fix(backend): 修复转发监听 IP 为 IPv6 时报missing port in address (#397)

* fix(backend): handle IPv6 forward bind IP ports

* test(handler): add IPv6 bindIP test cases for forward service config

---------

Co-authored-by: sagit <36596628+Sagit-chu@users.noreply.github.com>
Co-authored-by: sagitchu <sagitchu@gmail.com>

.agents/skills/security-scan/SKILL.md

@@ -1,165 +0,0 @@
----
-name: security-scan
-description: Scan your Claude Code configuration (.claude/ directory) for security vulnerabilities, misconfigurations, and injection risks using AgentShield. Checks CLAUDE.md, settings.json, MCP servers, hooks, and agent definitions.
-origin: ECC
----
-
-# Security Scan Skill
-
-Audit your Claude Code configuration for security issues using [AgentShield](https://github.com/affaan-m/agentshield).
-
-## When to Activate
-
-- Setting up a new Claude Code project
-- After modifying `.claude/settings.json`, `CLAUDE.md`, or MCP configs
-- Before committing configuration changes
-- When onboarding to a new repository with existing Claude Code configs
-- Periodic security hygiene checks
-
-## What It Scans
-
-| File | Checks |
-|------|--------|
-| `CLAUDE.md` | Hardcoded secrets, auto-run instructions, prompt injection patterns |
-| `settings.json` | Overly permissive allow lists, missing deny lists, dangerous bypass flags |
-| `mcp.json` | Risky MCP servers, hardcoded env secrets, npx supply chain risks |
-| `hooks/` | Command injection via interpolation, data exfiltration, silent error suppression |
-| `agents/*.md` | Unrestricted tool access, prompt injection surface, missing model specs |
-
-## Prerequisites
-
-AgentShield must be installed. Check and install if needed:
-
-```bash
-# Check if installed
-npx ecc-agentshield --version
-
-# Install globally (recommended)
-npm install -g ecc-agentshield
-
-# Or run directly via npx (no install needed)
-npx ecc-agentshield scan .
-```
-
-## Usage
-
-### Basic Scan
-
-Run against the current project's `.claude/` directory:
-
-```bash
-# Scan current project
-npx ecc-agentshield scan
-
-# Scan a specific path
-npx ecc-agentshield scan --path /path/to/.claude
-
-# Scan with minimum severity filter
-npx ecc-agentshield scan --min-severity medium
-```
-
-### Output Formats
-
-```bash
-# Terminal output (default) — colored report with grade
-npx ecc-agentshield scan
-
-# JSON — for CI/CD integration
-npx ecc-agentshield scan --format json
-
-# Markdown — for documentation
-npx ecc-agentshield scan --format markdown
-
-# HTML — self-contained dark-theme report
-npx ecc-agentshield scan --format html > security-report.html
-```
-
-### Auto-Fix
-
-Apply safe fixes automatically (only fixes marked as auto-fixable):
-
-```bash
-npx ecc-agentshield scan --fix
-```
-
-This will:
-- Replace hardcoded secrets with environment variable references
-- Tighten wildcard permissions to scoped alternatives
-- Never modify manual-only suggestions
-
-### Opus 4.6 Deep Analysis
-
-Run the adversarial three-agent pipeline for deeper analysis:
-
-```bash
-# Requires ANTHROPIC_API_KEY
-export ANTHROPIC_API_KEY=your-key
-npx ecc-agentshield scan --opus --stream
-```
-
-This runs:
-1. **Attacker (Red Team)** — finds attack vectors
-2. **Defender (Blue Team)** — recommends hardening
-3. **Auditor (Final Verdict)** — synthesizes both perspectives
-
-### Initialize Secure Config
-
-Scaffold a new secure `.claude/` configuration from scratch:
-
-```bash
-npx ecc-agentshield init
-```
-
-Creates:
-- `settings.json` with scoped permissions and deny list
-- `CLAUDE.md` with security best practices
-- `mcp.json` placeholder
-
-### GitHub Action
-
-Add to your CI pipeline:
-
-```yaml
-- uses: affaan-m/agentshield@v1
-  with:
-    path: '.'
-    min-severity: 'medium'
-    fail-on-findings: true
-```
-
-## Severity Levels
-
-| Grade | Score | Meaning |
-|-------|-------|---------|
-| A | 90-100 | Secure configuration |
-| B | 75-89 | Minor issues |
-| C | 60-74 | Needs attention |
-| D | 40-59 | Significant risks |
-| F | 0-39 | Critical vulnerabilities |
-
-## Interpreting Results
-
-### Critical Findings (fix immediately)
-- Hardcoded API keys or tokens in config files
-- `Bash(*)` in the allow list (unrestricted shell access)
-- Command injection in hooks via `${file}` interpolation
-- Shell-running MCP servers
-
-### High Findings (fix before production)
-- Auto-run instructions in CLAUDE.md (prompt injection vector)
-- Missing deny lists in permissions
-- Agents with unnecessary Bash access
-
-### Medium Findings (recommended)
-- Silent error suppression in hooks (`2>/dev/null`, `|| true`)
-- Missing PreToolUse security hooks
-- `npx -y` auto-install in MCP server configs
-
-### Info Findings (awareness)
-- Missing descriptions on MCP servers
-- Prohibitive instructions correctly flagged as good practice
-
-## Links
-
-- **GitHub**: [github.com/affaan-m/agentshield](https://github.com/affaan-m/agentshield)
-- **npm**: [npmjs.com/package/ecc-agentshield](https://www.npmjs.com/package/ecc-agentshield)

.claude/settings.json

@@ -1,84 +0,0 @@
-{
-  "hooks": {
-    "PostToolUse": [
-      {
-        "matcher": "Task",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code post-task"
-          }
-        ]
-      },
-      {
-        "matcher": "TodoWrite",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code post-todo"
-          }
-        ]
-      }
-    ],
-    "PreToolUse": [
-      {
-        "matcher": "Task",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code pre-task"
-          }
-        ]
-      }
-    ],
-    "SessionEnd": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code session-end"
-          }
-        ]
-      }
-    ],
-    "SessionStart": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code session-start"
-          }
-        ]
-      }
-    ],
-    "Stop": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code stop"
-          }
-        ]
-      }
-    ],
-    "UserPromptSubmit": [
-      {
-        "matcher": "",
-        "hooks": [
-          {
-            "type": "command",
-            "command": "entire hooks claude-code user-prompt-submit"
-          }
-        ]
-      }
-    ]
-  },
-  "permissions": {
-    "deny": [
-      "Read(./.entire/metadata/**)"
-    ]
-  }
-}

.claude/skills/security-scan

@@ -1 +0,0 @@
-../../.agents/skills/security-scan

.gitignore

@@ -62,6 +62,9 @@ go-gost/ss/
 .classpath
 .project
 .settings/
+
+# OpenCode session metadata
+.entire/
 bin/
 tmp/
 *.swp

.sisyphus/plans/001-setup-dev-environment.md

@@ -1,71 +0,0 @@
-# Plan: 搭建开发环境
-
-## 目标
-为 Flux Panel 项目安装所有缺失的开发依赖，使 3 个子项目都能本地开发和构建。
-
-## 当前状态
-
-### ✅ 已安装
-| 工具 | 版本 | 用途 |
-|------|------|------|
-| Node.js | v20.19.2 | vite-frontend |
-| npm | 9.2.0 | vite-frontend |
-| Go | 1.24.4 | go-gost |
-| Docker | 29.1.4 | 容器化部署 |
-
-### ❌ 缺失
-| 工具 | 需求版本 | 用途 |
-|------|----------|------|
-| Java | 21 | springboot-backend |
-| Maven | 3.x | 构建后端 |
-| Docker Compose | v2 | 容器编排 |
-
----
-
-## 执行任务
-
-### Task 1: 安装 Java 21
-```bash
-apt-get update && apt-get install -y openjdk-21-jdk
-```
-**验证**: `java -version` 应显示 openjdk 21
-
-### Task 2: 安装 Maven
-```bash
-apt-get install -y maven
-```
-**验证**: `mvn -v` 应显示 Maven 3.x
-
-### Task 3: 安装 Docker Compose Plugin
-```bash
-apt-get install -y docker-compose-plugin
-```
-**验证**: `docker compose version` 应显示版本号
-
-### Task 4: 安装前端依赖
-```bash
-cd /root/flux-panel/vite-frontend && npm install
-```
-**验证**: `node_modules/` 目录存在
-
-### Task 5: 验证后端可构建
-```bash
-cd /root/flux-panel/springboot-backend && mvn clean compile -q
-```
-**验证**: 编译成功无错误
-
-### Task 6: 验证 Go 模块
-```bash
-cd /root/flux-panel/go-gost && go mod download
-```
-**验证**: 依赖下载成功
-
----
-
-## 完成标准
-- [ ] `java -version` → openjdk 21
-- [ ] `mvn -v` → Maven 3.x
-- [ ] `docker compose version` → v2.x
-- [ ] 前端: `npm run dev` 可启动
-- [ ] 后端: `mvn compile` 成功
-- [ ] Go: `go build .` 成功

211-custom-ip-selection.md

@@ -1,33 +0,0 @@
-# Issue #211: 转发自定义监听IP / 隧道指定连接IP
-
-## 需求总结
-1. **节点**: 高级配置增加"额外IP地址"字段（逗号分隔）
-2. **转发**: 创建/编辑时可指定入口监听IP
-3. **隧道**: 配置出口节点时可指定连接IP
-
----
-
-## 任务清单
-
-### 后端
-- [x] 1. 数据模型扩展 - Node/ForwardPort/ChainTunnel 增加字段
-- [x] 2. Repository - CreateNode/UpdateNode 处理 extraIPs
-- [x] 3. Repository - resolveForwardIngress 使用 forward_port.in_ip
-- [x] 4. Repository - GetNodeAllIPs 辅助函数（返回节点所有可用IP）
-- [x] 5. Handler - 转发创建/更新处理 inIp 参数
-- [x] 6. Handler - 隧道出口节点处理 connectIp 参数
-- [x] 7. Handler - 节点API返回 extraIPs 字段
-
-### 前端
-- [x] 8. 节点编辑页 - 高级配置增加"额外IP"输入
-- [x] 9. 转发编辑弹窗 - 增加"监听IP"下拉选择
-- [x] 10. 隧道配置页 - 出口节点增加"连接IP"输入
-
----
-
-## 完成进度
-- 开始时间: 2026-03-02
-- 完成时间: 2026-03-02
-- 完成任务: 10/10
-- 后端完成: ✅
-- 前端完成: ✅

AGENTS.md

@@ -1,9 +1,9 @@
 # PROJECT KNOWLEDGE BASE
 
-**Generated:** Thu Feb 26 2026
-**Commit:** 21008cc
+**Generated:** Tue Mar 24 2026
+**Commit:** 8ebde9d
 **Branch:** main
-**Tag:** 2.1.5-rc15
+**Tag:** 2.1.9-rc10
 
 ## OVERVIEW
 FLVX (formerly Flux Panel) is a traffic forwarding management system built on a forked GOST v3 stack. It ships as a Go-based admin API (SQLite/PostgreSQL) + Vite/React UI + Go forwarding agent, with optional mobile WebView wrappers.

