OCDP/ocdp-go
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

dev #1

Merged
ivanwu merged 15 commits from dev into main 2026-05-22 09:41:12 +00:00
Conversation 0 Commits 15 Files Changed 185 +15872 -3686
169 changed files with 15872 additions and 3686 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 8f90cf0f0d - Show all commits

74
docs/UNRESOLVED-BUGS.md Normal file
View File

@ -0,0 +1,74 @@
# OCDP 未修复问题清单
**最后更新:** 2026-05-14 (Round 3 回归测试)
**测试覆盖:** 3 轮测试 (Round 1: v1 基线, Round 2: 配额+YAML, Round 3: 回归+新功能)
---
## 已知未修复 Bug (15 个)
### P1 — 高优先级 (1)
| # | 标题 | 严重度 | 描述 | Round |
|----|------|--------|------|-------|
| 1 | Detail API 返回 replicas: 0 | **P1** | `GET /instances/{id}` 始终返回 `replicas: 0`，与 List API 不一致 | R3 NEW |
### P2 — 中优先级 (8)
| # | 标题 | 严重度 | 描述 | Round |
|----|------|--------|------|-------|
| 2 | List API 移除 values 字段 | **P2** | List instances 不再返回 `values`，仅在详情API返回。可能是性能优化，但属于 API 行为变更 | R3 NEW |
| 3 | API 层无配额预检查 | **P2** | 后端接受所有部署请求（返回200），不验证是否超配额。K8s ResourceQuota 在 pod 级阻止，但 Helm release 仍创建 | R2 |
| 4 | Values 冲突时无警告 | **P2** | 同时提供 `values` JSON 和 `valuesYaml` 时，JSON 静默覆盖 YAML，无任何警告 | R2 |
| 5 | Tags 端点缺失 | **P2** | `GET /registries/{id}/repositories/{repo}/tags` 返回 404 | R1 |
| 6 | Metrics API 缺失 | **P2** | `GET /monitoring/clusters/{id}/metrics` 返回 404 | R1 |
| 7 | Stats API 缺失 | **P2** | `GET /clusters/{id}/stats` 返回 404 | R1 |
| 8 | Kubeconfig API 缺失 | **P2** | `GET /clusters/{id}/kubeconfig` 返回 404 | R1 |
| 9 | Namespace 静默覆盖 + HTTP 200 | **P2** | 用户部署到他人的 namespace 时，API 返回 201 但 namespace 被静默改为自己的。应返回 403 | R1 |
### P3 — 低优先级 (6)
| # | 标题 | 严重度 | 描述 | Round |
|----|------|--------|------|-------|
| 10 | 用户枚举漏洞 | **P3** | 不存在用户 "user not found" vs 存在用户 "invalid password"，错误消息不同 | R1 |
| 11 | 无登录速率限制 | **P3** | 10 次快速失败全部返回 401，无 429 或锁定 | R1 |
| 12 | Nginx 版本泄露 | **P3** | `Server: nginx/1.27.5` 响应头暴露精确版本 | R1 |
| 13 | CORS: * | **P3** | `Access-Control-Allow-Origin: *` 允许任意跨域 | R1 |
| 14 | 缺少安全响应头 | **P3** | 无 HSTS, X-Frame-Options, CSP, X-Content-Type-Options | R1 |
| 15 | `/health` 端点返回 SPA HTML | **P3** | 健康检查返回 index.html 而非 `{"status":"ok"}` | R1 |
---
## 已修复 (Round 3 验证通过)
| 原 Bug ID | 描述 | 修复后行为 |
|-----------|------|-----------|
| BUG-001 | Launch 按钮无反应 (P0) | ✅ 部署端到端正常 |
| BUG-002 | SPA 旧路由空白页 (P0) | ✅ 所有旧路由返回 SPA |
| BUG-003 | DELETE 返回 404 (P1) | ✅ 返回 HTTP 204 |
| BUG-004 | DELETE 空响应体 (P1) | ✅ HTTP 204 No Content |
| — | InstanceCard 无 scaling UI | ✅ +/- 按钮 + K8s API |
| — | ModifyModal values 为空 | ✅ Full Helm values + diff |
| — | Per-card Refresh button | ✅ 移除，改为 page-level |
---
## 修复优先级排序
```
立即修复 (P1):
1. Detail API replicas=0 → 从 K8s live state 同步
短期修复 (P2):
2. API 层配额预检查 → POST instances 时验证
3. Values 冲突警告 → 两者同时提供时返回 warning
4. Namespace 拒绝而非覆盖 → 返回 403
5. 缺失端点实现 (tags/stats/metrics/kubeconfig)
安全加固 (P3):
6. 登录错误消息统一 → "Invalid username or password"
7. 速率限制 → max 5/min per IP
8. Nginx: server_tokens off + 安全头
9. CORS 收紧 → 具体域名
10. /health → JSON 响应
```

164
docs/bug-report.md
View File

@ -1,164 +0,0 @@
# OCDP 系统测试 Bug 报告
**测试日期:** 2026-05-11
**测试环境:** http://10.6.80.114:18080
**集群:** k3s (dbf824f1-9962-4d8e-881e-870c75fdb6f5), k8s (23880994-dfe4-48d0-abc0-b49692cc630a)
**Harbor:** harbor.bwgdi.com (83b823af-873b-457c-912c-9ccde3cb12e6)
---
## 测试团队
| Agent | 角色 | 账号 |
|-------|------|------|
| user-a-agent | 前端 UI 测试 | test-user-a / TestUserA123! |
| user-b-agent | API/部署测试 | test-user-b / TestUserB123! |
| user-c-agent | 权限隔离测试 | test-user-c / TestUserC123! |
| security-agent | 安全测试 | admin + 普通用户 |
---
## Bug 列表 (按严重度排序)
### P0 - Blocker (核心功能不可用)
| ID | 标题 | 发现者 | 页面/端点 | 描述 |
|----|------|--------|-----------|------|
| BUG-001 | **Launch 按钮点击无任何反应** | user-a | `/artifact/registries` (TagCard) | Chart Browser 中 TagCard 的 "Launch" 按钮显示为可用状态 (`is_enabled() == True`)，但点击后无任何效果：不弹出 Launch Modal，无 URL 变化，无控制台错误。**核心"一键部署"流程完全阻塞** |
| BUG-002 | **SPA 直接路由返回空白页面** | user-a | `/clusters`, `/registries`, `/monitoring`, `/launch` | 直接访问 SPA 旧路由时只渲染 `<div id="root">` 空壳，React SPA 无法挂载。代码中已定义 redirect 映射但未生效（如 `/clusters` → `/configuration/clusters`） |
### P1 - 高 (High)
| ID | 标题 | 发现者 | 页面/端点 | 描述 |
|----|------|--------|-----------|------|
| BUG-003 | DELETE 实例返回 404 但实际成功删除 | user-b, user-c | `DELETE /clusters/{id}/instances/{id}` | 删除操作正确触发 `pending-delete` 状态转换，但 HTTP 返回 **404**（空 body），非预期 202/204。客户端误判为失败 |
| BUG-004 | DELETE 实例返回空响应体 | user-b | `DELETE /clusters/{id}/instances/{id}` | 用正确的 token 和 ID 请求，返回空 body（无 JSON），前端解析会失败 |
### P2 - 中 (Medium)
| ID | 标题 | 发现者 | 页面/端点 | 描述 |
|----|------|--------|-----------|------|
| BUG-005 | Tags 专用端点缺失 | user-b | `GET /registries/{id}/repositories/{repo}/tags` | 端点未实现，返回纯文本 "404 page not found"。虽可通过 `/artifacts` 获取 tag，但 API 不完整 |
| BUG-006 | 跨用户 namespace 部署时静默覆盖 | user-c | `POST /clusters/{id}/instances` | 用户请求部署到其他用户的 namespace 时，服务端静默使用自己的 namespace，返回 200 且无任何警告或提示 |
| BUG-007 | Clusters Metrics API 缺失 | user-b | `GET /monitoring/clusters/{id}/metrics` | 监控页面可能需要的数据端点未实现（404） |
| BUG-008 | Cluster Stats API 缺失 | user-b | `GET /clusters/{id}/stats` | 统计端点未实现（404） |
| BUG-009 | Kubeconfig API 缺失 | user-b | `GET /clusters/{id}/kubeconfig` | kubeconfig 签发端点未实现（404） |
| BUG-010 | "Launch" 按钮缺乏可访问性标识 | user-a | TagCard "Launch" | Chart 上的 "Launch" 按钮无 `aria-label`，与侧边栏 "Launch Instance" 导航项标签冲突，屏幕阅读器用户无法区分 |
### P3 - 低 (Low)
| ID | 标题 | 发现者 | 页面/端点 | 描述 |
|----|------|--------|-----------|------|
| BUG-011 | API 响应格式不一致 | user-b | 列表 API | Clusters/Registries 返回裸数组，Instances 返回 `{ "instances": [...], "total": N }` 包装对象 |
| BUG-012 | `/auth/me` 返回空的 token 字段 | user-b | `GET /auth/me` | 响应中包含 `"accessToken": ""` 和 `"refreshToken": ""` 空字段，复用了 login 响应 DTO 未清理 |
| BUG-013 | 登录接口存在用户枚举漏洞 | security | `POST /auth/login` | 不存在用户返回 "user not found"，存在用户返回 "invalid password"，攻击者可枚举有效用户名 |
| BUG-014 | 登录接口无速率限制 | security | `POST /auth/login` | 10 次连续请求全部返回 401，无 429 限流或锁定 |
| BUG-015 | Nginx 版本信息泄露 | security | HTTP Headers | `Server: nginx/1.27.5` 暴露精确版本号 |
| BUG-016 | CORS 配置过于宽松 | security | All API | `Access-Control-Allow-Origin: *` 允许任意跨域请求 |
| BUG-017 | 缺少安全响应头 | security | All pages | 缺少 HSTS、X-Frame-Options、Content-Security-Policy 等 |
| BUG-018 | `/health` 端点返回 SPA HTML | security | `GET /health` | 健康检查返回完整 index.html，非 JSON 状态响应 |
---
## 分类汇总
### 前端 Bug
| ID | 描述 | 严重度 |
|----|------|--------|
| BUG-001 | Launch 按钮无反应（核心功能阻塞） | P0 🔴 |
| BUG-002 | SPA 路由空白页 | P0 🔴 |
| BUG-010 | Launch 按钮缺少 aria-label | P2 🟡 |
### 后端 API Bug
| ID | 描述 | 严重度 |
|----|------|--------|
| BUG-003 | DELETE 返回 404 | P1 🟠 |
| BUG-004 | DELETE 空 body | P1 🟠 |
| BUG-005 | Tags 端点缺失 | P2 🟡 |
| BUG-007 | Metrics API 缺失 | P2 🟡 |
| BUG-008 | Stats API 缺失 | P2 🟡 |
| BUG-009 | Kubeconfig API 缺失 | P2 🟡 |
| BUG-011 | 响应格式不一致 | P3 🔵 |
| BUG-012 | auth/me 空 token 字段 | P3 🔵 |
| BUG-018 | /health 返回 HTML | P3 🔵 |
### 安全/权限 Bug
| ID | 描述 | 严重度 |
|----|------|--------|
| BUG-006 | Namespace 静默覆盖（安全但令人困惑） | P2 🟡 |
| BUG-013 | 用户枚举（错误消息差异） | P3 🔵 |
| BUG-014 | 无速率限制 | P3 🔵 |
| BUG-015 | Nginx 版本泄露 | P3 🔵 |
| BUG-016 | CORS Origin: * | P3 🔵 |
| BUG-017 | 缺少安全响应头 | P3 🔵 |
### 严重度分布
| 级别 | 数量 |
|------|------|
| P0 (Blocker) | 2 |
| P1 (High) | 2 |
| P2 (Medium) | 6 |
| P3 (Low) | 8 |
| **合计** | **18** |
---
## 测试通过项
### 认证
- [x] 有效凭据登录 (admin + 所有 test-user)
- [x] 无效凭据返回 401
- [x] 无 token 访问被保护 API 返回 401
- [x] 无效/篡改 JWT token 全部被拒绝
- [x] /auth/me 返回正确的用户信息
- [x] JWT payload 包含角色、权限、namespace
### Cluster / Registry API
- [x] 集群列表正常返回
- [x] 集群健康检查正常
- [x] Registry 列表正常返回
- [x] 通过 artifacts 端点浏览 repository 正常
- [x] 无效 registry/repository 返回恰当错误
### 权限隔离
- [x] GET /users 返回 403 (普通用户)
- [x] POST /auth/register 返回 403 (普通用户)
- [x] 用户无法访问其他用户的 workspace 资源
- [x] 用户无法部署到其他用户的 Kubernetes namespace
- [x] 安全架构：核心认证/授权/脱敏/隔离控制均正确实现
### 实例部署生命周期
- [x] 实例创建操作成功（pending-install）
- [x] 实例状态正确追踪（pending-install → deployed）
- [x] 实例删除正确转换状态（pending-delete → 消失）
- [x] 实例列表按 clusterId 正确过滤
### 安全测试通过项
- [x] XSS/SQLi 注入安全处理
- [x] 路径遍历攻击被阻止
- [x] JWT alg=none/无效格式被拒绝
- [x] 集群凭据和 Registry 密码脱敏显示 (••••••••)
- [x] 自注册端点需认证 (401)
---
## 建议修复优先级
### 立即修复 (P0)
1. **BUG-001**: 调查 Launch 按钮 onClick handler — TagCard 组件中 `onLaunch` prop 未正确传递给 LaunchModal，或 launch 状态 / artifactType 检查阻止了 modal 打开
2. **BUG-002**: 检查 React Router `<Navigate redirect>` 组件和 SPA 的 index.html 配置，确保旧路由正确重定向
### 尽快修复 (P1)
3. **BUG-003/004**: InstanceHandler.Delete 应返回 202 Accepted + `{"status":"deleting"}` 而非 404+空 body
### 短期修复 (P2)
4. 实现 `/metrics`, `/stats` 等缺失 API
5. Launch 按钮添加 `aria-label` 属性
6. Namespace 覆盖时返回警告或 403
### 安全加固 (P3)
7. 登录错误消息统一为 "Invalid username or password"
8. 实现速率限制
9. Nginx 安全加固：`server_tokens off` + 安全响应头
10. CORS 收紧为具体域名
11. 修复 `/health` 端点
12. 统一 API 响应格式

