 项目计划 §6）：
|
||||
// 普通列表 API P95 < 500ms；点赞确认 P95 < 2s（广播链路另由前端观测）。
|
||||
// thresholds: {
|
||||
// feed_list_duration: ['p(95)<500'],
|
||||
// feed_hot_duration: ['p(95)<500'],
|
||||
// like_confirm_duration: ['p(95)<2000'],
|
||||
// visit_report_duration: ['p(95)<500'],
|
||||
// 'checks': ['rate>0.99'],
|
||||
// http_req_failed: ['rate<0.01'],
|
||||
// },
|
||||
};
|
||||
|
||||
// k6 每个 VU 独立初始化模块：以下状态天然按 VU 隔离（观众身份）。
|
||||
const visitorId = uniqId(`visitor-${exec.vu.idInTest}`);
|
||||
const likeSessionId = uniqId(`sess-${exec.vu.idInTest}`);
|
||||
let visitReported = false;
|
||||
|
||||
export function setup() {
|
||||
if (!BASE || !ALBUM_TOKEN) {
|
||||
throw new Error('缺少 --env BASE=<ov 基址> 与 --env ALBUM_TOKEN=<公开相册 token>');
|
||||
}
|
||||
const res = http.get(`${BASE}/v1/public/albums/${ALBUM_TOKEN}`);
|
||||
if (res.status !== 200) {
|
||||
throw new Error(`相册不可访问（HTTP ${res.status}）：核对 BASE/ALBUM_TOKEN`);
|
||||
}
|
||||
return { albumTitle: res.json('title') };
|
||||
}
|
||||
|
||||
export default function () {
|
||||
// ① 浏览上报（每观众会话一次；窗口语义见文件头）
|
||||
if (!visitReported) {
|
||||
visitReported = true;
|
||||
const res = reportVisit(BASE, ALBUM_TOKEN, visitorId, uniqId('vr'));
|
||||
visitLatency.add(res.timings.duration);
|
||||
check(res, { 'visit report 200': (r) => r.status === 200 });
|
||||
}
|
||||
|
||||
// ② 最新分页（游标推进至 nextCursor=null 或 PAGES 页）
|
||||
const items = [];
|
||||
let cursor;
|
||||
for (let page = 0; page < PAGES && cursor !== null; page += 1) {
|
||||
const res = fetchFeedPage(BASE, ALBUM_TOKEN, {
|
||||
cursor,
|
||||
limit: PAGE_LIMIT,
|
||||
tag: 'list_public_photos',
|
||||
});
|
||||
listLatency.add(res.timings.duration);
|
||||
if (
|
||||
!check(res, {
|
||||
'list 200': (r) => r.status === 200,
|
||||
'list has items': (r) => Array.isArray(r.json('items')),
|
||||
})
|
||||
) {
|
||||
listBroken.add(1);
|
||||
break;
|
||||
}
|
||||
const body = res.json();
|
||||
for (const item of body.items) {
|
||||
items.push(item);
|
||||
}
|
||||
cursor = body.nextCursor; // null = 末页，循环条件止
|
||||
}
|
||||
|
||||
// ③ 热门排序一页（sort=hot 仅 ov 实现）
|
||||
if (Math.random() < HOT_SHARE) {
|
||||
const res = fetchFeedPage(BASE, ALBUM_TOKEN, {
|
||||
limit: PAGE_LIMIT,
|
||||
sort: 'hot',
|
||||
tag: 'list_public_photos_hot',
|
||||
});
|
||||
hotLatency.add(res.timings.duration);
|
||||
check(res, { 'hot list 200': (r) => r.status === 200 });
|
||||
}
|
||||
|
||||
// ④ 大图请求（presigned 直链 → 对象存储/CDN，非被测 API；tag 区分解读）
|
||||
if (items.length > 0 && Math.random() < BIGIMG_SHARE) {
|
||||
const item = items[Math.floor(Math.random() * items.length)];
|
||||
const url = item.browse && item.browse.url;
|
||||
if (url) {
|
||||
const res = http.get(url, { tags: { name: 'oss_image_get' } });
|
||||
imageLatency.add(res.timings.duration);
|
||||
check(res, { 'image 200': (r) => r.status === 200 });
|
||||
}
|
||||
}
|
||||
|
||||
// ⑤ 点赞批次（匿名端点；限速口径见 README——压测会话请注意 LIKE_IP_WINDOW_DELTA）
|
||||
if (items.length > 0 && Math.random() < LIKE_SHARE) {
|
||||