IMPLEMENTATION_PLAN.md

@@ -1,148 +0,0 @@
-# 限速功能重构实施计划
-
-## 一、需求概述
-
-**原始需求**: 限速功能当前绑定到具体隧道，需要改为不绑定隧道，创建限速后可以自由在隧道上限速，也可以在转发上限速。
-
-**核心变更**:
-1. 限速规则(SpeedLimit)与隧道的绑定关系改为可选
-2. 转发(Forward)支持独立的限速规则
-
----
-
-## 二、实施计划清单
-
-### 2.0 计划状态（审计更新：2026-02-26）
-
-- 总体状态：**进行中（未验收通过）**
-- 已完成：模型、仓储查询、限速 CRUD、控制面优先级、限速页与类型改造、编译与测试通过
-- 未完成：**Forward 独立限速写入链路**（前端表单 -> API handler -> repository 落库 `forward.speed_id`）
-
-### 2.1 后端模型层 (Model)
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| M1 | SpeedLimit.TunnelID 改为 sql.NullInt64 (可空) | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-| M2 | SpeedLimit.TunnelName 改为 sql.NullString (可空) | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-| M3 | Forward 添加 SpeedID sql.NullInt64 字段 | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-| M4 | ForwardRecord 添加 SpeedID sql.NullInt64 字段 | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-| M5 | SpeedLimitBackup.TunnelID 改为指针类型 | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-| M6 | ForwardBackup 添加 SpeedID *int64 字段 | `go-backend/internal/store/model/model.go` | ✅ 完成 |
-
-### 2.2 后端仓储层 (Repository)
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| R1 | ListSpeedLimits() 返回可空 tunnelId/tunnelName | `go-backend/internal/store/repo/repository.go` | ✅ 完成 |
-| R2 | ListForwards() 返回 speedId 字段 | `go-backend/internal/store/repo/repository.go` | ✅ 完成 |
-| R3 | CreateSpeedLimit() 参数 tunnelID 改为 *int64 | `go-backend/internal/store/repo/repository_mutations.go` | ✅ 完成 |
-| R4 | UpdateSpeedLimit() 参数 tunnelID 改为 *int64 | `go-backend/internal/store/repo/repository_mutations.go` | ✅ 完成 |
-| R5 | GetSpeedLimitTunnelID() 返回 sql.NullInt64 | `go-backend/internal/store/repo/repository_mutations.go` | ✅ 完成 |
-| R6 | exportSpeedLimits() 处理可空字段 | `go-backend/internal/store/repo/repository.go` | ✅ 完成 |
-| R7 | importSpeedLimits() 处理可空字段 | `go-backend/internal/store/repo/repository.go` | ✅ 完成 |
-| R8 | GetSpeedLimitSpeed() 新增方法 | `go-backend/internal/store/repo/repository_flow.go` | ✅ 完成 |
-| R9 | ListForwardsByTunnel() 返回 SpeedID | `go-backend/internal/store/repo/repository_control.go` | ✅ 完成 |
-| R10 | ListActiveForwardsByUser() 返回 SpeedID | `go-backend/internal/store/repo/repository_flow.go` | ✅ 完成 |
-| R11 | ListActiveForwardsByUserTunnel() 返回 SpeedID | `go-backend/internal/store/repo/repository_flow.go` | ✅ 完成 |
-| R12 | GetForwardRecord() 返回 SpeedID | `go-backend/internal/store/repo/repository_flow.go` | ✅ 完成 |
-
-### 2.3 后端处理器层 (Handler)
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| H1 | speedLimitCreate 处理可选 tunnelId | `go-backend/internal/http/handler/mutations.go` | ✅ 完成 |
-| H2 | speedLimitUpdate 处理可选 tunnelId | `go-backend/internal/http/handler/mutations.go` | ✅ 完成 |
-| H3 | speedLimitDelete 处理可空 tunnelID | `go-backend/internal/http/handler/mutations.go` | ✅ 完成 |
-
-### 2.4 后端控制平面 (Control Plane)
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| C1 | syncForwardServices 优先使用 Forward.SpeedID | `go-backend/internal/http/handler/control_plane.go` | ✅ 完成 |
-| C2 | 回退到 UserTunnel 的 speed limit | `go-backend/internal/http/handler/control_plane.go` | ✅ 完成 |
-
-### 2.5 前端类型定义 (TypeScript Types)
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| T1 | SpeedLimitApiItem.tunnelId 改为可选 | `vite-frontend/src/api/types.ts` | ✅ 完成 |
-| T2 | ForwardApiItem 添加 speedId 字段 | `vite-frontend/src/api/types.ts` | ✅ 完成 |
-| T3 | ForwardMutationPayload 添加 speedId 字段 | `vite-frontend/src/api/types.ts` | ✅ 完成 |
-| T4 | SpeedLimitMutationPayload.tunnelId 改为可选 | `vite-frontend/src/api/types.ts` | ✅ 完成 |
-
-### 2.6 前端页面组件
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| F1 | SpeedLimitRule 接口更新 | `vite-frontend/src/pages/limit.tsx` | ✅ 完成 |
-| F2 | SpeedLimitForm 接口更新 | `vite-frontend/src/pages/limit.tsx` | ✅ 完成 |
-| F3 | validateForm 移除 tunnelId 必填校验 | `vite-frontend/src/pages/limit.tsx` | ✅ 完成 |
-| F4 | Select 组件改为可选 | `vite-frontend/src/pages/limit.tsx` | ✅ 完成 |
-| F5 | 显示"未绑定"状态 | `vite-frontend/src/pages/limit.tsx` | ✅ 完成 |
-
-### 2.7 编译验证
-
-| 序号 | 任务 | 状态 |
-|------|------|------|
-| B1 | Go 后端编译通过 | ✅ 完成 |
-| B2 | TypeScript 类型检查通过 | ✅ 完成 |
-| B3 | `go test ./...` 全量通过 | ✅ 完成 |
-| B4 | `go test ./tests/contract/... -run SpeedLimit` 通过 | ✅ 完成 |
-
-### 2.8 Forward 独立限速写入链路补全（新增）
-
-| 序号 | 任务 | 文件 | 状态 |
-|------|------|------|------|
-| N1 | forwardCreate 支持接收并校验可选 speedId，写入 Forward.SpeedID | `go-backend/internal/http/handler/mutations.go` | ✅ 完成 |
-| N2 | forwardUpdate 支持更新/清空 speedId，并触发服务重下发 | `go-backend/internal/http/handler/mutations.go` | ✅ 完成 |
-| N3 | CreateForwardTx 支持落库 speed_id | `go-backend/internal/store/repo/repository_mutations.go` | ✅ 完成 |
-| N4 | UpdateForward 支持更新 speed_id | `go-backend/internal/store/repo/repository_mutations.go` | ✅ 完成 |
-| N5 | Forward 页面新增限速选择并透传 speedId | `vite-frontend/src/pages/forward.tsx` | ✅ 完成 |
-| N6 | Forward 相关契约测试补充 speedId 写入/清空断言 | `go-backend/tests/contract/forward_contract_test.go` | ✅ 完成 |
-
----
-
-## 三、优先级说明
-
-限速规则应用优先级:
-1. **Forward.SpeedID** - 转发级别的限速 (最高优先)
-2. **UserTunnel.SpeedID** - 用户隧道权限级别的限速 (回退)
-
----
-
-## 四、数据库兼容性
-
-- SpeedLimit 表: `tunnel_id` 和 `tunnel_name` 字段改为可空 (GORM AutoMigrate 自动处理)
-- Forward 表: 新增 `speed_id` 可空字段 (GORM AutoMigrate 自动处理)
-
----
-
-## 五、验证检查项
-
-### 5.1 功能验证（审计后）
-
-- [x] 创建不限速规则的限速 (不绑定隧道)
-- [x] 创建绑定隧道的限速 (兼容旧逻辑)
-- [x] 编辑限速规则，切换隧道绑定状态
-- [ ] 删除限速规则
-- [ ] 转发列表正确显示 speedId
-
-### 5.2 API 验证（审计后）
-
-- [x] GET /api/speed-limit/list 返回可选 tunnelId
-- [x] POST /api/speed-limit/create 接受可选 tunnelId
-- [x] POST /api/speed-limit/update 接受可选 tunnelId
-- [ ] GET /api/forward/list 返回 speedId
-
-### 5.3 兼容性验证（审计后）
-
-- [x] 现有绑定隧道的限速规则继续正常工作
-- [ ] 现有 UserTunnel 的限速继续正常工作
-- [ ] 备份/恢复功能正常
-
-### 5.4 Forward 独立限速闭环验证（新增）
-
-- [x] POST /api/forward/create 接受 speedId 并写入 `forward.speed_id`
-- [x] POST /api/forward/update 可更新/清空 speedId
-- [x] Forward 表单可选择限速并提交 speedId
-- [ ] `syncForwardServices` 实际使用 Forward.SpeedID 而非仅回退 UserTunnel.SpeedID

README.md

@@ -33,13 +33,13 @@ curl -L https://raw.githubusercontent.com/Sagit-chu/flux-panel/main/install.sh -
 #### 安装特定版本
 从 [Releases](https://github.com/Sagit-chu/flux-panel/releases) 页面复制对应版本的安装命令，脚本会自动安装该版本而非最新版。
 