92
docs/bugs-user-a.md
View File

@ -1,92 +0,0 @@
# OCDP Platform QA Report - test-user-a
**Date:** 2026-05-11
**Environment:** http://10.6.80.114:18080
**User:** test-user-a (non-admin)
## Summary
- **Total Bugs Found:** 3
- **Screenshots Taken:** 12
- **Test Status:** 7/8 areas covered, 1 blocked (Launch button non-functional)
---
## Bug List
### Bug #1: Direct SPA Routes Return Empty Pages (🔴 HIGH)
- **Page:** Multiple — `/clusters`, `/registries`, `/monitoring`, `/launch`
- **Action:** Navigate directly to these URLs
- **Actual:** Returns only the React `<div id="root">` shell with no rendered content (~0 chars body text). The SPA fails to mount when hitting these routes directly.
- **Expected:** Should either render content or redirect to correct working routes:
- `/clusters` → `/configuration/clusters`
- `/registries` → `/configuration/registries`
- `/monitoring` → `/monitoring/clusters`
- `/launch` → `/artifact/registries`
- **Severity:** HIGH — Users who bookmark or type these URLs see blank pages
- **Screenshot:** `01-login` (representative of empty state)
**Working routes for reference:**
- `/configuration/clusters` ✅
- `/configuration/registries` ✅
- `/monitoring/clusters` ✅
- `/artifact/registries` ✅
- `/artifact/instances` ✅
---
### Bug #2: Launch Button Does Nothing When Clicked (🔴 HIGH)
- **Page:** Chart Browser (`/artifact/registries`)
- **Action:**
1. Navigate to `/artifact/registries`
2. Registry `harbor-bwgdi` loads with 13 charts
3. Expand `charts/chromadb` folder
4. Tag `0.1.4` appears with "Launch" and "Copy" buttons
5. Click the "Launch" button
- **Actual:** No visible reaction — no modal opens, no URL change, no console error. The button is not disabled (no `disabled` attribute, no `aria-disabled`), is visibly styled as active (`bg-blue-50 text-blue-700 border-blue-200 shadow-sm`), and Playwright confirms `is_enabled() == True`. The React onClick handler produces no observable effect.
- **Expected:** Clicking "Launch" on a chart tag should open a deployment form/dialog with cluster selector, instance name, namespace, and values configuration fields.
- **Severity:** HIGH — Core platform feature (deploying Helm charts) is completely blocked
- **Screenshot:** `04-chart-expanded`
---
### Bug #3: Ambiguous "Launch" Button Labels (🟡 MEDIUM)
- **Page:** Chart Browser (`/artifact/registries`)
- **Action:** Inspect button accessible names
- **Actual:** Both the sidebar navigation item "Launch Instance" and the chart action button "Launch" appear on the same page. The chart action button has no distinguishing `aria-label` or accessible description. The "Copy" button next to it has a `title="Copy pull command"` attribute, but "Launch" does not.
- **Expected:** The chart action should have a descriptive label like `aria-label="Launch chart chromadb version 0.1.4"` to differentiate from the nav item.
- **Severity:** MEDIUM — Accessibility concern; minor confusion for sighted users with multiple "Launch" targets
---
## Test Results by Area
| Area | Status | Notes |
|------|--------|-------|
| Login | ✅ PASS | test-user-a login successful, redirect to `/home` |
| Home Page | ✅ PASS | All cards visible, nav clicks work, no Users section |
| Sidebar Nav | ✅ PASS | All 6 items navigate correctly, Users hidden |
| Chart Browser | ❌ BLOCKED | Registry loads, charts expand, but **Launch button dead** |
| Instances | ✅ PASS | Empty state, filter, refresh all work |
| Monitoring | ✅ PASS | 2 clusters, health data, CPU/Memory/GPU stats all load |
| Config - Clusters | ✅ PASS | Both clusters listed, Add form opens |
| Config - Registries | ✅ PASS | Harbor registry listed, Add form opens |
| Direct Routes | ❌ FAIL | 4 routes return empty pages |
## Screenshots
- `01-login` → `/tmp/ocdp-qa-screenshots/01-login.png`
- `02-home` → `/tmp/ocdp-qa-screenshots/02-home.png`
- `02-home-full` → `/tmp/ocdp-qa-screenshots/02-home-full.png`
- `04-chart-browser` → `/tmp/ocdp-qa-screenshots/04-chart-browser.png`
- `04-chart-expanded` → `/tmp/ocdp-qa-screenshots/04-chart-expanded.png`
- `04-launch-modal` → `/tmp/ocdp-qa-screenshots/04-launch-modal.png`
- `05-instances` → `/tmp/ocdp-qa-screenshots/05-instances.png`
- `06-monitoring` → `/tmp/ocdp-qa-screenshots/06-monitoring.png`
- `07-clusters` → `/tmp/ocdp-qa-screenshots/07-clusters.png`
- `07-add-cluster-form` → `/tmp/ocdp-qa-screenshots/07-add-cluster-form.png`
- `08-registries` → `/tmp/ocdp-qa-screenshots/08-registries.png`
- `08-add-registry-form` → `/tmp/ocdp-qa-screenshots/08-add-registry-form.png`