const item = items[Math.floor(Math.random() * items.length)];
|
||||
const delta = 1 + Math.floor(Math.random() * 3); // 契约 1..20，取保守值
|
||||
const res = likeBatch(
|
||||
BASE,
|
||||
ALBUM_TOKEN,
|
||||
item.photoId,
|
||||
likeSessionId,
|
||||
uniqId('batch'),
|
||||
delta,
|
||||
);
|
||||
likeLatency.add(res.timings.duration);
|
||||
if (res.status === 200 && res.json('batchAccepted') === true) {
|
||||
likeAccepted.add(1);
|
||||
} else {
|
||||
likeRejected.add(1); // 429=限速触发（压测口径问题，见 README）；非 200=链路异常
|
||||
}
|
||||
}
|
||||
|
||||
sleep(THINK_MIN + Math.random() * (THINK_MAX - THINK_MIN));
|
||||
}
|
||||
@@ -0,0 +1,144 @@
|
||||
// loadtest/upload-burst.js —— L1/L2 工作人员并发上传压测场景（P5-prep Task 3）。
|
||||
//
|
||||
// M 名工作人员（VU）并发执行小图上传全链：presign → 对象存储 PUT → complete，
|
||||
// 并可选轮询「公开可见」（传输完成→发布，worker 变体流水线 + 发布事务异步）
|
||||
// 与「ov 侧可见」（cn→ov 同步端到端）。数据面与 scripts/prod-smoke.sh ③④ 同径。
|
||||
//
|
||||
// 兼作 read-feed.js 的数据种子：ALBUM_ID 缺省时 setup 建时间戳相册并打印
|
||||
// albumId/publicToken——用打印出的 ALBUM_TOKEN 跑 read-feed.js 即得固定测试相册。
|
||||
//
|
||||
// 用法（README.md 有完整参数表）：
|
||||
// k6 run --env BASE_CN=https://photo.example.cn --env ACCOUNT=a --env PASSWORD=p \
|
||||
// --env VUS=3 --env ITERATIONS=20 --env WAIT_PUBLISH=1 upload-burst.js
|
||||
//
|
||||
// 注意：fixtures 用内嵌 800x600 小图（3118 字节）——只压「链路与并发正确性」；
|
||||
// 真实尺寸图片的吞吐/处理时延另用真实样本按计划 §6 注记单独验证。
|
||||
import { Counter, Trend } from 'k6/metrics';
|
||||
import { check } from 'k6';
|
||||
import {
|
||||
createAlbum,
|
||||
envInt,
|
||||
envStr,
|
||||
login,
|
||||
pollUntilVisible,
|
||||
uniqId,
|
||||
uploadPhoto,
|
||||
} from './lib/helpers.js';
|
||||
|
||||
// ── 参数 ──
|
||||
const BASE_CN = envStr('BASE_CN', envStr('BASE', ''));
|
||||
const BASE_OV = envStr('BASE_OV', ''); // 留空 = 跳过 ov 同步观测
|
||||
const ACCOUNT = envStr('ACCOUNT', '');
|
||||
const PASSWORD = envStr('PASSWORD', '');
|
||||
const ALBUM_ID = envStr('ALBUM_ID', ''); // 空 = setup 建新相册（种子模式）
|
||||
const VUS = envInt('VUS', 3); // L1=3 / L2=10（同时上传人数）
|
||||
const ITERATIONS = envInt('ITERATIONS', 20); // 每人上传张数
|
||||
const WAIT_PUBLISH = envStr('WAIT_PUBLISH', '0') === '1'; // 轮询 cn 公开可见
|
||||
const WAIT_SYNC = envStr('WAIT_SYNC', '0') === '1'; // 轮询 ov 公开可见（需 BASE_OV）
|
||||
const POLL_TIMEOUT = envInt('POLL_TIMEOUT', 60); // 轮询上限秒
|
||||
const POLL_INTERVAL = envInt('POLL_INTERVAL', 2); // 轮询间隔秒
|
||||
|
||||
// ── 指标 ──
|
||||
const presignLatency = new Trend('upload_presign_duration');
|
||||
const putLatency = new Trend('upload_oss_put_duration'); // 对象存储外部依赖
|
||||
const completeLatency = new Trend('upload_complete_duration');
|
||||
const chainLatency = new Trend('upload_chain_duration'); // 三步端到端
|
||||
const publishWait = new Trend('publish_wait_duration'); // complete → cn 公开可见
|
||||
const syncWait = new Trend('ov_sync_wait_duration'); // complete → ov 公开可见
|
||||
const publishTimeout = new Counter('publish_timeout_total');
|
||||
const syncTimeout = new Counter('ov_sync_timeout_total');
|
||||
const uploadFailed = new Counter('upload_failed_total');