-面板端（以 2.1.0 为例）：
+面板端（以 2.1.9-beta6 为例）：
 ```bash
-curl -L https://github.com/Sagit-chu/flux-panel/releases/download/2.1.0/panel_install.sh -o panel_install.sh && chmod +x panel_install.sh && ./panel_install.sh
+curl -L https://github.com/Sagit-chu/flux-panel/releases/download/2.1.9-beta6/panel_install.sh -o panel_install.sh && chmod +x panel_install.sh && ./panel_install.sh
 ```
-节点端（以 2.1.0 为例）：
+节点端（以 2.1.9-beta6 为例）：
 ```bash
-curl -L https://github.com/Sagit-chu/flux-panel/releases/download/2.1.0/install.sh -o install.sh && chmod +x install.sh && ./install.sh
+curl -L https://github.com/Sagit-chu/flux-panel/releases/download/2.1.9-beta6/install.sh -o install.sh && chmod +x install.sh && ./install.sh
 ```
 
 #### PostgreSQL 部署（Docker Compose）

go-backend/AGENTS.md

@@ -1,5 +1,10 @@
 # GO BACKEND KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Go-based Admin API for FLVX. Replaced legacy Spring Boot backend.
 **Stack:** Go 1.24, net/http (std lib), GORM + SQLite/PostgreSQL (glebarez/sqlite - CGO-free).

go-backend/internal/health/checker.go

@@ -0,0 +1,402 @@
+package health
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"log"
+	"net"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"go-backend/internal/monitoring"
+	"go-backend/internal/store/model"
+	"go-backend/internal/store/repo"
+	"go-backend/internal/ws"
+)
+
+type nodeCommander interface {
+	SendCommand(nodeID int64, cmdType string, data interface{}, timeout time.Duration) (ws.CommandResult, error)
+}
+
+const serviceMonitorReportInterval = 30 * time.Second // DB write interval per monitor
+
+type Checker struct {
+	repo      *repo.Repository
+	commander nodeCommander
+	lastRun   map[int64]int64
+	inFlight  map[int64]struct{}
+
+	// In-memory latest result per monitor (for real-time API reads)
+	latestResults map[int64]*model.ServiceMonitorResult
+	lastDBWrite   map[int64]int64 // last DB write timestamp per monitorID
+
+	mu       sync.RWMutex
+	cancel   context.CancelFunc
+	wg       sync.WaitGroup
+	checking int32 // atomic flag: 1 = runChecks running, 0 = idle
+}
+
+func NewChecker(repo *repo.Repository, commander nodeCommander) *Checker {
+	return &Checker{
+		repo:          repo,
+		commander:     commander,
+		lastRun:       make(map[int64]int64),
+		inFlight:      make(map[int64]struct{}),
+		latestResults: make(map[int64]*model.ServiceMonitorResult),
+		lastDBWrite:   make(map[int64]int64),
+	}
+}
+
+// GetLatestCached returns the in-memory latest results (updated every 1s).
+// Returns nil if no results are cached.
+func (c *Checker) GetLatestCached() []*model.ServiceMonitorResult {
+	c.mu.RLock()
+	defer c.mu.RUnlock()
+	results := make([]*model.ServiceMonitorResult, 0, len(c.latestResults))
+	for _, r := range c.latestResults {
+		results = append(results, r)
+	}
+	return results
+}
+
+func (c *Checker) Start(ctx context.Context) {
+	c.mu.Lock()
+	ctx, cancel := context.WithCancel(ctx)
+	c.cancel = cancel
+	c.mu.Unlock()
+
+	c.runChecks(ctx)
+
+	for {
+		limits := c.loadServiceMonitorLimits()
+		scanInterval := time.Duration(limits.CheckerScanIntervalSec) * time.Second
+		if scanInterval <= 0 {
+			scanInterval = 1 * time.Second
+		}
+
+		timer := time.NewTimer(scanInterval)
+		select {
+		case <-ctx.Done():
+			timer.Stop()
+			return
+		case <-timer.C:
+			c.runChecks(ctx)
+		}
+	}
+}
+
+func (c *Checker) Stop() {
+	c.mu.Lock()
+	if c.cancel != nil {
+		c.cancel()
+	}
+	c.mu.Unlock()
+	c.wg.Wait()
+}
+
+func (c *Checker) RunOnce(m *model.ServiceMonitor) (*model.ServiceMonitorResult, error) {
+	if c == nil {
+		return nil, errors.New("checker not initialized")
+	}
+	if m == nil {
+		return nil, errors.New("monitor is nil")
+	}
+	limits := c.loadServiceMonitorLimits()
+	return c.executeCheck(m, time.Now().UnixMilli(), limits), nil
+}
+
+func (c *Checker) runChecks(ctx context.Context) {
+	// Skip if previous round is still running (interval < timeout guard)
+	if !atomic.CompareAndSwapInt32(&c.checking, 0, 1) {
+		return
+	}
+	defer atomic.StoreInt32(&c.checking, 0)
+
+	if c == nil || c.repo == nil {
+		return
+	}
+
+	limits := c.loadServiceMonitorLimits()
+	monitors, err := c.repo.ListEnabledServiceMonitors()
+	if err != nil {
+		log.Printf("service monitor scheduler failed op=list_enabled err=%v", err)
+		return
+	}
+	if len(monitors) == 0 {
+		return
+	}
+
+	// Use persisted result timestamps to avoid restart bursts.
+	latest, err := c.repo.GetLatestServiceMonitorResults()
+	if err != nil {
+		log.Printf("service monitor scheduler failed op=get_latest_results err=%v", err)
+		latest = nil
+	}
+	persistedLast := make(map[int64]int64, len(latest))
+	for _, r := range latest {
+		if r.MonitorID <= 0 || r.Timestamp <= 0 {
+			continue
+		}
+		persistedLast[r.MonitorID] = r.Timestamp
+	}
+
+	now := time.Now().UnixMilli()
+	due := make([]model.ServiceMonitor, 0, len(monitors))
+	for _, m := range monitors {
+		select {
+		case <-ctx.Done():
+			return
+		default:
+		}
+
+		intervalSec := m.IntervalSec
+		if intervalSec <= 0 {
+			intervalSec = limits.DefaultIntervalSec
+		}
+		if intervalSec < limits.MinIntervalSec {
+			intervalSec = limits.MinIntervalSec
+		}
+		intervalMs := int64(intervalSec) * 1000
+
+		c.mu.Lock()
+		if _, ok := c.inFlight[m.ID]; ok {
+			c.mu.Unlock()
+			continue
+		}
+
+		lastSeen := persistedLast[m.ID]
+		if v := c.lastRun[m.ID]; v > lastSeen {
+			lastSeen = v
+		}
+		if lastSeen > 0 && intervalMs > 0 && now-lastSeen < intervalMs {
+			c.mu.Unlock()
+			continue
+		}
+
+		c.inFlight[m.ID] = struct{}{}
+		// Use now as a best-effort guard against overlapping scans; the final
+		// timestamp is updated again when the result is persisted.
+		c.lastRun[m.ID] = now
+		c.mu.Unlock()
+
+		due = append(due, m)
+	}
+	if len(due) == 0 {
+		return
+	}
+
+	workerLimit := limits.WorkerLimit
+	if workerLimit <= 0 {
+		workerLimit = 1
+	}
+	if workerLimit > len(due) {
+		workerLimit = len(due)
+	}
+
+	jobs := make(chan model.ServiceMonitor, len(due))
+	for _, m := range due {
+		jobs <- m
+	}
+	close(jobs)
+
+	reportIntervalMs := int64(serviceMonitorReportInterval / time.Millisecond)
+
+	for i := 0; i < workerLimit; i++ {
+		c.wg.Add(1)
+		go func() {
+			defer c.wg.Done()
+			for {
+				select {
+				case <-ctx.Done():
+					return
+				case m, ok := <-jobs:
+					if !ok {
+						return
+					}
+					ts := time.Now().UnixMilli()
+					result := c.executeCheck(&m, ts, limits)
+
+					// Always update in-memory cache for real-time reads
+					c.mu.Lock()
+					c.latestResults[m.ID] = result
+					c.lastRun[m.ID] = result.Timestamp
+					delete(c.inFlight, m.ID)
+
+					// Only write to DB every 30s per monitor
+					lastWrite := c.lastDBWrite[m.ID]
+					writeToDB := ts-lastWrite >= reportIntervalMs
+					if writeToDB {
+						c.lastDBWrite[m.ID] = ts
+					}
+					c.mu.Unlock()
+
+					if writeToDB {
+						if err := c.repo.InsertServiceMonitorResult(result); err != nil {
+							log.Printf("monitoring write failed op=service_monitor_result.insert monitor_id=%d err=%v", result.MonitorID, err)
+						}
+					}
+				}
+			}
+		}()
+	}
+}
+
+func (c *Checker) executeCheck(m *model.ServiceMonitor, timestamp int64, limits monitoring.ServiceMonitorLimits) *model.ServiceMonitorResult {
+	result := &model.ServiceMonitorResult{
+		MonitorID: m.ID,
+		NodeID:    m.NodeID,
+		Timestamp: timestamp,
+	}
+
+	timeoutSec := m.TimeoutSec
+	if timeoutSec <= 0 {
+		timeoutSec = limits.DefaultTimeoutSec
+	}
+	if timeoutSec < limits.MinTimeoutSec {
+		timeoutSec = limits.MinTimeoutSec
+	}
+	if timeoutSec > limits.MaxTimeoutSec {
+		timeoutSec = limits.MaxTimeoutSec
+	}
+
+	timeout := time.Duration(timeoutSec) * time.Second
+
+	// When nodeId is set, run checks on the specified node.
+	if m.NodeID > 0 {
+		c.checkOnNode(m, timeoutSec, timeout, result)
+		return result
+	}
+
+	switch strings.ToLower(strings.TrimSpace(m.Type)) {
+	case "tcp":
+		c.checkTCP(m.Target, timeout, result)
+	case "icmp":
+		result.Success = 0
+		result.ErrorMessage = "ICMP 监控必须指定执行节点"
+	default:
+		result.Success = 0
+		result.ErrorMessage = fmt.Sprintf("不支持的检查类型: %s", m.Type)
+	}
+
+	return result
+}
+
+func (c *Checker) loadServiceMonitorLimits() monitoring.ServiceMonitorLimits {
+	defaults := monitoring.DefaultServiceMonitorLimits()
+	if c == nil || c.repo == nil {
+		return defaults
+	}
+	cfg, err := c.repo.GetConfigsByNames([]string{
+		monitoring.ConfigServiceMonitorCheckerScanIntervalSec,
+		monitoring.ConfigServiceMonitorWorkerLimit,
+		monitoring.ConfigServiceMonitorMinIntervalSec,
+		monitoring.ConfigServiceMonitorDefaultIntervalSec,
+		monitoring.ConfigServiceMonitorMinTimeoutSec,
+		monitoring.ConfigServiceMonitorDefaultTimeoutSec,
+		monitoring.ConfigServiceMonitorMaxTimeoutSec,
+	})
+	if err != nil {
+		return defaults
+	}
+	return monitoring.ServiceMonitorLimitsFromConfigMap(cfg)
+}
+
+type serviceMonitorCheckRequest struct {
+	MonitorID  int64  `json:"monitorId"`
+	Type       string `json:"type"`
+	Target     string `json:"target"`
+	TimeoutSec int    `json:"timeoutSec"`
+}
+
+func (c *Checker) checkOnNode(m *model.ServiceMonitor, timeoutSec int, timeout time.Duration, result *model.ServiceMonitorResult) {
+	if c == nil || m == nil || result == nil {
+		return
+	}
+	if c.commander == nil {
+		result.Success = 0
+		result.ErrorMessage = "节点检查不可用"
+		return
+	}
+
+	checkType := strings.ToLower(strings.TrimSpace(m.Type))
+	if checkType != "tcp" && checkType != "icmp" {
+		result.Success = 0
+		result.ErrorMessage = fmt.Sprintf("不支持的检查类型: %s", m.Type)
+		return
+	}
+	if strings.TrimSpace(m.Target) == "" {
+		result.Success = 0
+		result.ErrorMessage = "检查目标为空"
+		return
+	}
+
+	req := serviceMonitorCheckRequest{
+		MonitorID:  m.ID,
+		Type:       checkType,
+		Target:     m.Target,
+		TimeoutSec: timeoutSec,
+	}
+
+	cmdTimeout := timeout
+	if cmdTimeout < 2*time.Second {
+		cmdTimeout = 2 * time.Second
+	}
+	cmdTimeout = cmdTimeout + 2*time.Second
+
+	cmdRes, err := c.commander.SendCommand(m.NodeID, "ServiceMonitorCheck", req, cmdTimeout)
+	if err != nil {
+		result.Success = 0
+		result.ErrorMessage = err.Error()
+		return
+	}
+	if cmdRes.Data == nil {
+		result.Success = 0
+		result.ErrorMessage = "节点返回为空"
+		return
+	}
+
+	if v, ok := cmdRes.Data["success"]; ok {
+		if b, ok := v.(bool); ok {
+			if b {
+				result.Success = 1
+			} else {
+				result.Success = 0
+			}
+		}
+	}
+	if v, ok := cmdRes.Data["latencyMs"]; ok {
+		if f, ok := v.(float64); ok {
+			result.LatencyMs = f
+		}
+	}
+	if v, ok := cmdRes.Data["statusCode"]; ok {
+		if f, ok := v.(float64); ok {
+			result.StatusCode = int(f)
+		}
+	}
+	if v, ok := cmdRes.Data["errorMessage"]; ok {
+		if s, ok := v.(string); ok {
+			result.ErrorMessage = s
+		}
+	}
+}
+
+func (c *Checker) checkTCP(target string, timeout time.Duration, result *model.ServiceMonitorResult) {
+	start := time.Now()
+
+	conn, err := net.DialTimeout("tcp", target, timeout)
+	latency := time.Since(start)
+
+	result.LatencyMs = float64(latency.Milliseconds())
+
+	if err != nil {
+		result.Success = 0
+		result.ErrorMessage = err.Error()
+		return
+	}
+	_ = conn.Close()
+	result.Success = 1
+}