149
docs/bugs-user-b.md
View File

@ -1,149 +0,0 @@
# Bug Report: test-user-b QA Test
**Tester:** test-user-b (user role)
**Date:** 2026-05-11
**Environment:** http://10.6.80.114:18080
---
## Bug 1: Repository Tags Endpoint Returns 404
**Endpoint:** `GET /api/v1/registries/{registryId}/repositories/{repository}/tags`
**Status Code:** 404
**Response Body:** `404 page not found` (plain text, not JSON)
**Expected:** Should return a list of tags for the chart/artifact.
**Actual:** The dedicated tags endpoint is not implemented or routes incorrectly. The artifacts endpoint (`/repositories/{repository}/artifacts`) does work and returns tag info.
**Severity:** Medium — tags are still discoverable via artifacts endpoint but the dedicated tags API is broken.
---
## Bug 2: DELETE Instance Returns Empty Response Body
**Endpoint:** `DELETE /api/v1/clusters/{clusterId}/instances/{instanceId}`
**Status Code:** 200
**Response Body:** (empty — no content at all)
**Expected:** Should return a confirmation JSON body (e.g., `{"message": "Instance deletion initiated", "id": "..."}`) or at minimum a 202 Accepted with status details.
**Actual:** Returns a completely empty body. The instance does transition to `pending-delete` state, but the API consumer receives no feedback.
**Severity:** Medium — operation works but API consumer gets no confirmation.
---
## Bug 3: Cluster Stats Endpoint Returns 404
**Endpoint:** `GET /api/v1/clusters/{clusterId}/stats`
**Status Code:** 404
**Response Body:** `404 page not found` (plain text)
**Expected:** Should return cluster resource statistics (CPU, memory, pod counts, etc.) or a proper JSON error if not implemented.
**Actual:** Endpoint is not implemented — returns a raw 404 with no JSON error structure.
**Severity:** Low — but given the user has `monitoring:clusters:view` permission, this is a missing feature.
---
## Bug 4: Kubeconfig Endpoint Returns 404
**Endpoint:** `GET /api/v1/clusters/{clusterId}/kubeconfig`
**Status Code:** 404
**Response Body:** `404 page not found` (plain text)
**Expected:** Should return a kubeconfig file content or JSON error. User has `kubeconfig:issue:own` permission.
**Actual:** Endpoint is not implemented.
**Severity:** Low — the permission exists but the endpoint does nothing.
---
## Bug 5: Monitoring Metrics Endpoint Returns 404
**Endpoint:** `GET /api/v1/monitoring/clusters/{clusterId}/metrics`
**Status Code:** 404
**Response Body:** `404 page not found` (plain text)
**Expected:** Monitoring metrics data. User has `monitoring:clusters:view` permission.
**Actual:** Endpoint not found.
**Severity:** Low — monitoring permissions exist but backend endpoints missing.
---
## Bug 6: Inconsistent API Response Format (Array vs Object Wrapper)
**Clusters and Registries** return bare arrays:
```json
[
{ "id": "...", "name": "k3s", ... }
]
```
**Instances** returns an object wrapper:
```json
{
"instances": [
{ "id": "...", "name": "test-nginx-b", ... }
],
"total": 1
}
```
**Expected:** Consistent response format across all list endpoints. Either all return bare arrays or all use the `{ "items": [...], "total": N }` wrapper pattern.
**Severity:** Low — API consistency issue. Makes client code harder to write generically.
---
## Bug 7: auth/me Returns Empty Token Fields
**Endpoint:** `GET /api/v1/auth/me`
**Response includes empty/unpopulated fields:**
```json
{
"accessToken": "",
"refreshToken": "",
...
}
```
**Expected:** Either remove these fields from the `/auth/me` response (they are only meaningful in login/refresh responses) or populate them with valid values.
**Actual:** Emptry string values for both token fields create confusion about whether they should be present.
**Severity:** Low — cosmetic issue, but suggests the DTO is reusing the login response struct without clearing token fields.
---
## Summary
| # | Bug | Severity | Category |
|---|-----|----------|----------|
| 1 | Tags endpoint 404 | Medium | Missing Implementation |
| 2 | DELETE returns empty body | Medium | API Response Quality |
| 3 | Cluster stats endpoint 404 | Low | Missing Implementation |
| 4 | Kubeconfig endpoint 404 | Low | Missing Implementation |
| 5 | Monitoring metrics endpoint 404 | Low | Missing Implementation |
| 6 | Inconsistent list response format | Low | API Consistency |
| 7 | auth/me returns empty tokens | Low | API Response Quality |
**Passed Tests:**
- Login/authentication ✓
- Auth/me user info ✓
- Cluster listing ✓
- Cluster health check ✓
- Registry listing ✓
- Repository browsing (artifacts) ✓
- Instance deployment (nginx chart) ✓
- Instance status tracking (pending-install → deployed) ✓
- Instance deletion (async, transitions to pending-delete then removed) ✓
- Error handling for invalid repository ✓
- Error handling for missing required fields ✓
- Auth rejects invalid tokens ✓
- Auth rejects missing tokens ✓
- Instance cleanup confirmed ✓

109
docs/bugs-user-c.md
View File

@ -1,109 +0,0 @@
# QA Report: Permission Isolation & Multi-Tenancy Testing — test-user-c
**Tester:** test-user-c (role: `user`)
**Date:** 2026-05-11
**Environment:** http://10.6.80.114:18080
## Summary
Test-user-c is a standard `user` role with namespace `ocdp-u-test-c`, workspace `71459030-7166-4c79-b53c-81c61da4c313`. Permissions follow the `manage_own` / `view` pattern — no admin-level permissions.
---
## Test Results
### 1. Login & Basic Access ✅
| Test | Result | Notes |
|------|--------|-------|
| POST /auth/login | ✅ Pass | Token issued, role=`user`, workspace/namespace correctly assigned |
| GET /auth/me | ✅ Pass | Returns correct user profile with permissions |
| GET /clusters | ✅ Pass | Sees all `global_shared` clusters (k8s, k3s) |
| GET /registries | ✅ Pass | Sees all `global_shared` registries (harbor) |
### 2. Admin Endpoint Protection
| Test | Result | Notes |
|------|--------|-------|
| GET /api/v1/users | ✅ **403 Forbidden** | Properly blocked — `permission denied` |
| POST /auth/register | ✅ **403 Forbidden** | Cannot register new users as non-admin |
| GET /api/v1/admin/* | ✅ **404** | Admin route prefix doesn't exist (not a bypass risk) |
### 3. Frontend Access
| Test | Result | Notes |
|------|--------|-------|
| GET /configuration/users | ⚠️ **200 (OK)** | SPA returns index.html — expected. Auth is enforced via API, not routes. |
| GET /configuration/clusters | ⚠️ **200 (OK)** | Same — SPA behavior. |
| GET /configuration/registries | ⚠️ **200 (OK)** | Same. |
**Risk: Low.** This is standard SPA behavior. Authorization is enforced at the API level. However, if the frontend relies solely on hiding UI elements rather than checking permissions, users who manually navigate could see empty/error states.
### 4. Namespace Isolation Enforcement
| Test | Result | Notes |
|------|--------|-------|
| Deploy with `namespace: ocdp-u-test-a` | ⚠️ **Silently overridden** | Server ignored requested namespace and used `ocdp-u-test-c` instead. **No warning or error returned.** |
| PATCH to change namespace | ✅ **404** | PATCH endpoint doesn't exist — namespace cannot be changed after creation |
🔴 **Bug: Silent namespace override (Low severity)**
When a user specifies a namespace that doesn't belong to them in the instance creation request, the server silently overrides it with the user's own namespace. This is secure (prevents cross-namespace deployment) but:
- The user receives HTTP 200 with the overridden value — no indication that their request was modified
- The response does not differentiate between "user's own namespace" and "requested namespace"
- This could lead to user confusion about where their resources were actually deployed
- It's unclear whether the user's Helm values also get silently overridden (e.g., the `values.namespace` field)
### 5. Resource Isolation
| Test | Result | Notes |
|------|--------|-------|
| GET instances with other workspaceId query param | ✅ **Isolated** | Returns only own instances (workspaceId filter is server-enforced) |
| DELETE on own instance | ⚠️ **Async deletion** | Returns HTTP 404 on DELETE itself, but instance transitions to `pending-delete` then disappears |
🔴 **Bug: DELETE returns 404 on success (Medium severity)**
When deleting an instance via `DELETE /clusters/{clusterId}/instances/{instanceId}`:
- The instance transitions to `pending-delete` status
- But the HTTP response status code is **404** rather than 200/202/204
- The first raw DELETE call returns an empty body (causing JSON parse errors)
- This is an API inconsistency — async deletions should return HTTP 202 Accepted
### 6. Monitoring & Other Endpoints
| Test | Result | Notes |
|------|--------|-------|
| GET /monitoring/clusters/.../pods | ✅ **404** | Monitoring endpoints not implemented for this cluster type |
| POST /kubeconfig | ✅ **404** | Kubeconfig endpoint not implemented |
These endpoints return 404 which is acceptable behavior for features not yet implemented.
---
## Security Assessment
### Works as Intended ✅
- Admin endpoints (`/users`, `/auth/register`) properly return 403
- User cannot access other users' instances via workspaceId manipulation
- User cannot deploy into other users' Kubernetes namespaces
- No PATCH/PUT verbs available to modify existing instance namespaces
- No admin-specific route paths leak data
### Bugs Found
1. **DELETE returns 404 on successful async deletion** (Medium)
- Endpoint: `DELETE /clusters/{id}/instances/{id}`
- After call, instance status becomes `pending-delete` and eventually disappears
- But the HTTP response is `404` with empty body
- Expected: `202 Accepted` with a `status: "deleted"` or similar response
- Risk: Clients interpreting HTTP 404 as "not found" will retry or report errors incorrectly
2. **Silent namespace override without user feedback** (Low)
- Endpoint: `POST /clusters/{id}/instances`
- When requesting deployment into another user's namespace, the server silently uses the caller's namespace
- No warning, no error, no indication in the response
- Expected: Either `403 Forbidden` with "cannot deploy into namespace owned by another user" or a response field indicating the override occurred
- Risk: Low for security (the override correctly prevents cross-tenant deployment), but could cause user confusion
### No Critical Vulnerabilities Found
- No privilege escalation vectors identified
- No data leakage across workspaces
- No ability to access or manipulate other users' resources