|
||||
|
||||
export const options = {
|
||||
scenarios: {
|
||||
uploaders: {
|
||||
executor: 'per-vu-iterations',
|
||||
vus: VUS,
|
||||
iterations: ITERATIONS,
|
||||
maxDuration: '30m',
|
||||
},
|
||||
},
|
||||
// thresholds 占位 —— D05 确认后取消注释（候选口径 = 项目计划 §6：
|
||||
// 传输完成到发布 P95 < 10s，上传耗时单独统计不计入处理时延）：
|
||||
// thresholds: {
|
||||
// upload_presign_duration: ['p(95)<500'],
|
||||
// upload_complete_duration: ['p(95)<2000'],
|
||||
// publish_wait_duration: ['p(95)<10000'],
|
||||
// http_req_failed: ['rate<0.01'],
|
||||
// },
|
||||
};
|
||||
|
||||
export function setup() {
|
||||
if (!BASE_CN || !ACCOUNT || !PASSWORD) {
|
||||
throw new Error('缺少 --env BASE_CN=<cn 基址> --env ACCOUNT=… --env PASSWORD=…');
|
||||
}
|
||||
if (WAIT_SYNC && !BASE_OV) {
|
||||
throw new Error('WAIT_SYNC=1 需要 --env BASE_OV=<ov 基址>');
|
||||
}
|
||||
const token = login(BASE_CN, ACCOUNT, PASSWORD);
|
||||
let albumId = ALBUM_ID;
|
||||
let publicToken = '';
|
||||
if (!albumId) {
|
||||
const album = createAlbum(BASE_CN, token, `k6-upload-${Date.now()}`);
|
||||
albumId = album.albumId;
|
||||
publicToken = album.publicToken;
|
||||
console.log(
|
||||
`[seed] 已建相册 albumId=${albumId} ALBUM_TOKEN=${publicToken}` +
|
||||
'（read-feed.js 用此 token；照片不会自动清理，软删见 README）',
|
||||
);
|
||||
} else {
|
||||
// 复用既有相册：publicToken 由调用方提供（种子模式首次运行会打印）；
|
||||
// 未提供则 WAIT_PUBLISH/WAIT_SYNC 轮询自动跳过（默认函数内按空值守护）。
|
||||
publicToken = envStr('ALBUM_TOKEN', '');
|
||||
console.log(
|
||||
`[seed] 复用相册 albumId=${albumId}` +
|
||||
(publicToken ? ` ALBUM_TOKEN=${publicToken}` : '（未传 ALBUM_TOKEN，跳过公开可见轮询）'),
|
||||
);
|
||||
}
|
||||
return { token, albumId, publicToken };
|
||||
}
|
||||
|
||||
export default function (data) {
|
||||
try {
|
||||
const result = uploadPhoto(BASE_CN, data.token, data.albumId, uniqId('k6up'));
|
||||
presignLatency.add(result.timings.presign);
|
||||
putLatency.add(result.timings.put);
|
||||
completeLatency.add(result.timings.complete);
|
||||
chainLatency.add(result.timings.chainMs);
|
||||
check(result, { 'upload chain ok': (r) => Boolean(r.photoId) });
|
||||
|
||||
if (WAIT_PUBLISH && data.publicToken && result.photoId) {
|
||||
const poll = pollUntilVisible(
|
||||
BASE_CN,
|
||||
data.publicToken,
|
||||
result.photoId,
|
||||
POLL_TIMEOUT,
|
||||
POLL_INTERVAL,
|
||||
'publish_poll_cn',
|
||||
);
|
||||
if (poll.found) {
|
||||
publishWait.add(poll.waitedMs);
|
||||
} else {
|
||||
publishTimeout.add(1);
|
||||
}
|
||||
}
|
||||
if (WAIT_SYNC && BASE_OV && data.publicToken && result.photoId) {
|
||||
const poll = pollUntilVisible(
|
||||
BASE_OV,
|
||||
data.publicToken,
|
||||
result.photoId,
|
||||
POLL_TIMEOUT,
|
||||
POLL_INTERVAL,
|
||||
'publish_poll_ov',
|
||||
);
|
||||
if (poll.found) {
|
||||
syncWait.add(poll.waitedMs);
|
||||
} else {
|
||||
syncTimeout.add(1);
|
||||
}
|
||||
}
|
||||
} catch (err) {
|
||||
uploadFailed.add(1);
|
||||
console.error(`upload chain failed: ${err.message}`);
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user