go-backend/internal/health/checker_test.go

@@ -0,0 +1,477 @@
+package health
+
+import (
+	"context"
+	"net"
+	"testing"
+	"time"
+
+	"go-backend/internal/monitoring"
+	"go-backend/internal/store/model"
+	"go-backend/internal/store/repo"
+	"go-backend/internal/ws"
+)
+
+type fakeCommander struct {
+	lastNodeID int64
+	lastType   string
+	lastData   interface{}
+	res        ws.CommandResult
+	err        error
+}
+
+type delayedCommander struct {
+	delayByMonitorID map[int64]time.Duration
+}
+
+func (d *delayedCommander) SendCommand(nodeID int64, cmdType string, data interface{}, _ time.Duration) (ws.CommandResult, error) {
+	_ = nodeID
+	_ = cmdType
+	if req, ok := data.(serviceMonitorCheckRequest); ok {
+		if delay := d.delayByMonitorID[req.MonitorID]; delay > 0 {
+			time.Sleep(delay)
+		}
+	}
+	return ws.CommandResult{
+		Success: true,
+		Data: map[string]interface{}{
+			"success":   true,
+			"latencyMs": float64(1),
+		},
+	}, nil
+}
+
+func (f *fakeCommander) SendCommand(nodeID int64, cmdType string, data interface{}, _ time.Duration) (ws.CommandResult, error) {
+	f.lastNodeID = nodeID
+	f.lastType = cmdType
+	f.lastData = data
+	return f.res, f.err
+}
+
+func TestTCPHealthCheckViaMonitor(t *testing.T) {
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+	defer listener.Close()
+	addr := listener.Addr().String()
+
+	go func() {
+		for {
+			conn, err := listener.Accept()
+			if err != nil {
+				return
+			}
+			conn.Close()
+		}
+	}()
+
+	t.Run("successful tcp check", func(t *testing.T) {
+		checker := NewChecker(nil, nil)
+		limits := checker.loadServiceMonitorLimits()
+		now := time.Now().UnixMilli()
+		monitor := &model.ServiceMonitor{
+			Type:       "tcp",
+			Target:     addr,
+			TimeoutSec: 5,
+		}
+		result := checker.executeCheck(monitor, now, limits)
+		if result.Success != 1 {
+			t.Fatalf("expected success, got error: %s", result.ErrorMessage)
+		}
+		if result.LatencyMs < 0 {
+			t.Fatalf("expected non-negative latency, got %f", result.LatencyMs)
+		}
+	})
+
+	t.Run("failed tcp check - connection refused", func(t *testing.T) {
+		checker := NewChecker(nil, nil)
+		limits := checker.loadServiceMonitorLimits()
+		now := time.Now().UnixMilli()
+		monitor := &model.ServiceMonitor{
+			Type:       "tcp",
+			Target:     "127.0.0.1:1",
+			TimeoutSec: 1,
+		}
+		result := checker.executeCheck(monitor, now, limits)
+		if result.Success == 1 {
+			t.Fatalf("expected failure for connection refused")
+		}
+		if result.ErrorMessage == "" {
+			t.Fatalf("expected error message")
+		}
+	})
+}
+
+func TestCheckerRunChecks(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+	defer listener.Close()
+	tcpAddr := listener.Addr().String()
+
+	go func() {
+		for {
+			conn, err := listener.Accept()
+			if err != nil {
+				return
+			}
+			conn.Close()
+		}
+	}()
+
+	now := time.Now().UnixMilli()
+
+	monitors := []*model.ServiceMonitor{
+		{
+			Name:        "TCP Monitor",
+			Type:        "tcp",
+			Target:      tcpAddr,
+			IntervalSec: 60,
+			TimeoutSec:  5,
+			NodeID:      0,
+			Enabled:     1,
+			CreatedTime: now,
+			UpdatedTime: now,
+		},
+		{
+			Name:        "TCP Monitor 2",
+			Type:        "tcp",
+			Target:      tcpAddr,
+			IntervalSec: 60,
+			TimeoutSec:  5,
+			NodeID:      0,
+			Enabled:     1,
+			CreatedTime: now,
+			UpdatedTime: now,
+		},
+		{
+			Name:        "Disabled Monitor",
+			Type:        "tcp",
+			Target:      "127.0.0.1:1",
+			IntervalSec: 60,
+			TimeoutSec:  5,
+			NodeID:      0,
+			Enabled:     0,
+			CreatedTime: now,
+			UpdatedTime: now,
+		},
+	}
+
+	for _, m := range monitors {
+		if err := r.CreateServiceMonitor(m); err != nil {
+			t.Fatalf("create monitor: %v", err)
+		}
+	}
+
+	monitors[2].Enabled = 0
+	if err := r.UpdateServiceMonitor(monitors[2]); err != nil {
+		t.Fatalf("update disabled monitor: %v", err)
+	}
+
+	enabledMonitors, err := r.ListEnabledServiceMonitors()
+	if err != nil {
+		t.Fatalf("list enabled monitors: %v", err)
+	}
+	if len(enabledMonitors) != 2 {
+		t.Fatalf("expected 2 enabled monitors, got %d", len(enabledMonitors))
+	}
+
+	checker := NewChecker(r, nil)
+
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+
+	go checker.Start(ctx)
+	time.Sleep(500 * time.Millisecond)
+
+	results, err := r.GetServiceMonitorResults(monitors[0].ID, 10)
+	if err != nil {
+		t.Fatalf("get tcp results: %v", err)
+	}
+	if len(results) == 0 {
+		t.Fatalf("expected at least one result for tcp monitor")
+	}
+	for _, res := range results {
+		if res.Success != 1 {
+			t.Fatalf("expected success for tcp monitor, got failure: %s", res.ErrorMessage)
+		}
+	}
+
+	results2, err := r.GetServiceMonitorResults(monitors[1].ID, 10)
+	if err != nil {
+		t.Fatalf("get tcp results 2: %v", err)
+	}
+	if len(results2) == 0 {
+		t.Fatalf("expected at least one result for tcp monitor 2")
+	}
+	for _, res := range results2 {
+		if res.Success != 1 {
+			t.Fatalf("expected success for tcp monitor 2, got failure: %s", res.ErrorMessage)
+		}
+	}
+
+	disabledResults, err := r.GetServiceMonitorResults(monitors[2].ID, 10)
+	if err != nil {
+		t.Fatalf("get disabled results: %v", err)
+	}
+	if len(disabledResults) != 0 {
+		t.Fatalf("expected no results for disabled monitor, got %d", len(disabledResults))
+	}
+}
+
+func TestCheckerUnsupportedType(t *testing.T) {
+	checker := NewChecker(nil, nil)
+	limits := checker.loadServiceMonitorLimits()
+	now := time.Now().UnixMilli()
+	monitor := &model.ServiceMonitor{
+		Type:       "http",
+		Target:     "https://example.com",
+		TimeoutSec: 5,
+	}
+	result := checker.executeCheck(monitor, now, limits)
+	if result.Success == 1 {
+		t.Fatalf("expected failure for unsupported type")
+	}
+	if result.ErrorMessage == "" {
+		t.Fatalf("expected error message for unsupported type")
+	}
+}
+
+func TestCheckerDefaultTimeout(t *testing.T) {
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+	defer listener.Close()
+	addr := listener.Addr().String()
+
+	go func() {
+		for {
+			conn, err := listener.Accept()
+			if err != nil {
+				return
+			}
+			conn.Close()
+		}
+	}()
+
+	checker := NewChecker(nil, nil)
+	limits := checker.loadServiceMonitorLimits()
+	now := time.Now().UnixMilli()
+	monitor := &model.ServiceMonitor{
+		Type:       "tcp",
+		Target:     addr,
+		TimeoutSec: 0,
+	}
+	result := checker.executeCheck(monitor, now, limits)
+	if result.Success != 1 {
+		t.Fatalf("expected success with default timeout, got error: %s", result.ErrorMessage)
+	}
+}
+
+func TestCheckerStop(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	listener, err := net.Listen("tcp", "127.0.0.1:0")
+	if err != nil {
+		t.Fatalf("listen: %v", err)
+	}
+	defer listener.Close()
+
+	go func() {
+		for {
+			conn, err := listener.Accept()
+			if err != nil {
+				return
+			}
+			conn.Close()
+		}
+	}()
+
+	now := time.Now().UnixMilli()
+	monitor := &model.ServiceMonitor{
+		Name:        "Test Monitor",