209
docs/regression-full-report.md Normal file
View File

@ -0,0 +1,209 @@
# OCDP 回归测试完整报告 (Round 3)
**测试日期:** 2026-05-14
**环境:** http://10.6.80.114:18080
**集群:** k3s v1.28.0 (dbf824f1-9962-4d8e-881e-870c75fdb6f5), k8s (23880994-dfe4-48d0-abc0-b49692cc630a)
**Harbor:** harbor.bwgdi.com (83b823af-873b-457c-912c-9ccde3cb12e6)
---
## 代码变更概要 (Commit b88fe24 +)
| Commit | 变更 |
|--------|------|
| b88fe24 | fix: real K8s replicas in list API, full Helm values in modify YAML editor |
| 96d42ee | fix: scale replicas in response, YAML lineWidth, delta values, modified keys |
| 4441f58 | fix: direct K8s scaling, replicas from K8s API, button labels, modify fetch |
| 49b92e6 | fix: UI redesign — horizontal instance rows, proper scaling, readable tag cards |
| 28ecb2e | feat: scale instances, --reuse-values, values diff, UI redesign, hover animations |
| 87eaaa5 | fix: remove per-card Refresh button, consolidate to page-level refresh |
---
## 测试结果总览
### 总评分: 85/100
| 测试领域 | 状态 | 问题数 |
|----------|------|--------|
| 前端 UI / Launch / Routes | ✅ ALL PASS | 0 |
| API CRUD / 部署生命周期 | ⚠️ 2 issues | 2 |
| 配额执行 | ⚠️ 1 issue (pre-existing) | 1 |
| Values 优先级 / gpuMem | ⚠️ 1 issue | 1 |
| 权限隔离 | ✅ ALL PASS | 0 |
| 安全测试 | ⚠️ 3 issues (pre-existing) | 3 |
| DELETE 行为 | ✅ FIXED | 0 |
| InstanceCard / Scaling UI | ✅ ALL PASS | 0 |
| ModifyModal / Values Diff | ✅ ALL PASS | 0 |
---
## 新发现 Bug
### 🆕 BUG-R3-001: Detail API replicas=0 与 List API 不一致 (High)
| 属性 | 值 |
|------|-----|
| 严重度 | **P1 (High)** |
| 端点 | `GET /clusters/{id}/instances/{instance_id}` |
| 现象 | List API 返回正确 replicas（如 1, 5），但 Detail API 始终返回 `replicas: 0` |
| 根因 | Detail endpoint 从数据库读取实例记录，replicas 字段未同步自 K8s 实时状态 |
| 影响 | 前端依赖 Detail API 的页面（如刷新后详情页）显示错误的副本数 |
| 修复建议 | Detail endpoint 也从 K8s live state 填充 replicas，或确保数据库同步 |
### 🆕 BUG-R3-002: List API 移除 values 字段
| 属性 | 值 |
|------|-----|
| 严重度 | **P2 (Medium)** |
| 端点 | `GET /clusters/{id}/instances` |
| 现象 | List API 响应不再包含 `values` 字段（之前版本有）；values 仅在单实例 GET 中返回 |
| 影响 | 依赖 list API values 的测试脚本和前端组件会 break |
| 备注 | 可能是故意的性能优化，但属于 API 行为变更 |
---
## 已修复 Bug (验证通过)
| Bug ID | 描述 | 之前状态 | 现在状态 | 验证 |
|--------|------|----------|----------|------|
| BUG-001 | Launch 按钮无反应 | P0 Blocker | ✅ FIXED | 部署端到端成功 |
| BUG-002 | SPA 旧路由空白页 | P0 Blocker | ✅ FIXED | 所有旧路由返回 SPA |
| BUG-003 | DELETE 返回 404 | P1 High | ✅ FIXED | 返回 HTTP 204 |
| BUG-004 | DELETE 空 body | P1 High | ✅ FIXED | HTTP 204 No Content |
| - | InstanceCard 无 scaling UI | New Feature | ✅ ADDED | +/- 按钮 + API 调用 |
| - | ModifyModal values 为空 | Bug | ✅ FIXED | Full Helm values 返回 |
---
## 仍未修复的已知问题
| Bug ID | 描述 | 严重度 | 状态 |
|--------|------|--------|------|
| BUG-013 | 用户枚举（login 错误消息差异） | Medium | 未修复 |
| BUG-014 | 无速率限制 | Medium | 未修复 |
| BUG-015 | Nginx 版本泄露 | Low | 未修复 |
| BUG-016 | CORS: * | Low | 未修复 |
| BUG-017 | 缺少安全响应头 | Low | 未修复 |
| BUG-018 | /health 返回 HTML | Low | 未修复 |
| BUG-005 | Tags 端点 404 | Medium | 未修复 |
| BUG-006 | Namespace 静默覆盖无警告 | Medium | 未修复 |
| BUG-007-009 | Metrics/Stats/Kubeconfig 端点缺失 | Medium | 未修复 |
| BUG-011 | API 响应格式不一致 | Low | 未修复 |
| BUG-012 | /auth/me 空 token 字段 | Low | 未修复 |
| - | API 层配额预检查 | New Feature | 未实现 |
| - | Values 冲突时无警告 | UX | 未修复 |
---
## 详细测试结果
### 1. 前端 UI (test-user-a)
| 测试项 | 结果 |
|--------|------|
| 登录页加载 | ✅ HTTP 200, SPA 398 bytes |
| JWT 认证 | ✅ role=user, 10 permissions, quota 字段 |
| 所有页面路由 | ✅ 6 个路由全部返回 398 bytes (非空白) |
| 旧路由重定向 | ✅ /clusters, /registries, /launch, /monitoring 全部返回 SPA |
| Chart Browser | ✅ nginx:22.1.1 可发现，Launch 按钮可用 |
| 部署 Pipeline | ✅ pending-install → deployed (~15s) |
| InstanceCard Scale UI | ✅ +/- 按钮，副本数显示 |
| InstanceCard Actions | ✅ Entries/Diag/Modify/Delete 全部可见 |
| ModifyModal YAML Editor | ✅ full Helm values, lineWidth:0, diff 检测 |
| TagCard UI | ✅ 色标 (chart=blue, image=green), Copy helm pull 命令, LATEST badge |
| DELETE | ✅ HTTP 204 |
### 2. API 后端 (test-user-b)
| 测试项 | 结果 |
|--------|------|
| Login + /auth/me | ✅ 完整 profile (quota, namespace, permissions) |
| Cluster list | ✅ 2 clusters |
| Registry list | ✅ 1 registry (harbor-bwgdi) |
| Repository artifacts | ✅ Harbor API proxy 正常 |
| Cluster health | ✅ k3s healthy, v1.28.0 |
| 部署 nginx (default) | ✅ deployed, replicas=1 |
| 部署 nginx (over-quota) | ⚠️ 接受部署 (API 无预检查) |
| 实例状态轮询 | ✅ ~20s 到达 deployed |
| List API replicas | ✅ 正确显示 1/5 |
| Detail API replicas | ❌ 返回 0 (BUG-R3-001) |
| 实例删除 | ✅ HTTP 204 (BUG-003 FIXED) |
| 缺失端点 | ✅ 正常 404 |
### 3. Values 优先级 (test-user-c)
| 方法 | 结果 |
|------|------|
| values JSON only | ✅ gpuMem=10000 正确接受 |
| valuesYaml only | ✅ YAML -> JSON 解析正确 |
| 同时提供 (冲突) | ✅ values JSON 覆盖 valuesYaml |
| 冲突警告 | ❌ 无警告（建议添加） |
| 默认值 (空 values) | ✅ 使用 chart 内置 defaults |
| gpuMem=10000 | ✅ 整数 MB scalar 正确 |
| 清理 | ✅ 所有 3 个实例已删除 |
### 4. 权限隔离
| 测试项 | 结果 |
|--------|------|
| Admin lists users | ✅ 10 users |
| test-admin-d lists users | ✅ 10 users (admin role works) |
| test-user-c GET /users | ✅ 403 Forbidden |
| test-user-c POST /auth/register | ✅ 403 Forbidden |
| Cross-tenant deploy (c→b) | ✅ Silent override to ocdp-u-test-c (secure) |
| Deploy into own namespace | ✅ 成功 |
| Instance visibility | ✅ 仅看到自己的实例 |
| Disable user | ✅ 用户被禁用 |
| Disabled user login | ✅ 401 |
| Re-enable user | ✅ 恢复 |
| Self-registration | ✅ 401 (必须认证) |
### 5. 安全 (回归)
| 测试项 | 结果 |
|--------|------|
| 未认证端点 | ✅ 全部 401 |
| JWT 验证 | ✅ 篡改 token 被拒绝 |
| XSS/SQLi | ✅ 安全处理 |
| 敏感数据脱敏 | ✅ creds 显示为 •••••••• |
| 普通用户权限提升 | ✅ 403 blocked |
| 用户枚举 | ❌ 不同错误消息 (未修复) |
| 速率限制 | ❌ 无 429 (未修复) |
| CORS * + 缺失安全头 | ❌ 未修复 |
| Nginx 版本泄露 | ❌ "nginx/1.27.5" (未修复) |
| /auth/me 空 token | ✅ tokens 已正确清空 |
---
## 优先修复清单
### 立即修复 (P0/P1)
1. **BUG-R3-001**: Detail API replicas=0 — 从 K8s live state 填充
2. **配额预检查** — POST instances 时验证请求资源 ≤ 用户配额
### 尽快修复 (P2)
3. Values 冲突时添加 API 警告
4. Tags/Metrics/Stats/Kubeconfig 端点实现
5. Namespace 静默覆盖返回警告
6. List API values 字段回归或文档化
### 安全加固 (P3)
7. 登录错误消息统一
8. 速率限制
9. Nginx 安全头 + 关闭 server_tokens
10. CORS 收紧
---
## 对比: Round 1 vs Round 3
| 指标 | Round 1 (2026-05-11) | Round 3 (2026-05-14) |
|------|---------------------|---------------------|
| Total Bugs | 18 | 15 (3 fixed, 12 remain, 2 new) |
| P0 Blockers | 2 (Launch, Routes) | 0 |
| P1 High | 2 (DELETE 404, empty body) | 1 (Detail replicas=0) |
| 新功能 | - | Scaling UI, Values Diff, YAML editor |
| 安全漏洞 | 6 | 5 (token fields fixed) |
**结论:** 代码修改有效，3 个关键 Bug 已修复，新增了 scaling 和 values diff 功能。仍有 12 个已知问题和 2 个新问题待修复。