+		Type:        "tcp",
+		Target:      listener.Addr().String(),
+		IntervalSec: 60,
+		TimeoutSec:  5,
+		NodeID:      0,
+		Enabled:     1,
+		CreatedTime: now,
+		UpdatedTime: now,
+	}
+	if err := r.CreateServiceMonitor(monitor); err != nil {
+		t.Fatalf("create monitor: %v", err)
+	}
+
+	checker := NewChecker(r, nil)
+
+	ctx := context.Background()
+	go checker.Start(ctx)
+
+	time.Sleep(100 * time.Millisecond)
+	checker.Stop()
+
+	results, err := r.GetServiceMonitorResults(monitor.ID, 10)
+	if err != nil {
+		t.Fatalf("get results: %v", err)
+	}
+	if len(results) == 0 {
+		t.Fatalf("expected at least one result before stop")
+	}
+}
+
+func TestCheckerRunsOnNodeWhenNodeIDSet(t *testing.T) {
+	fake := &fakeCommander{
+		res: ws.CommandResult{
+			Success: true,
+			Data: map[string]interface{}{
+				"success":      false,
+				"latencyMs":    float64(12),
+				"errorMessage": "unreachable",
+			},
+		},
+	}
+	checker := NewChecker(nil, fake)
+	limits := checker.loadServiceMonitorLimits()
+	now := time.Now().UnixMilli()
+	monitor := &model.ServiceMonitor{
+		ID:         99,
+		Type:       "icmp",
+		Target:     "8.8.8.8",
+		TimeoutSec: 2,
+		NodeID:     123,
+	}
+	res := checker.executeCheck(monitor, now, limits)
+	if fake.lastNodeID != 123 {
+		t.Fatalf("expected command to be sent to node 123, got %d", fake.lastNodeID)
+	}
+	if fake.lastType != "ServiceMonitorCheck" {
+		t.Fatalf("expected ServiceMonitorCheck command, got %s", fake.lastType)
+	}
+	if res.Success != 0 {
+		t.Fatalf("expected failed result from node check")
+	}
+	if res.ErrorMessage != "unreachable" {
+		t.Fatalf("expected errorMessage unreachable, got %q", res.ErrorMessage)
+	}
+}
+
+func TestCheckerDoesNotBurstOnRestartWhenRecentResultsExist(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	now := time.Now().UnixMilli()
+	monitor := &model.ServiceMonitor{
+		Name:        "recent-monitor",
+		Type:        "tcp",
+		Target:      "127.0.0.1:1",
+		IntervalSec: 60,
+		TimeoutSec:  1,
+		NodeID:      0,
+		Enabled:     1,
+		CreatedTime: now,
+		UpdatedTime: now,
+	}
+	if err := r.CreateServiceMonitor(monitor); err != nil {
+		t.Fatalf("create monitor: %v", err)
+	}
+	if err := r.InsertServiceMonitorResult(&model.ServiceMonitorResult{
+		MonitorID: monitor.ID,
+		NodeID:    0,
+		Timestamp: now - 10_000,
+		Success:   1,
+	}); err != nil {
+		t.Fatalf("seed recent result: %v", err)
+	}
+
+	checker := NewChecker(r, nil)
+	ctx, cancel := context.WithCancel(context.Background())
+	go checker.Start(ctx)
+	// Give the initial scan a chance to run.
+	time.Sleep(200 * time.Millisecond)
+	cancel()
+	checker.Stop()
+
+	results, err := r.GetServiceMonitorResults(monitor.ID, 10)
+	if err != nil {
+		t.Fatalf("get results: %v", err)
+	}
+	if len(results) != 1 {
+		t.Fatalf("expected no immediate rerun (1 result), got %d", len(results))
+	}
+}
+
+func TestCheckerConcurrencyPreventsSlowMonitorBlockingOthers(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	now := time.Now().UnixMilli()
+	// Force worker limit to at least 2 for this test.
+	_ = r.UpsertConfig(monitoring.ConfigServiceMonitorWorkerLimit, "2", now)
+
+	slow := &model.ServiceMonitor{
+		Name:        "slow",
+		Type:        "icmp",
+		Target:      "8.8.8.8",
+		IntervalSec: 60,
+		TimeoutSec:  1,
+		NodeID:      123,
+		Enabled:     1,
+		CreatedTime: now,
+		UpdatedTime: now,
+	}
+	if err := r.CreateServiceMonitor(slow); err != nil {
+		t.Fatalf("create slow monitor: %v", err)
+	}
+	fast := &model.ServiceMonitor{
+		Name:        "fast",
+		Type:        "icmp",
+		Target:      "1.1.1.1",
+		IntervalSec: 60,
+		TimeoutSec:  1,
+		NodeID:      123,
+		Enabled:     1,
+		CreatedTime: now,
+		UpdatedTime: now,
+	}
+	if err := r.CreateServiceMonitor(fast); err != nil {
+		t.Fatalf("create fast monitor: %v", err)
+	}
+
+	cmd := &delayedCommander{delayByMonitorID: map[int64]time.Duration{slow.ID: 800 * time.Millisecond}}
+	checker := NewChecker(r, cmd)
+	ctx, cancel := context.WithCancel(context.Background())
+	go checker.Start(ctx)
+
+	// Fast monitor should complete even while slow one is still running.
+	time.Sleep(250 * time.Millisecond)
+	results, err := r.GetServiceMonitorResults(fast.ID, 10)
+	if err != nil {
+		t.Fatalf("get fast results: %v", err)
+	}
+	if len(results) == 0 {
+		t.Fatalf("expected fast monitor to have results without waiting for slow")
+	}
+
+	cancel()
+	checker.Stop()
+}

go-backend/internal/http/handler/AGENTS.md

@@ -1,6 +1,9 @@
 # BACKEND HTTP HANDLER KNOWLEDGE BASE
 
-**Generated:** Thu Feb 26 2026
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
 
 ## OVERVIEW
 HTTP request handlers for FLVX Admin API. Core business logic layer.

go-backend/internal/http/handler/control_plane.go

@@ -280,6 +280,16 @@ func (h *Handler) syncForwardServicesWithWarnings(forward *forwardRecord, method
 	for _, fp := range ports {
 		if limiterID != nil && speed != nil {
 			if err := h.ensureLimiterOnNode(fp.NodeID, *limiterID, *speed); err != nil {
+				// If the limiter push fails because the node is offline, skip it with a warning
+				if isNodeOfflineOrTimeoutError(err) {
+					node, _ := h.getNodeRecord(fp.NodeID)
+					nodeName := fmt.Sprintf("%d", fp.NodeID)
+					if node != nil && strings.TrimSpace(node.Name) != "" {
+						nodeName = strings.TrimSpace(node.Name)
+					}
+					warnings = append(warnings, fmt.Sprintf("节点 %s 不在线，已跳过下发", nodeName))
+					continue
+				}
 				return nil, err
 			}
 		}
@@ -308,6 +318,12 @@ func (h *Handler) syncForwardServicesWithWarnings(forward *forwardRecord, method
 				warnings = append(warnings, warning)
 			}
 		}
+		// When a node is offline, skip it with a warning instead of failing.
+		// This lets users modify forward rules even when some entry nodes are down.
+		if err != nil && isNodeOfflineOrTimeoutError(err) {
+			warnings = append(warnings, fmt.Sprintf("节点 %s 不在线，已跳过下发", node.Name))
+			continue
+		}
 		if err != nil {
 			return warnings, fmt.Errorf("节点 %s 下发失败: %w", node.Name, err)
 		}
@@ -1561,10 +1577,11 @@ func buildForwardServiceConfigs(baseName string, forward *forwardRecord, tunnel
 		}
 		var serviceAddr string
 		if bindIP != "" {
-			if strings.Contains(bindIP, ":") {
-				serviceAddr = processServerAddress(bindIP)
+			trimmedBindIP := strings.TrimSpace(bindIP)
+			if _, _, err := net.SplitHostPort(trimmedBindIP); err == nil {
+				serviceAddr = processServerAddress(trimmedBindIP)
 			} else {
-				serviceAddr = processServerAddress(fmt.Sprintf("%s:%d", bindIP, port))
+				serviceAddr = processServerAddress(net.JoinHostPort(strings.Trim(trimmedBindIP, "[]"), strconv.Itoa(port)))
 			}
 		} else {
 			serviceAddr = processServerAddress(fmt.Sprintf("%s:%d", listenerAddr, port))

go-backend/internal/http/handler/control_plane_test.go

@@ -421,6 +421,63 @@ func TestBuildForwardServiceConfigs_BindIPAlreadyContainsPort(t *testing.T) {
 	}
 }
 