284
docs/security/bugs-security.md
View File

@ -1,284 +0,0 @@
# OCDP Security Audit Report
**Date:** 2026-05-11
**Target:** http://10.6.80.114:18080
**API Base:** http://10.6.80.114:18080/api/v1
---
## Finding 1: User Enumeration via Login Error Messages
| Field | Value |
|-------|-------|
| **Test** | Authentication Error Disclosure |
| **Severity** | **Medium** |
| **Endpoint** | `POST /api/v1/auth/login` |
| **Status** | Confirmed |
### What I Did
```bash
# Non-existent user
curl -s -X POST http://10.6.80.114:18080/api/v1/auth/login \
-H "Content-Type: application/json" \
-d '{"username":"nonexistent_user_xyz","password":"test123"}'
# Existing user with wrong password
curl -s -X POST http://10.6.80.114:18080/api/v1/auth/login \
-H "Content-Type: application/json" \
-d '{"username":"admin","password":"wrongpassword"}'
```
### Expected
Both requests should return the same generic error message (e.g., "Invalid credentials") to prevent username enumeration.
### Actual
- Non-existent user: `{"error":"Login failed","message":"user not found","code":401}`
- Existing user: `{"error":"Login failed","message":"invalid password","code":401}`
The error messages are different, allowing an attacker to determine whether a username exists in the system.
### Impact
An attacker can enumerate valid usernames by observing the error message difference. This is the first step in a targeted brute force or credential stuffing attack.
### Recommendation
Return identical error messages for both cases, e.g., `"Invalid username or password"`.
---
## Finding 2: No Rate Limiting on Login Endpoint
| Field | Value |
|-------|-------|
| **Test** | Brute Force Protection |
| **Severity** | **Medium** |
| **Endpoint** | `POST /api/v1/auth/login` |
| **Status** | Confirmed |
### What I Did
```bash
for i in $(seq 1 10); do
curl -s -o /dev/null -w "%{http_code}" \
-X POST http://10.6.80.114:18080/api/v1/auth/login \
-H "Content-Type: application/json" \
-d '{"username":"admin","password":"wrongpassword"}'
done
```
### Expected
After a threshold (e.g., 5 failed attempts), the server should return HTTP 429 Too Many Requests or temporarily lock the account.
### Actual
All 10 rapid sequential attempts returned HTTP 401. No rate limiting, no account lockout, no progressive delay.
### Impact
An attacker can brute force passwords without restriction. Combined with Finding 1 (user enumeration), the attack surface is increased.
### Recommendation
- Implement rate limiting on the login endpoint (e.g., max 5 attempts per minute per IP).
- Consider account lockout after N failed attempts.
- Add progressive response delays after repeated failures.
---
## Finding 3: Server Version Disclosure
| Field | Value |
|-------|-------|
| **Test** | Information Disclosure |
| **Severity** | **Low** |
| **Endpoint** | All (HTTP response headers) |
| **Status** | Confirmed |
### What I Did
```bash
curl -s -D - http://10.6.80.114:18080/ | head -10
```
### Expected
Server header should be generic (e.g., `Server: nginx`) or removed entirely.
### Actual
```http
Server: nginx/1.27.5
```
### Impact
Knowing the exact nginx version helps attackers target known vulnerabilities for that specific version.
### Recommendation
Disable or obfuscate the Server header in nginx configuration:
```nginx
server_tokens off;
```
---
## Finding 4: Permissive CORS Policy
| Field | Value |
|-------|-------|
| **Test** | CORS Misconfiguration |
| **Severity** | **Low** |
| **Endpoint** | All API endpoints |
| **Status** | Confirmed |
### What I Did
```bash
curl -s -D - http://10.6.80.114:18080/api/v1/auth/login \
-X POST -H "Content-Type: application/json" \
-d '{"username":"test","password":"test"}'
```
### Expected
CORS `Access-Control-Allow-Origin` should be restricted to the application's origin (e.g., `http://10.6.80.114:18080`) rather than allowing all origins.
### Actual
```http
Access-Control-Allow-Origin: *
Access-Control-Allow-Credentials: true
Access-Control-Allow-Methods: GET, POST, PUT, DELETE, OPTIONS
Access-Control-Allow-Headers: Content-Type, Authorization, X-Requested-With
Access-Control-Max-Age: 86400
```
### Impact
Any website can make cross-origin requests to the API. If a user is logged in, a malicious site could potentially make authenticated API calls on their behalf (CSRF-style attack, though mitigated by the Bearer token requirement).
### Recommendation
Restrict `Access-Control-Allow-Origin` to the specific frontend origin(s) instead of `*`.
---
## Finding 5: Missing Security Headers
| Field | Value |
|-------|-------|
| **Test** | Security Headers Audit |
| **Severity** | **Low** |
| **Endpoint** | All |
| **Status** | Confirmed |
### What I Did
```bash
curl -s -D - http://10.6.80.114:18080/ | head -20
```
### Expected
Security headers should include:
- `Strict-Transport-Security`
- `X-Content-Type-Options: nosniff`
- `X-Frame-Options: DENY`
- `Content-Security-Policy`
### Actual
None of these security headers are present in responses.
### Impact
Increases attack surface for clickjacking, MIME-type confusion, and XSS attacks.
### Recommendation
Add the following headers to nginx configuration:
```
add_header X-Frame-Options "DENY" always;
add_header X-Content-Type-Options "nosniff" always;
add_header X-XSS-Protection "0" always;
add_header Strict-Transport-Security "max-age=31536000; includeSubDomains" always;
add_header Content-Security-Policy "default-src 'self'; script-src 'self'; style-src 'self' 'unsafe-inline';" always;
```
---
## Finding 6: `/health` Endpoint Returns HTML Instead of Health Status
| Field | Value |
|-------|-------|
| **Test** | Health Endpoint Behavior |
| **Severity** | **Low** |
| **Endpoint** | `GET /health` |
| **Status** | Confirmed |
### What I Did
```bash
curl -s http://10.6.80.114:18080/health
```
### Expected
A health check endpoint should return a structured JSON response (e.g., `{"status":"healthy"}`) with HTTP 200.
### Actual
Returns the full `index.html` SPA page with HTTP 200:
```html
<!doctype html>
<html lang="en">
<head>
<title>OCDP Platform</title>
...
```
### Impact
Not a direct vulnerability, but misconfigured health checks can cause false positives in monitoring/load balancer health checks. It also means the SPA is served at `/health`, which is unexpected.
### Recommendation
Implement a dedicated health endpoint that returns `{"status":"ok"}` with appropriate content type, or remove the `/health` route if not needed.
---
## Tests Passed (No Issues Found)
| Test | Result |
|------|--------|
| **1. Unauthenticated Access** | **PASS** - All business endpoints return 401 |
| **2. JWT Token Manipulation** | **PASS** - Tampered tokens, alg=none, invalid formats all rejected (401) |
| **3. XSS/SQLi Testing** | **PASS** - Script injection, SQLi patterns safely handled |
| **4. IDOR - Instance Access** | **PASS** - No instances deployed to test; cluster/registry isolation confirmed working |
| **5. Sensitive Data Masking** | **PASS** - Cluster certs/keys and registry passwords masked as `••••••••` |
| **6. Self-Registration** | **PASS** - Registration endpoint requires authentication (401) |
| **7. Path Traversal** | **PASS** - Path traversal attempts return index.html (not /etc/passwd) |
| **8. Admin Permission Escalation** | **PASS** - Regular users blocked from admin endpoints (403) |
---
## Summary
| Severity | Count | Findings |
|----------|-------|----------|
| Critical | 0 | — |
| High | 0 | — |
| **Medium** | **2** | User enumeration, No rate limiting |
| **Low** | **4** | Server version disclosure, Permissive CORS, Missing security headers, `/health` returns HTML |
| **Total** | **6** | |
The platform's core security controls (authentication, JWT validation, authorization, sensitive data masking) are properly implemented. The main areas for improvement are authentication hardening (rate limiting, user enumeration) and HTTP security hardening (headers, CORS).

156
docs/test2-quota.md
View File

@ -1,156 +0,0 @@
# Resource Quota Enforcement Test Report
**Date:** 2026-05-11
**Tester:** test-user-b
**Namespace:** ocdp-u-test-b
**User Quota:** cpu=2, memory=4Gi, gpu=0, gpumem=0
---
## Test Summary
| Test | Description | Expected | Actual | Result |
|------|-------------|----------|--------|--------|
| A | Deploy nginx (default, within quota) | Success | Deployed (status: `deployed`) | ✅ PASS |
| B | Deploy nginx (cpu=4, mem=8Gi, replicas=5, exceeds quota) | Blocked by quota | Helm release created, Service created, all pods blocked by ResourceQuota (status: `pending-install`) | ⚠️ PARTIAL |
| C | Deploy vllm-serve with gpu=1 (gpu quota = 0) | Blocked by quota | Helm release created, all pods blocked by ResourceQuota (status: `pending-install`) | ⚠️ PARTIAL |
---
## Detailed Results
### Test A: Deploy nginx within quota limits
- **Instance:** `quota-test-nginx` (ed846c33-3631-4d54-adce-c7f00210176f)
- **Chart:** charts/nginx:22.1.1
- **Values:** defaults
- **API Response:** HTTP 200, status: `pending-install`
- **Final Status after 21s:** `deployed` ("Instance deployed successfully")
- **K8s Resource Usage:** requests.cpu=100m/2, requests.memory=128Mi/4Gi
### Test B: Deploy nginx exceeding quota
- **Instance:** `quota-test-nginx-2` (36c0350f-089c-41c2-a66e-e93539c00d52)
- **Chart:** charts/nginx:22.1.1
- **Values:** replicaCount=5, resources.limits.cpu=4/memory=8Gi, resources.requests.cpu=2/memory=4Gi
- **API Response:** HTTP 200, status: `pending-install`
- **Final Status (observed for 90s+):** `pending-install` (never transitioned to `deployed` or `failed`)
- **K8s Behavior:**
- Helm release created: `sh.helm.release.v1.quota-test-nginx-2.v1`
- TLS secret created
- Service created, IP assigned
- Deployment created, ReplicaSet scaled up
- **All pod creations FAILED** with: `Error creating: pods "..." is forbidden: exceeded quota: tenant-quota, requested: requests.cpu=2,requests.memory=4Gi, used: requests.cpu=100m,requests.memory=128Mi, limited: requests.cpu=2,requests.memory=4Gi`
### Test C: Deploy GPU instance (gpu quota = 0)
- **Instance:** `quota-test-gpu` (a0d692c8-cdf8-4248-a6d4-1468ad4a7cc7)
- **Chart:** charts/vllm-serve:0.6.0
- **Values:** resources.gpuLimit=1, resources.gpuMem=5000
- **API Response:** HTTP 200, status: `pending-install`
- **Final Status (observed for 30s+):** `pending-install`
- **K8s Behavior:**
- vllm-serve chart defaults: requests.cpu=8, requests.memory=16Gi, requests.nvidia.com/gpu=1, requests.nvidia.com/gpumem=5k
- All pods blocked: `exceeded quota: tenant-quota, requested: requests.cpu=8,requests.memory=16Gi,requests.nvidia.com/gpu=1,..., limited: requests.cpu=2,requests.memory=4Gi,requests.nvidia.com/gpu=0`
---
## Key Findings
### 1. No API-Level (Pre-flight) Quota Enforcement
The backend API accepts **all** deployment requests regardless of whether they exceed the user's quota. There is no validation at the API layer that checks:
- Whether the requested resources exceed the user's quota limits
- Whether the user's quota is already fully consumed by existing deployments
**Evidence:** All three deployments returned HTTP 200 with `status: pending-install`. The backend logs contain zero quota-related entries.
### 2. Kubernetes ResourceQuota Enforces at Pod Level
The Kubernetes `ResourceQuota` object `tenant-quota` in namespace `ocdp-u-test-b` does enforce limits, but only at the **pod creation** level:
```yaml
spec:
hard:
requests.cpu: "2"
requests.memory: 4Gi
requests.nvidia.com/gpu: "0"
requests.nvidia.com/gpumem: "0"
```
When pods exceed quota, Kubernetes explicitly refuses to create them with a clear error message.
However, Helm releases, Services, Deployments, and ReplicaSets are **still created** even when pods are blocked.
### 3. Stuck at "pending-install"
Instances that exceed quota remain stuck in `pending-install` status **indefinitely** — they never transition to `deployed`, `failed`, or any error status. The OCDP platform does not detect the ResourceQuota rejection and update the instance status accordingly. The only way to know about the failure is to check Kubernetes events directly:
```bash
kubectl get events -n ocdp-u-test-b
```
### 4. GPU Quota Enforcement
Users with `gpu=0` quota **can** submit deployments referencing GPU-enabled charts. The API does not reject them. Only the K8s ResourceQuota blocks pod creation at runtime. This could lead to:
- Unnecessary Helm releases and resource overhead in the cluster
- Confusion for users whose deployments appear to hang at `pending-install`
### 5. Quota Exposed in Login Response
The login response includes quota information:
```json
{
"quotaCpu": "2",
"quotaMemory": "4Gi",
"quotaGpu": "0",
"quotaGpuMemory": "0"
}
```
This could be used by the frontend to show usage limits, but no pre-flight check uses it server-side.
---
## Recommendations
1. **Add pre-flight quota validation** in the backend API: before accepting a deployment, check whether the requested resources (from chart values) would exceed the user's quota. Return HTTP 4xx with a clear error message.
2. **Handle "pending-install" timeout**: implement a watcher that detects when a Helm release has been created but pods remain stuck (e.g., due to ResourceQuota) and:
- Update instance status to `failed` with a descriptive `statusReason`
- Clean up the Helm release, Service, etc.
- Optionally surface the K8s error message via the API
3. **GPU quota pre-check**: if a chart requests GPU resources and the user's `gpu=0`, reject the deployment at the API level before creating any Kubernetes resources.
4. **UI quota indicator**: show remaining quota (used vs. hard limit) on the deployment form so users know their limits before submitting.
---
## ResourceQuota YAML (for reference)
```yaml
apiVersion: v1
kind: ResourceQuota
metadata:
name: tenant-quota
namespace: ocdp-u-test-b
labels:
ocdp.io/managed-by: ocdp
ocdp.io/tenant: ocdp-u-test-b
spec:
hard:
requests.cpu: "2"
requests.memory: 4Gi
requests.nvidia.com/gpu: "0"
requests.nvidia.com/gpumem: "0"
```
---
## Cleanup Verification
All test instances were removed after testing:
- `quota-test-nginx` ✅ deleted (pods terminated, helm release removed, quota back to 0)
- `quota-test-nginx-2` ✅ cleaned up (no pods created, resources released)
- `quota-test-gpu` ✅ cleaned up (no pods created, resources released)
- ResourceQuota used: all resources at 0

141
docs/test2-report.md
View File

@ -1,141 +0,0 @@
# OCDP 第二次测试报告
**测试日期:** 2026-05-11
**测试环境:** http://10.6.80.114:18080
---
## 测试1: 资源配额限额
### 测试方法
使用 test-user-b（quota: cpu=2, mem=4Gi, gpu=0, gpumem=0）在 k3s 集群部署 nginx chart
### 测试结果
| 测试 | 操作 | 预期 | 实际 | 结论 |
|------|------|------|------|------|
| Test A | 部署 nginx（默认值，在配额内） | 成功 | 部署完成，状态 deployed | ✅ |
| Test B | 部署 nginx（requests.cpu=2, mem=4Gi, replica=5，超配额） | 被配额阻止 | Helm release 创建成功，所有 Pod 被 ResourceQuota 阻塞，状态永远 stuck 在 pending-install | ⚠️ 部分通过 |
| Test C | 部署 vllm-serve（gpuLimit=1，gpu配额=0） | 被配额阻止 | Helm release 创建成功，Pod 被 ResourceQuota 阻塞，状态 pending-install | ⚠️ 部分通过 |
### 关键发现
**1. 没有 API 层的预检查配额验证**
- 后端 API 无条件接受所有部署请求，不检查是否超配额
- 所有超配额请求返回 HTTP 200 + status: pending-install
- 后端日志中**没有任何配额相关的条目**
**2. K8s ResourceQuota 在 Pod 级别执行**
- `tenant-quota` ResourceQuota 对象确实存在并执行限制
- 当 Pod 超配额时，K8s 明确拒绝创建并给出错误消息
- 但 Helm release、Service、Deployment、ReplicaSet **仍然被创建**
**3. 实例永远 stuck 在 "pending-install"**
- 超配额的实例永远不会转换到 deployed/failed/error
- OCDP 平台不检测 ResourceQuota 拒绝事件
- 唯一知道失败的方式是直接查 K8s events
**4. GPU 配额绕过**
- gpu=0 的用户可以提交需要 GPU 的 chart 部署
- K8s ResourceQuota 最终会阻止，但 Helm release 等资源已被创建
**5. 有效的 ResourceQuota 配置**
```yaml
apiVersion: v1
kind: ResourceQuota
metadata:
name: tenant-quota
namespace: ocdp-u-test-b
spec:
hard:
requests.cpu: "2"
requests.memory: 4Gi
requests.nvidia.com/gpu: "0"
requests.nvidia.com/gpumem: "0"
```
### 建议
1. **添加 API 层预检查配额验证** — 在接受部署前检查请求资源是否超过用户配额
2. **处理 pending-install 超时** — 监控 Helm release 创建后 Pod 是否 stuck，更新状态为 failed
3. **GPU 配额预检查** — 如果 chart 需要 GPU 而用户 gpu=0，在 API 层拒绝
4. **UI 配额指示器** — 在部署表单上显示剩余配额
---
## 测试2: values.yaml 覆盖优先级
### 测试方法
使用 test-user-c（quota: cpu=4, mem=8Gi, gpu=1, gpumem=5000）部署 vllm-serve:0.6.0 chart
### 测试结果
| 方法 | 提交方式 | 是否部署成功 | 存储的值 | 结论 |
|------|----------|-------------|---------|------|
| 方法1 | `values` JSON 字段 | ✅ | cpuRequest=2, gpuMem=10000 | JSON 值被准确接受和存储 |
| 方法2 | `valuesYaml` 字符串 | ✅ | cpuRequest=4, gpuMem=10000 | YAML 被正确解析为结构化 values |
| 方法3 | 同时提供 `values` + `valuesYaml`（冲突） | ✅ 无任何错误/警告 | **values JSON 全胜** | `values` JSON 静默覆盖 `valuesYaml` |
| 方法4 | 不提供任何 values（使用 chart 默认） | ✅ | 仅 namespace | chart 默认值不存储在 API 响应中 |
### 优先级最终结论
| 优先级 | 来源 | 说明 |
|--------|------|------|
| **最高** | `values` JSON 字段 | 请求体中的结构化 JSON |
| **中** | `valuesYaml` 字符串 | 请求体中的 YAML 字符串 |
| **最低** | Chart 内置 values.yaml | Helm chart 打包的默认值 |
### 冲突测试详细结果
当同时提供 `values` 和 `valuesYaml` 且值冲突时：
- `values` JSON 字段**完全覆盖** `valuesYaml`
- **没有任何错误或警告**返回给用户
- 两者被合并到统一的 DB `values` 字段
### gpuMem=10000 行为
- 整数值 `10000` 在 `values` JSON 和 `valuesYaml` 中都被**正确接受**
- 无单位转换（作为整数 MB 标量存储）
- 符合项目规范
### 建议
1. **记录优先级顺序** — 用户需知道同时提供两者时 values JSON 优先
2. **添加冲突警告** — 当两个字段存在冲突值时应返回警告
3. **考虑废弃一个字段** — values 和 valuesYaml 语义重复易混淆
---
## 测试3: 前端 UI 溢出/滚动/刷新
### 测试方法
Playwright + 源码分析，测试 1920/768/375 三个视口
### 测试结果
**总体结论: PASS** — 没有导致功能问题的关键溢出问题
| 测试项 | 结果 | 详情 |
|--------|------|------|
| 水平溢出 | ✅ 无问题 | 所有视口均无水平溢出 |
| 文本截断 | ⚠️ 1 个低风险 | InstanceCard h3 标题 truncate 无 title tooltip |
| 响应式设计 | ✅ 正确 | sm/md/lg/xl 断点覆盖完整 |
| 滚动行为 | ✅ 流畅 | Sidebar 和内容区独立滚动，overscroll-contain 防滚动穿透 |
| 模态框布局 | ✅ 正确 | body scroll lock + 内容独立滚动 |
| 页面刷新 | ✅ 正常 | 受保护路由正确重定向到登录页 |
| 颜色对比度 | ⚠️ 1 个中风险 | 登录页错误文本 red-400 在白色背景上仅 2.5:1 (WCAG AA 要求 4.5:1) |
### 通过的细分项
- Chart Browser 全高 + overflow-y-auto 布局 ✅
- InstanceCard 操作按钮网格 grid-cols-2/3/5 响应正确 ✅
- Tabs 支持 overflow-x-auto 水平滚动 ✅
- 用户管理表格 overflow-x-auto ✅
- iOS 触摸滚动 (`-webkit-overflow-scrolling: touch`) 已配置 ✅
### 建议
1. 将登录页错误文本从 text-red-400 改为 text-red-600/700
2. InstanceCard h3 标题添加 title 属性
---
## 综合建议
1. 添加 API 层配额预检查
2. 处理 pending-install 超时 + 状态更新
3. 记录 values 覆盖优先级并添加冲突警告
4. 统一 values JSON/YAML 的 API 设计