+func TestBuildForwardServiceConfigs_IPv6BindIP(t *testing.T) {
+	tests := []struct {
+		name     string
+		bindIP   string
+		port     int
+		wantAddr string
+	}{
+		{
+			name:     "pure ipv6 without port",
+			bindIP:   "2001:db8::1",
+			port:     22000,
+			wantAddr: "[2001:db8::1]:22000",
+		},
+		{
+			name:     "bracketed ipv6 without port",
+			bindIP:   "[2001:db8::2]",
+			port:     22001,
+			wantAddr: "[2001:db8::2]:22001",
+		},
+		{
+			name:     "bracketed ipv6 with port",
+			bindIP:   "[2001:db8::3]:8080",
+			port:     55555,
+			wantAddr: "[2001:db8::3]:8080",
+		},
+		{
+			name:     "ipv6 link-local with zone",
+			bindIP:   "fe80::1%eth0",
+			port:     22002,
+			wantAddr: "[fe80::1%eth0]:22002",
+		},
+		{
+			name:     "ipv6 localhost",
+			bindIP:   "::1",
+			port:     22003,
+			wantAddr: "[::1]:22003",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			forward := &forwardRecord{RemoteAddr: "1.2.3.4:80", Strategy: "fifo", TunnelID: 7}
+			node := &nodeRecord{TCPListenAddr: "[::]", UDPListenAddr: "[::]"}
+			services := buildForwardServiceConfigs("1_2_0", forward, nil, node, tt.port, tt.bindIP, nil, false)
+			if len(services) != 2 {
+				t.Fatalf("expected 2 services, got %d", len(services))
+			}
+			for _, svc := range services {
+				addr, _ := svc["addr"].(string)
+				if addr != tt.wantAddr {
+					t.Fatalf("expected addr %q, got %q", tt.wantAddr, addr)
+				}
+			}
+		})
+	}
+}
+
 func TestProcessServerAddress_StripsURLSchemeAndPath(t *testing.T) {
 	tests := []struct {
 		name string

go-backend/internal/http/handler/dual_stack_test.go

@@ -345,21 +345,39 @@ func TestSelectTunnelDialHost_V6Only_PreferV4Fallback(t *testing.T) {
 	}
 }
 
-func TestSelectTunnelDialHost_Incompatible(t *testing.T) {
+func TestSelectTunnelDialHost_CrossVersion_V4ToV6(t *testing.T) {
+	// v4-only -> v6-only: 跨版本支持，应成功返回 v6 地址
 	from := v4OnlyNode("from", "10.0.0.1")
 	to := v6OnlyNode("to", "2001:db8::2")
-	_, err := selectTunnelDialHost(from, to, "", "")
-	if err == nil {
-		t.Fatal("expected error for incompatible nodes (v4-only -> v6-only)")
+	host, err := selectTunnelDialHost(from, to, "", "")
+	if err != nil {
+		t.Fatalf("unexpected error for cross-version (v4-only -> v6-only): %v", err)
+	}
+	if host != "2001:db8::2" {
+		t.Fatalf("expected v6 address for cross-version, got %q", host)
 	}
 }
 
-func TestSelectTunnelDialHost_Incompatible_Reverse(t *testing.T) {
+func TestSelectTunnelDialHost_CrossVersion_V6ToV4(t *testing.T) {
+	// v6-only -> v4-only: 跨版本支持，应成功返回 v4 地址
 	from := v6OnlyNode("from", "2001:db8::1")
 	to := v4OnlyNode("to", "10.0.0.2")
+	host, err := selectTunnelDialHost(from, to, "", "")
+	if err != nil {
+		t.Fatalf("unexpected error for cross-version (v6-only -> v4-only): %v", err)
+	}
+	if host != "10.0.0.2" {
+		t.Fatalf("expected v4 address for cross-version, got %q", host)
+	}
+}
+
+func TestSelectTunnelDialHost_TrulyIncompatible(t *testing.T) {
+	// 真正不兼容：两个节点都没有任何 IP
+	from := &nodeRecord{Name: "empty-from", ServerIPv4: "", ServerIPv6: "", ServerIP: ""}
+	to := &nodeRecord{Name: "empty-to", ServerIPv4: "", ServerIPv6: "", ServerIP: ""}
 	_, err := selectTunnelDialHost(from, to, "", "")
 	if err == nil {
-		t.Fatal("expected error for incompatible nodes (v6-only -> v4-only)")
+		t.Fatal("expected error for nodes with no IP addresses")
 	}
 }
 

go-backend/internal/http/handler/flow_policy.go

@@ -22,6 +22,7 @@ type userTunnelPolicy struct {
 	OutFlow  int64
 	ExpTime  int64
 	Status   int
+	Num      int
 }
 
 type gostConfigSnapshot struct {
@@ -363,6 +364,16 @@ func (h *Handler) ensureUserTunnelForwardAllowed(userID int64, tunnelID int64, n
 		return err
 	}
 
+	if user.Num > 0 {
+		currentForwardCount, err := h.repo.CountActiveForwardsByUser(userID)
+		if err != nil {
+			return err
+		}
+		if currentForwardCount >= int64(user.Num) {
+			return errors.New("转发数量已达上限")
+		}
+	}
+
 	userTunnelID, _, _, err := h.resolveUserTunnelAndLimiter(userID, tunnelID)
 	if err != nil {
 		return err
@@ -392,6 +403,16 @@ func (h *Handler) ensureUserTunnelForwardAllowed(userID int64, tunnelID int64, n
 		return errors.New("该隧道流量已超额，禁止开启转发")
 	}
 
+	if policy.Num > 0 {
+		currentTunnelForwardCount, err := h.repo.CountActiveForwardsByUserTunnel(userID, tunnelID)
+		if err != nil {
+			return err
+		}
+		if currentTunnelForwardCount >= int64(policy.Num) {
+			return errors.New("该隧道转发数量已达上限")
+		}
+	}
+
 	return nil
 }
 
@@ -442,7 +463,7 @@ func (h *Handler) getUserTunnelPolicy(userTunnelID int64) (*userTunnelPolicy, er
 	return &userTunnelPolicy{
 		ID: ut.ID, UserID: ut.UserID, TunnelID: ut.TunnelID,
 		Flow: ut.Flow, InFlow: ut.InFlow, OutFlow: ut.OutFlow,
-		ExpTime: ut.ExpTime, Status: ut.Status,
+		ExpTime: ut.ExpTime, Status: ut.Status, Num: ut.Num,
 	}, nil
 }
 

go-backend/internal/http/handler/handler.go

@@ -16,17 +16,21 @@ import (
 	"time"
 
 	"go-backend/internal/auth"
+	"go-backend/internal/health"
 	"go-backend/internal/http/middleware"
 	"go-backend/internal/http/response"
+	"go-backend/internal/metrics"
 	"go-backend/internal/security"
 	"go-backend/internal/store/repo"
 	"go-backend/internal/ws"
 )
 
 type Handler struct {
-	repo      *repo.Repository
-	jwtSecret string
-	wsServer  *ws.Server
+	repo        *repo.Repository
+	jwtSecret   string
+	wsServer    *ws.Server
+	metrics     *metrics.IngestionService
+	healthCheck *health.Checker
 
 	captchaMu     sync.Mutex
 	captchaTokens map[string]int64
@@ -38,8 +42,12 @@ type Handler struct {
 
 	upgradeMu              sync.Mutex
 	pendingUpgradeRedeploy map[int64]struct{}
+
+	qualityProber *tunnelQualityProber
 }
 
+const monitorTunnelQualityEnabledConfigKey = "monitor_tunnel_quality_enabled"
+
 type loginRequest struct {
 	Username  string `json:"username"`
 	Password  string `json:"password"`
@@ -83,10 +91,32 @@ func New(repo *repo.Repository, jwtSecret string) *Handler {
 		repo:                   repo,
 		jwtSecret:              jwtSecret,
 		wsServer:               ws.NewServer(repo, jwtSecret),
+		metrics:                metrics.NewIngestionService(repo),
+		healthCheck:            nil,
 		captchaTokens:          make(map[string]int64),
 		pendingUpgradeRedeploy: make(map[int64]struct{}),
 	}
+	h.healthCheck = health.NewChecker(repo, h.wsServer)
+	h.qualityProber = newTunnelQualityProber(h)
 	h.wsServer.SetNodeOnlineHook(h.onNodeOnline)
+	h.wsServer.SetNodeMetricHook(func(nodeID int64, info ws.SystemInfo) {
+		metricInfo := metrics.SystemInfo{
+			Uptime:           info.Uptime,
+			BytesReceived:    info.BytesReceived,
+			BytesTransmitted: info.BytesTransmitted,
+			CPUUsage:         info.CPUUsage,
+			MemoryUsage:      info.MemoryUsage,
+			DiskUsage:        info.DiskUsage,
+			Load1:            info.Load1,
+			Load5:            info.Load5,
+			Load15:           info.Load15,
+			TCPConns:         info.TCPConns,
+			UDPConns:         info.UDPConns,
+			NetInSpeed:       info.NetInSpeed,
+			NetOutSpeed:      info.NetOutSpeed,
+		}
+		h.metrics.RecordNodeMetric(nodeID, metricInfo)
+	})
 	return h
 }
 
@@ -200,6 +230,24 @@ func (h *Handler) Register(mux *http.ServeMux) {
 	mux.HandleFunc("/api/v1/announcement/get", h.getAnnouncement)
 	mux.HandleFunc("/api/v1/announcement/update", h.updateAnnouncement)
 