271
docs/test2-ui-overflow.md
View File

@ -1,271 +0,0 @@
# QA Report: UI Layout Overflow & Responsiveness Test
**Date:** 2026-05-11
**Environment:** http://10.6.80.114:18080
**Browser:** Chromium (Playwright headless)
**Test Credentials:** test-user-a / TestUserA123!
---
## Test Results Summary
| # | Test | Status | Issues Found |
|----|------|--------|-------------|
| 1 | Login Page Layout | ✅ Pass | 1 Low |
| 2 | Home Page | ✅ Pass | 0 |
| 3 | Chart Browser (Registries) | ✅ Pass | 0 |
| 4 | Instances Page | ✅ Pass | 0 |
| 5 | Monitoring Page | ✅ Pass | 0 |
| 6 | Tablet Responsive (768px) | ✅ Pass | 0 |
| 7 | Mobile Responsive (375px) | ✅ Pass | 0 |
| 8 | Deep DOM Overflow Analysis | ✅ Pass | 0 |
| 9 | Source Code CSS Pattern Audit | ✅ Pass | 2 Info |
| 10 | Text Visibility & Contrast | ⚠️ 1 Issue | 1 Medium |
---
## 1. Login Page (AuthPage.tsx)
**Location:** `frontend/src/features/auth/pages/AuthPage.tsx`
**Layout:**
- Form card is `max-w-md` (448px), horizontally centered via `flex items-center justify-center`
- Desktop viewport (1920×1080): card is perfectly centered (checked via bounding rect)
- Background: `bg-slate-50` with gradient overlay
- Card: `bg-white/95 backdrop-blur-xl` with `shadow-2xl`
**Responsive:**
- Padding: `px-4 sm:px-6` — increases from 16px → 24px on `sm:` breakpoint
- Card padding: `p-6 sm:p-7`
- Icon: `w-11 h-11` — fixed size, not responsive
### ✅ Issue #1-LOW: Login error text color contrast
- **File:** `AuthPage.tsx:96`
- **Pattern:** `<p className="text-red-400 text-center text-sm">`
- **Problem:** `text-red-400` (`#f87171`) on white background has a contrast ratio of ~2.5:1, which fails WCAG AA (minimum 4.5:1 for normal text). Error messages may be hard to read for users with visual impairments.
- **Recommendation:** Use `text-red-600` or `text-red-700` for error text on white backgrounds.
---
## 2. Home Page
**Location:** `frontend/src/features/home/pages/HomePage.tsx`
**Layout:**
- Main container: `min-h-full bg-slate-50 px-4 py-6 sm:px-6 lg:px-8`
- Two-column layout on large screens: `lg:grid-cols-[1.4fr_0.8fr]`
- Feature cards: `md:grid-cols-3`
- Quick actions: `md:grid-cols-3`
**Scroll:** ScrollHeight=1080, Viewport=1080 — content fits exactly without scrolling on 1080p.
**Overflow:** No horizontal overflow detected. Proper use of responsive padding and grid columns.
### Passing — no issues found.
---
## 3. Chart Browser / Registries
**Location:** `frontend/src/features/artifact/registries/pages/ArtifactBrowserPage.tsx`
**Layout (Desktop):**
- Main layout: `flex-1 flex overflow-hidden bg-slate-50` (sidebar + detail panes)
- Sidebar tree: `flex-1 overflow-y-auto custom-scrollbar`
- Detail pane: `flex-1 flex flex-col bg-white overflow-hidden`
- Tag grid: `grid-cols-1 md:grid-cols-2 xl:grid-cols-3 gap-4`
**Tablet (768px):** No overflow. Grid collapses to 2 columns.
**Mobile (375px):** No overflow. Grid collapses to 1 column.
### Key Patterns Found:
- `RepositoryItem.tsx:212` — `<span className="text-sm text-gray-200 font-mono truncate" title={repository}>` — proper truncation with `title` tooltip
- `ArtifactBrowserPage.tsx:336` — `<p className="text-[11px] text-slate-500 truncate">` — uses 11px text with truncation
- `TagCard.tsx` — uses truncation with `title` attribute for long names
### Passing — no overflow issues found.
---
## 4. Instances Page
**Location:** `frontend/src/features/artifact/instances/pages/InstancesManagementPage.tsx`
**Component:** `InstanceCard.tsx`
**Layout:**
- Cluster cards: responsive grid `clusters.length > 1 ? 'md:grid-cols-3' : 'md:grid-cols-2'`
- Instance cards listed in single column then `lg:grid-cols-2 gap-6`
- Action buttons grid: `grid-cols-2 gap-2 md:grid-cols-3 xl:grid-cols-5`
### ✅ Issue #2-INFO: Action button text truncation on InstanceCards
- **File:** `InstanceCard.tsx:285-327`
- **Pattern:**
```
<div className="grid grid-cols-2 gap-2 md:grid-cols-3 xl:grid-cols-5">
<button>
<span className="truncate">Refresh</span>
</button>
</div>
```
- **Analysis:** At `grid-cols-2` (small screens), two buttons share each row. The buttons use `min-w-0` which allows them to shrink, and `truncate` on the text span. However, the button text is short ("Refresh", "Entries", "Diagnostics", "Modify", "Delete"), so truncation is unlikely to occur in practice.
- **Mitigation:** Each `<span>` has `title` attribute on parent button, providing tooltip fallback.
- **Verdict:** Acceptable — button labels are intentionally short and tooltips are present.
### ✅ Issue #3-INFO: Header text truncation with tooltip
- **File:** `InstanceCard.tsx:185`
- **Pattern:** `<h3 className="text-xl font-bold text-slate-950 truncate">{instanceName}</h3>`
- **Analysis:** Instance names could be long, `truncate` will clip with ellipsis. No `title` attribute on this element — unlike repository text below it.
- **Recommendation:** Add `title={instanceName}` to the `<h3>` element for tooltip on overflow.
### Passing — no critical overflow issues found.
---
## 5. Monitoring / Clusters
**Location:** `frontend/src/features/monitoring/clusters/`
**Layout:**
- Cluster cards grid: `grid-cols-1 sm:grid-cols-2 lg:grid-cols-4 gap-4`
- Card header: `<h3 className="text-lg font-semibold text-slate-900 truncate">` with cluster name
- Metrics: `grid-cols-2 sm:grid-cols-4 gap-4 mb-3`
- Resource bars: `overflow-hidden` for proper progress bar clipping
- Node details: `grid-cols-1 lg:grid-cols-2 gap-3`
**Overflow Check:** ScrollWidth = clientWidth at all tested viewports — no horizontal overflow.
**Responsive:**
- 1920px: 4 columns of cluster cards
- 768px: 2 columns
- 375px: 1 column
### Passing — no issues found.
---
## 6. Sidebar Layout
**Location:** `frontend/src/shared/components/layout/SidebarLayout/`
**Layout:**
- Parent: `min-h-screen flex bg-dark text-primary overflow-hidden`
- Nav: `flex-1 p-3 space-y-1 overflow-y-auto` — independently scrollable
- Footer: Fixed at bottom, `p-3 text-xs text-muted`
**Scroll Analysis:**
- Content area has `overflow-y-auto`, so sidebar nav items scroll independently when they exceed viewport height
- The footer anchors to the bottom of the sidebar (not the scroll area)
- At 1080px viewport, sidebar content fits without scrolling
**Potential Concern:** If many nav items are added, the footer will push below the fold and the user must scroll the nav to see it. The `overflow-y-auto` on the `<nav>` element handles this correctly.
### Passing — no issues found.
---
## 7. Tabs Component
**Location:** `frontend/src/shared/components/layout/Tabs.tsx`
**Pattern:**
```
<div className="flex gap-2 sm:gap-4 overflow-x-auto scrollbar-thin scrollbar-thumb-gray-600 scrollbar-track-transparent">
<button className="whitespace-nowrap flex-shrink-0">
```
**Analysis:**
- Uses `overflow-x-auto` — horizontal scroll appears when tabs exceed container width
- `whitespace-nowrap` prevents tab text from wrapping
- `flex-shrink-0` prevents tab items from compressing
- Custom thin scrollbar styling for webkit browsers
**Edge Case:** On very small viewports with many tabs, users need to horizontally scroll. This is acceptable UX for a tabs pattern.
### Passing — no issues found.
---
## 8. Modal Component
**Location:** `frontend/src/shared/components/layout/Modal.tsx`
**Pattern:**
- Body scroll lock: sets `document.body.style.overflow = 'hidden'` on open
- Modal overlay: `fixed inset-0 z-[90] flex items-start sm:items-center justify-center overflow-y-auto p-4 sm:p-6`
- Content: `max-h-[calc(100vh-12rem)] sm:max-h-[calc(100vh-10rem)] overflow-y-auto`
**Analysis:**
- Properly handles body scroll prevention
- Modal content is independently scrollable when content exceeds viewport
- `padding-right` compensation prevents layout shift when scrollbar disappears
### Passing — no issues found.
---
## 9. Deep DOM Overflow Analysis
The Playwright script ran a comprehensive scan of ALL DOM elements, checking:
- `overflow: hidden` elements where `scrollWidth > clientWidth` (clipped content)
- `text-overflow: ellipsis` elements where content overflows
- `white-space: nowrap` causing overflow
- Tiny text (< 10px)
**Result: No overflow:hidden clipping detected, no text truncation overflow detected on the monitoring page (last page tested).**
Note: The test scans the current page's DOM after navigating through pages. Some truncation exists (`InstanceCard.tsx`, `RepositoryItem.tsx`, etc.) but all uses are intentional and include `title` tooltip fallbacks.
---
## 10. CSS Pattern Summary
### Overflow Patterns Used in Codebase
| Pattern | Used For | Risk |
|---------|----------|------|
| `overflow-hidden` | Card containers, modal wrappers, progress bars | Low — decorative/structural |
| `overflow-y-auto` | Scrollable content areas (sidebar nav, modals, detail panes) | None — intentional scroll |
| `overflow-x-auto` | Tabs, data tables | Low — scroll indicator needed |
| `truncate` | Instance names, repository names, tags, button labels | Low — tooltips provided on most |
| `whitespace-nowrap` | Tab items, table headers | Low — paired with overflow-x-auto |
| `line-clamp-1` | Registry descriptions | None — CSS line clamp |
### Responsive Breakpoints Used
| Breakpoint | Usage |
|-----------|-------|
| `sm:` (640px) | Login form padding, cluster form layout, button layouts, tabs spacing |
| `md:` (768px) | Grid columns (2-3 cols), card layouts, diagnostics modal |
| `lg:` (1024px) | Two-column layouts, 4-col monitoring grids |
| `xl:` (1280px) | 5 action button columns, 3-col tag grids |
| `2xl:` (1536px) | Not used |
### Fixed Widths Checked
No problematic fixed widths found. All layouts use `max-w-` constraints (`max-w-md`, `max-w-6xl`, `max-w-7xl`) rather than fixed pixel widths.
---
## 11. Page Refresh Behavior
The SPA uses React Router. When navigating to authenticated routes:
- `ProtectedRoute` component checks `isAuthenticated` and `isAllowed`
- If not authenticated, users are redirected to login page (`/`)
- After login, `navigate("/home", { replace: true })` navigates to home
- Page refresh at any route should redirect to login if token is expired
**Note:** The `/login` route path does not exist in the SPA router — login is handled by `AuthPage` rendered at the root `/` path when the user is unauthenticated.
---
## Final Verdict
| Category | Score |
|----------|-------|
| Horizontal Overflow | ✅ No issues at any viewport |
| Text Truncation | ⚠️ InstanceCard `<h3>` missing tooltip fallback |
| Responsive Design | ✅ Proper breakpoints at sm/md/lg/xl |
| Scroll Behavior | ✅ Sidebar and content areas properly scrollable |
| Color Contrast | ⚠️ Login error text (red-400) fails WCAG AA |
| Modal Layout | ✅ Body scroll lock + content scroll work correctly |
| Page Refresh | ✅ Protected routes redirect to login |
**Overall: PASS** — Two minor issues found (color contrast on login error text, missing tooltip on InstanceCard title), neither causing functional problems.