+	mux.HandleFunc("/api/v1/monitor/access", h.monitorAccessHandler)
+	mux.HandleFunc("/api/v1/monitor/nodes/", h.monitorNodeMetricsHandler)
+	mux.HandleFunc("/api/v1/monitor/nodes", h.monitorNodeListHandler)
+	mux.HandleFunc("/api/v1/monitor/tunnels", h.monitorTunnelListHandler)
+	mux.HandleFunc("/api/v1/monitor/tunnels/quality", h.monitorTunnelQualityHandler)
+	mux.HandleFunc("/api/v1/monitor/tunnels/", h.monitorTunnelMetrics)
+	mux.HandleFunc("/api/v1/monitor/services", h.monitorServiceListHandler)
+	mux.HandleFunc("/api/v1/monitor/services/create", h.monitorServiceCreate)
+	mux.HandleFunc("/api/v1/monitor/services/update", h.monitorServiceUpdate)
+	mux.HandleFunc("/api/v1/monitor/services/delete", h.monitorServiceDelete)
+	mux.HandleFunc("/api/v1/monitor/services/run", h.monitorServiceRun)
+	mux.HandleFunc("/api/v1/monitor/services/latest-results", h.monitorServiceLatestResultsHandler)
+	mux.HandleFunc("/api/v1/monitor/services/limits", h.monitorServiceLimitsHandler)
+	mux.HandleFunc("/api/v1/monitor/services/", h.monitorServiceResultsHandler)
+	mux.HandleFunc("/api/v1/monitor/permission/list", h.monitorPermissionList)
+	mux.HandleFunc("/api/v1/monitor/permission/assign", h.monitorPermissionAssign)
+	mux.HandleFunc("/api/v1/monitor/permission/remove", h.monitorPermissionRemove)
+
 	mux.HandleFunc("/flow/test", h.flowTest)
 	mux.HandleFunc("/flow/config", h.flowConfig)
 	mux.HandleFunc("/flow/upload", h.flowUpload)
@@ -728,6 +776,8 @@ func (h *Handler) flowUpload(w http.ResponseWriter, r *http.Request) {
 	if err == nil && strings.TrimSpace(raw) != "" {
 		var items []flowItem
 		if json.Unmarshal([]byte(raw), &items) == nil {
+			nowMs := time.Now().UnixMilli()
+			h.recordTunnelMetricsFromFlowItems(node.ID, items, nowMs)
 			for _, item := range items {
 				h.processFlowItem(node.ID, item)
 			}
@@ -838,11 +888,32 @@ func normalizeAndValidateConfigValue(key, value string) (string, error) {
 		}
 
 		return pngDataURLPrefix + payload, nil
+	case monitorTunnelQualityEnabledConfigKey:
+		normalized := strings.TrimSpace(strings.ToLower(value))
+		switch normalized {
+		case "true", "false":
+			return normalized, nil
+		default:
+			return "", fmt.Errorf("隧道质量检测开关配置值无效")
+		}
 	default:
 		return value, nil
 	}
 }
 
+func (h *Handler) isTunnelQualityMonitoringEnabled() bool {
+	if h == nil || h.repo == nil {
+		return true
+	}
+
+	cfg, err := h.repo.GetConfigByName(monitorTunnelQualityEnabledConfigKey)
+	if err != nil || cfg == nil {
+		return true
+	}
+
+	return strings.TrimSpace(strings.ToLower(cfg.Value)) != "false"
+}
+
 func (h *Handler) userPackage(w http.ResponseWriter, r *http.Request) {
 	if r.Method != http.MethodPost {
 		response.WriteJSON(w, response.ErrDefault("请求失败"))
@@ -1190,6 +1261,21 @@ func nullableNullInt64(v sql.NullInt64) interface{} {
 	return nil
 }
 
+// flowCryptoCache caches AES crypto instances by secret to avoid per-request SHA256+GCM init.
+var flowCryptoCache sync.Map
+
+func getOrCreateFlowCrypto(secret string) *security.AESCrypto {
+	if v, ok := flowCryptoCache.Load(secret); ok {
+		return v.(*security.AESCrypto)
+	}
+	c, err := security.NewAESCrypto(secret)
+	if err != nil {
+		return nil
+	}
+	flowCryptoCache.Store(secret, c)
+	return c
+}
+
 func readAndDecryptFlowBody(body io.ReadCloser, secret string) (string, error) {
 	defer body.Close()
 	raw, err := io.ReadAll(body)
@@ -1210,8 +1296,8 @@ func readAndDecryptFlowBody(body io.ReadCloser, secret string) (string, error) {
 		return text, nil
 	}
 
-	crypto, err := security.NewAESCrypto(secret)
-	if err != nil {
+	crypto := getOrCreateFlowCrypto(secret)
+	if crypto == nil {
 		return text, nil
 	}
 	plain, err := crypto.Decrypt(wrap.Data)

go-backend/internal/http/handler/jobs.go

@@ -18,12 +18,15 @@ func (h *Handler) StartBackgroundJobs() {
 	ctx, cancel := context.WithCancel(context.Background())
 	h.jobsCancel = cancel
 	h.jobsStarted = true
-	h.jobsWG.Add(3)
+	h.jobsWG.Add(6)
 	h.jobsMu.Unlock()
 
 	go h.runHourlyStatsLoop(ctx)
 	go h.runDailyMaintenanceLoop(ctx)
 	go h.runNodeRenewalCycleLoop(ctx)
+	go h.runMetricsIngestion(ctx)
+	go h.runHealthChecks(ctx)
+	go h.runTunnelQualityProber(ctx)
 }
 
 func (h *Handler) StopBackgroundJobs() {
@@ -47,6 +50,29 @@ func (h *Handler) StopBackgroundJobs() {
 	h.jobsWG.Wait()
 }
 
+func (h *Handler) runMetricsIngestion(ctx context.Context) {
+	defer h.jobsWG.Done()
+	if h.metrics != nil {
+		h.metrics.Start(ctx)
+	}
+}
+
+func (h *Handler) runHealthChecks(ctx context.Context) {
+	defer h.jobsWG.Done()
+	if h.healthCheck != nil {
+		h.healthCheck.Start(ctx)
+	}
+}
+
+func (h *Handler) runTunnelQualityProber(ctx context.Context) {
+	defer h.jobsWG.Done()
+	if h == nil || h.qualityProber == nil || !h.isTunnelQualityMonitoringEnabled() {
+		return
+	}
+
+	h.qualityProber.Start(ctx)
+}
+
 func (h *Handler) runHourlyStatsLoop(ctx context.Context) {
 	defer h.jobsWG.Done()
 