110
docs/test2-values-priority.md
View File

@ -1,110 +0,0 @@
# Test Report: values.yaml Override Priority
**Date:** 2026-05-11
**Tester:** test-user-c
**Cluster:** dbf824f1-9962-4d8e-881e-870c75fdb6f5
**Chart:** charts/vllm-serve:0.6.0
**Namespace:** ocdp-u-test-c
---
## Test Results
### Method 1: `values` JSON field only (vllm-values-json)
- **Deployment:** ✅ Success (status: pending-install, no errors)
- **Submitted values:**
```json
{ "cpuRequest": 2, "gpuLimit": 1, "gpuMem": 10000, "memoryLimit": "4Gi" }
```
- **Stored values (from API response):**
```json
{ "cpuRequest": 2, "gpuLimit": 1, "gpuMem": 10000, "memoryLimit": "4Gi" }
```
- **Result:** Values were accepted and stored exactly as provided. No chart defaults were merged into the stored representation (e.g., `shmSize: "8Gi"` from chart defaults is absent).
### Method 2: `valuesYaml` string field only (vllm-values-yaml)
- **Deployment:** ✅ Success (status: pending-install, no errors)
- **Submitted valuesYaml:**
```yaml
resources:
cpuRequest: 4
gpuLimit: 1
gpuMem: 10000
memoryLimit: "8Gi"
model:
huggingfaceName: "Qwen/Qwen2.5-0.5B-Instruct"
```
- **Stored values (parsed and stored in DB):**
```json
{ "cpuRequest": 4, "gpuLimit": 1, "gpuMem": 10000, "memoryLimit": "8Gi" }
```
- **Result:** The YAML string was correctly parsed into the structured `values` field in the database. YAML parsing works correctly.
### Method 3: Both `values` JSON AND `valuesYaml` with conflict (vllm-conflict-test)
- **Deployment:** ✅ Success (status: pending-install, **no error or warning returned**)
- **`values` JSON submitted:**
```json
{ "cpuRequest": 4, "memoryLimit": "8Gi", "huggingfaceName": "Qwen/Qwen2.5-0.5B-Instruct" }
```
- **`valuesYaml` submitted:**
```yaml
resources:
cpuRequest: 2
memoryLimit: "4Gi"
model:
huggingfaceName: "Qwen/Qwen2.5-7B-Instruct"
```
- **Stored values:**
```json
{ "cpuRequest": 4, "gpuLimit": 1, "gpuMem": 10000, "memoryLimit": "8Gi", "huggingfaceName": "Qwen/Qwen2.5-0.5B-Instruct" }
```
- **Result:** The `values` JSON field **won every conflict**. The `valuesYaml` values (`cpuRequest: 2`, `memoryLimit: "4Gi"`, `Qwen/Qwen2.5-7B-Instruct`) were completely overridden by the `values` JSON values (`cpuRequest: 4`, `memoryLimit: "8Gi"`, `Qwen/Qwen2.5-0.5B-Instruct`). No error or warning was presented to the user.
### Method 4: No values (chart defaults, vllm-defaults-test)
- **Deployment:** ✅ Success (status: pending-install, no errors)
- **Stored values:**
```json
{ "namespace": "ocdp-u-test-c" }
```
- **Result:** Only the auto-injected `namespace` was stored. Chart defaults (`cpuRequest: 8`, `memoryLimit: "16Gi"`, etc.) are not stored in the API response — they are resolved at Helm deploy time.
---
## Key Findings
### 1. Override Priority Order (when both fields provided)
| Priority | Source | Description |
|----------|--------|-------------|
| **Highest** | `values` JSON field | Structured JSON object in the request body |
| **Lowest** | `valuesYaml` string field | Raw YAML string in the request body |
| **Baseline** | Chart built-in `values.yaml` | Default values packaged in the Helm chart |
### 2. Conflict Resolution
When both `values` and `valuesYaml` are provided with conflicting values:
- **`values` JSON wins** — the structured JSON field takes priority over the YAML string
- **No error or warning** is returned to the user
- The system silently prefers the `values` JSON field
### 3. `gpuMem=10000` Behavior
- The integer value `10000` was **accepted without issues** in both `values` JSON and `valuesYaml` formats
- No normalization or unit conversion was applied (stored as-is: `10000`)
- Consistent with the project convention that `nvidia.com/gpumem` is treated as a vendor integer MB scalar
### 4. All values are stored in a unified `values` field in the DB
Both `values` JSON and `valuesYaml` inputs are converted to a single structured `values` JSON object in the database. The API response always returns the structured `values` field regardless of how the input was provided.
---
## Recommendations
1. **Document the priority order** — users should know that when providing both `values` and `valuesYaml`, the `values` JSON field takes precedence and no error is raised.
2. **Consider returning a warning** when both fields are provided with conflicting values, as silent override could cause confusion.
3. **The naming convention** (`values` vs `valuesYaml`) can be misleading since both ultimately serve the same purpose. Consider deprecating one in the API to avoid ambiguity.

30
tasks/session-notes.md
View File

@ -1,30 +0,0 @@
# OCDP 系统测试 - 完成报告
## 已交付文档
| 文档 | 路径 | 大小 |
|------|------|------|
| 用户操作指南 | `docs/user-guide.md` | 752 lines |
| 测试场景设计 | `docs/test-scenarios.md` | 67KB, 12 分类, 100+ 用例 |
| 测试用户凭据 | `docs/test-users.json` | 4 个账号 |
| 综合 Bug 报告 | `docs/bug-report.md` | 18 个 Bug, 含安全发现 |
| user-a 测试报告 | `docs/bugs-user-a.md` | 前端 UI 发现 |
| user-b 测试报告 | `docs/bugs-user-b.md` | API/部署发现 |
| user-c 测试报告 | `docs/bugs-user-c.md` | 权限隔离发现 |
| 安全测试报告 | `docs/security/bugs-security.md` | 6 个安全发现 |
## Bug 统计: 18 个
| 严重度 | 数量 | 说明 |
|--------|------|------|
| **P0 (Blocker)** | 2 | Launch 按钮无反应、SPA 路由空白页 |
| **P1 (High)** | 2 | DELETE 404 + 空 body |
| **P2 (Medium)** | 6 | 缺失 API、静默 namespace 覆盖、无障碍问题 |
| **P3 (Low)** | 8 | 响应格式、安全头缺失、CORS、用户枚举等 |
| **Total** | **18** | |
## 测试团队
- user-a-agent ✅ (前端测试)
- user-b-agent ✅ (API/部署测试)
- user-c-agent ✅ (权限隔离测试)
- security-agent ✅ (安全测试)

43
tasks/todo.md
View File

@ -1,36 +1,11 @@
# OCDP 第二次测试 - 完成
# OCDP 最终文档结构
## 交付文档
| 文档 | 路径 | 内容 |
## docs/ 目录 (已清理)
| 文件 | 用途 | 状态 |
|------|------|------|
| 综合报告 | `docs/test2-report.md` | 3 个测试的完整结果 |
| 配额测试详情 | `docs/test2-quota.md` | 配额限额详细分析 |
| Values 优先级测试 | `docs/test2-values-priority.md` | values 覆盖测试+冲突测试 |
| UI 溢出/滚动/刷新 | `docs/test2-ui-overflow.md` | Playwright + 源码分析 |
## 核心发现
### 1. 资源配额
| 发现 | 影响 |
|------|------|
| ✅ K8s ResourceQuota 对象正确创建并生效 | cpu/gpu/mem 限制在 pod 级别执行 |
| ❌ **无 API 层预检查** | 后端接受所有部署请求，配额耗尽时 pod stuck pending-install |
| ❌ **GPU 配额可绕过** | gpu=0 用户能提交需要 GPU 的 chart |
| ❌ **实例不会自动 failed** | 超配额实例永远 stuck 在 pending-install |
### 2. Values 覆盖优先级
| 优先级 | 来源 | 说明 |
|--------|------|------|
| 🥇 **最高** | `values` JSON 字段 | 结构化 JSON - 覆盖一切 |
| 🥈 **中** | `valuesYaml` 字符串 | 被 values JSON 覆盖 |
| 🥉 **最低** | Chart 内置 values.yaml | 默认基线 |
| ⚠️ **冲突时静默覆盖，无警告** | 两者都提供时 values JSON 全胜 | |
### 3. 前端 UI
| 测试 | 结论 |
|------|------|
| 水平溢出 | ✅ 无问题 |
| 响应式 | ✅ sm/md/lg/xl 正确 |
| 滚动 | ✅ 流畅 |
| 刷新 | ✅ 正常 |
| 颜色对比度 | ⚠️ 登录错误文本 red-400 WCAG AA 不合格 |
| `user-guide.md` | 用户操作指南 | ✅ 永久参考 |
| `test-scenarios.md` | 100+ 测试用例设计 | ✅ 永久参考 |
| `test-users.json` | 4 个测试账号凭证 | ✅ 永久参考 |
| `regression-full-report.md` | 最新综合回归报告 | ✅ 可删除（下一个版本） |
| `UNRESOLVED-BUGS.md` | 未修复问题清单 (15 个) | ✅ 当前版本 |