go-backend/internal/http/handler/monitoring.go

@@ -0,0 +1,940 @@
+package handler
+
+import (
+	"log"
+	"net/http"
+	"strconv"
+	"strings"
+	"time"
+
+	"go-backend/internal/http/response"
+	"go-backend/internal/monitoring"
+	"go-backend/internal/store/model"
+)
+
+const (
+	defaultMetricsRangeMs = int64(60 * 60 * 1000)      // 1h
+	maxMetricsRangeMs     = int64(24 * 60 * 60 * 1000) // 24h
+)
+
+func (h *Handler) resolveServiceMonitorLimits() monitoring.ServiceMonitorLimits {
+	defaults := monitoring.DefaultServiceMonitorLimits()
+	if h == nil || h.repo == nil {
+		return defaults
+	}
+	cfg, err := h.repo.GetConfigsByNames([]string{
+		monitoring.ConfigServiceMonitorCheckerScanIntervalSec,
+		monitoring.ConfigServiceMonitorWorkerLimit,
+		monitoring.ConfigServiceMonitorMinIntervalSec,
+		monitoring.ConfigServiceMonitorDefaultIntervalSec,
+		monitoring.ConfigServiceMonitorMinTimeoutSec,
+		monitoring.ConfigServiceMonitorDefaultTimeoutSec,
+		monitoring.ConfigServiceMonitorMaxTimeoutSec,
+	})
+	if err != nil {
+		return defaults
+	}
+	return monitoring.ServiceMonitorLimitsFromConfigMap(cfg)
+}
+
+func (h *Handler) monitorNodeMetricsHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	path := r.URL.Path
+	prefix := "/api/v1/monitor/nodes/"
+	if !strings.HasPrefix(path, prefix) {
+		response.WriteJSON(w, response.ErrDefault("无效的路径"))
+		return
+	}
+
+	rest := strings.TrimPrefix(path, prefix)
+	if strings.HasSuffix(rest, "/metrics/latest") {
+		h.handleNodeMetricsLatest(w, r, strings.TrimSuffix(rest, "/metrics/latest"))
+		return
+	}
+	if strings.HasSuffix(rest, "/metrics") {
+		h.handleNodeMetrics(w, r, strings.TrimSuffix(rest, "/metrics"))
+		return
+	}
+
+	response.WriteJSON(w, response.ErrDefault("无效的路径"))
+}
+
+type monitorNodeListItem struct {
+	ID          int64  `json:"id"`
+	Inx         int    `json:"inx"`
+	Name        string `json:"name"`
+	Status      int    `json:"status"`
+	Version     string `json:"version"`
+	UpdatedTime int64  `json:"updatedTime"`
+}
+
+func (h *Handler) monitorNodeListHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	nodes, err := h.repo.ListMonitorNodes()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	items := make([]monitorNodeListItem, 0, len(nodes))
+	for _, n := range nodes {
+		updated := int64(0)
+		if n.UpdatedTime.Valid {
+			updated = n.UpdatedTime.Int64
+		}
+		items = append(items, monitorNodeListItem{
+			ID:          n.ID,
+			Inx:         n.Inx,
+			Name:        n.Name,
+			Status:      n.Status,
+			Version:     n.Version.String,
+			UpdatedTime: updated,
+		})
+	}
+
+	response.WriteJSON(w, response.OK(items))
+}
+
+type monitorTunnelListItem struct {
+	ID          int64  `json:"id"`
+	Inx         int    `json:"inx"`
+	Name        string `json:"name"`
+	Status      int    `json:"status"`
+	UpdatedTime int64  `json:"updatedTime"`
+}
+
+func (h *Handler) monitorTunnelListHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	tunnels, err := h.repo.ListMonitorTunnels()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	items := make([]monitorTunnelListItem, 0, len(tunnels))
+	for _, t := range tunnels {
+		items = append(items, monitorTunnelListItem{
+			ID:          t.ID,
+			Inx:         t.Inx,
+			Name:        t.Name,
+			Status:      t.Status,
+			UpdatedTime: t.UpdatedTime,
+		})
+	}
+
+	response.WriteJSON(w, response.OK(items))
+}
+
+func (h *Handler) handleNodeMetrics(w http.ResponseWriter, r *http.Request, nodeIDStr string) {
+	nodeID, err := strconv.ParseInt(nodeIDStr, 10, 64)
+	if err != nil || nodeID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的节点ID"))
+		return
+	}
+
+	now := time.Now().UnixMilli()
+	startMs := now - defaultMetricsRangeMs
+	endMs := now
+
+	if s := r.URL.Query().Get("start"); s != "" {
+		if v, err := strconv.ParseInt(s, 10, 64); err == nil {
+			startMs = v
+		}
+	}
+	if e := r.URL.Query().Get("end"); e != "" {
+		if v, err := strconv.ParseInt(e, 10, 64); err == nil {
+			endMs = v
+		}
+	}
+	if startMs <= 0 || endMs <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+		return
+	}
+	if endMs < startMs {
+		response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+		return
+	}
+	if endMs-startMs > maxMetricsRangeMs {
+		response.WriteJSON(w, response.ErrDefault("时间范围过大"))
+		return
+	}
+
+	metrics, err := h.repo.GetNodeMetrics(nodeID, startMs, endMs)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(metrics))
+}
+
+func (h *Handler) handleNodeMetricsLatest(w http.ResponseWriter, _ *http.Request, nodeIDStr string) {
+	nodeID, err := strconv.ParseInt(nodeIDStr, 10, 64)
+	if err != nil || nodeID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的节点ID"))
+		return
+	}
+
+	metric, err := h.repo.GetLatestNodeMetric(nodeID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	if metric == nil {
+		response.WriteJSON(w, response.OK(nil))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(metric))
+}
+
+func (h *Handler) monitorTunnelQualityHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	// Try in-memory cache first
+	if h.qualityProber != nil {
+		items := h.qualityProber.GetAll()
+		if len(items) > 0 {
+			response.WriteJSON(w, response.OK(items))
+			return
+		}
+	}
+
+	// Fallback to database (latest per tunnel)
+	qualities, err := h.repo.GetLatestTunnelQualities()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	snapshots := make([]tunnelQualitySnapshot, 0, len(qualities))
+	for _, q := range qualities {
+		snapshots = append(snapshots, tunnelQualitySnapshot{
+			TunnelID:           q.TunnelID,
+			EntryToExitLatency: q.EntryToExitLatency,
+			ExitToBingLatency:  q.ExitToBingLatency,
+			EntryToExitLoss:    q.EntryToExitLoss,
+			ExitToBingLoss:     q.ExitToBingLoss,
+			Success:            q.Success == 1,
+			ErrorMessage:       q.ErrorMessage,
+			Timestamp:          q.Timestamp,
+			ChainDetails:       q.ChainDetails,
+		})
+	}
+	response.WriteJSON(w, response.OK(snapshots))
+}
+
+// monitorTunnelQualityHistory returns quality probe history for charting.
+// GET /api/v1/monitor/tunnels/{id}/quality?start=...&end=...
+// Mirrors monitorTunnelMetrics / monitorServiceResultsHandler pattern.
+func (h *Handler) monitorTunnelQualityHistory(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	tunnelIDStr := extractPathParam(r.URL.Path, "/api/v1/monitor/tunnels/", "/quality")
+	tunnelID, err := strconv.ParseInt(tunnelIDStr, 10, 64)
+	if err != nil || tunnelID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的隧道ID"))
+		return
+	}
+
+	now := time.Now().UnixMilli()
+	startMs := now - defaultMetricsRangeMs
+	endMs := now
+
+	if s := r.URL.Query().Get("start"); s != "" {
+		if v, err := strconv.ParseInt(s, 10, 64); err == nil {
+			startMs = v
+		}
+	}
+	if e := r.URL.Query().Get("end"); e != "" {
+		if v, err := strconv.ParseInt(e, 10, 64); err == nil {
+			endMs = v
+		}
+	}
+	if startMs <= 0 || endMs <= 0 || endMs < startMs {
+		response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+		return
+	}
+	if endMs-startMs > maxMetricsRangeMs {
+		response.WriteJSON(w, response.ErrDefault("时间范围过大"))
+		return
+	}
+
+	results, err := h.repo.GetTunnelQualityHistory(tunnelID, startMs, endMs)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(results))
+}
+
+func (h *Handler) monitorTunnelMetrics(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	path := r.URL.Path
+	prefix := "/api/v1/monitor/tunnels/"
+	if !strings.HasPrefix(path, prefix) {
+		response.WriteJSON(w, response.ErrDefault("无效的路径"))
+		return
+	}
+
+	rest := strings.TrimPrefix(path, prefix)
+
+	// Route: /api/v1/monitor/tunnels/{id}/quality
+	if strings.HasSuffix(rest, "/quality") {
+		h.monitorTunnelQualityHistory(w, r)
+		return
+	}
+
+	// Route: /api/v1/monitor/tunnels/{id}/metrics (original)
+	tunnelIDStr := extractPathParam(path, prefix, "/metrics")
+	tunnelID, err := strconv.ParseInt(tunnelIDStr, 10, 64)
+	if err != nil || tunnelID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的隧道ID"))
+		return
+	}
+
+	now := time.Now().UnixMilli()
+	startMs := now - defaultMetricsRangeMs
+	endMs := now
+
+	if s := r.URL.Query().Get("start"); s != "" {
+		if v, err := strconv.ParseInt(s, 10, 64); err == nil {
+			startMs = v
+		}
+	}
+	if e := r.URL.Query().Get("end"); e != "" {
+		if v, err := strconv.ParseInt(e, 10, 64); err == nil {
+			endMs = v
+		}
+	}
+	if startMs <= 0 || endMs <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+		return
+	}
+	if endMs < startMs {
+		response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+		return
+	}
+	if endMs-startMs > maxMetricsRangeMs {
+		response.WriteJSON(w, response.ErrDefault("时间范围过大"))
+		return
+	}
+
+	metrics, err := h.repo.GetTunnelMetricsAggregated(tunnelID, startMs, endMs)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(metrics))
+}
+
+func (h *Handler) monitorServiceListHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	monitors, err := h.repo.ListServiceMonitors()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(monitors))
+}
+
+type createServiceMonitorRequest struct {
+	Name        string `json:"name"`
+	Type        string `json:"type"`
+	Target      string `json:"target"`
+	IntervalSec int    `json:"intervalSec"`
+	TimeoutSec  int    `json:"timeoutSec"`
+	NodeID      int64  `json:"nodeId"`
+	Enabled     *int   `json:"enabled"`
+}
+
+func (h *Handler) monitorServiceCreate(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	var req createServiceMonitorRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+
+	name := strings.TrimSpace(req.Name)
+	if name == "" {
+		response.WriteJSON(w, response.ErrDefault("名称不能为空"))
+		return
+	}
+
+	monitorType := strings.ToLower(strings.TrimSpace(req.Type))
+	if monitorType != "tcp" && monitorType != "icmp" {
+		response.WriteJSON(w, response.ErrDefault("类型必须是 tcp 或 icmp"))
+		return
+	}
+
+	target := strings.TrimSpace(req.Target)
+	if target == "" {
+		response.WriteJSON(w, response.ErrDefault("目标地址不能为空"))
+		return
+	}
+
+	limits := h.resolveServiceMonitorLimits()
+
+	intervalSec := req.IntervalSec
+	if intervalSec <= 0 {
+		intervalSec = limits.DefaultIntervalSec
+	}
+	if intervalSec < limits.MinIntervalSec {
+		intervalSec = limits.MinIntervalSec
+	}
+
+	timeoutSec := req.TimeoutSec
+	if timeoutSec <= 0 {
+		timeoutSec = limits.DefaultTimeoutSec
+	}
+	if timeoutSec < limits.MinTimeoutSec {
+		timeoutSec = limits.MinTimeoutSec
+	}
+	if timeoutSec > limits.MaxTimeoutSec {
+		timeoutSec = limits.MaxTimeoutSec
+	}
+
+	enabled := 1
+	if req.Enabled != nil {
+		if *req.Enabled == 0 || *req.Enabled == 1 {
+			enabled = *req.Enabled
+		}
+	}
+
+	now := time.Now().UnixMilli()
+	if req.NodeID > 0 {
+		n, err := h.repo.GetNodeByID(req.NodeID)
+		if err != nil {
+			response.WriteJSON(w, response.Err(-2, err.Error()))
+			return
+		}
+		if n == nil {
+			response.WriteJSON(w, response.ErrDefault("节点不存在"))
+			return
+		}
+	}
+	m := &model.ServiceMonitor{
+		Name:        name,
+		Type:        monitorType,
+		Target:      target,
+		IntervalSec: intervalSec,
+		TimeoutSec:  timeoutSec,
+		NodeID:      req.NodeID,
+		Enabled:     enabled,
+		CreatedTime: now,
+		UpdatedTime: now,
+	}
+	if m.Type == "icmp" && m.NodeID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("ICMP 监控必须选择执行节点"))
+		return
+	}
+	// enabled is already normalized above.
+
+	if err := h.repo.CreateServiceMonitor(m); err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(m))
+}
+
+type updateServiceMonitorRequest struct {
+	ID          int64  `json:"id"`
+	Name        string `json:"name"`
+	Type        string `json:"type"`
+	Target      string `json:"target"`
+	IntervalSec int    `json:"intervalSec"`
+	TimeoutSec  int    `json:"timeoutSec"`
+	NodeID      *int64 `json:"nodeId"`
+	Enabled     *int   `json:"enabled"`
+}
+
+func (h *Handler) monitorServiceUpdate(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	var req updateServiceMonitorRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+
+	if req.ID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的监控ID"))
+		return
+	}
+
+	existing, err := h.repo.GetServiceMonitor(req.ID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	if existing == nil {
+		response.WriteJSON(w, response.ErrDefault("监控不存在"))
+		return
+	}
+
+	name := strings.TrimSpace(req.Name)
+	if name != "" {
+		existing.Name = name
+	}
+
+	monitorType := strings.ToLower(strings.TrimSpace(req.Type))
+	if monitorType == "tcp" || monitorType == "icmp" {
+		existing.Type = monitorType
+	}
+
+	target := strings.TrimSpace(req.Target)
+	if target != "" {
+		existing.Target = target
+	}
+
+	limits := h.resolveServiceMonitorLimits()
+
+	if req.IntervalSec > 0 {
+		intervalSec := req.IntervalSec
+		if intervalSec < limits.MinIntervalSec {
+			intervalSec = limits.MinIntervalSec
+		}
+		existing.IntervalSec = intervalSec
+	}
+	if req.TimeoutSec > 0 {
+		timeoutSec := req.TimeoutSec
+		if timeoutSec < limits.MinTimeoutSec {
+			timeoutSec = limits.MinTimeoutSec
+		}
+		if timeoutSec > limits.MaxTimeoutSec {
+			timeoutSec = limits.MaxTimeoutSec
+		}
+		existing.TimeoutSec = timeoutSec
+	}
+
+	if req.NodeID != nil {
+		existing.NodeID = *req.NodeID
+	}
+	if req.Enabled != nil {
+		if *req.Enabled == 0 || *req.Enabled == 1 {
+			existing.Enabled = *req.Enabled
+		}
+	}
+
+	existing.UpdatedTime = time.Now().UnixMilli()
+	if existing.Type == "icmp" && existing.NodeID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("ICMP 监控必须选择执行节点"))
+		return
+	}
+	if existing.NodeID > 0 {
+		n, err := h.repo.GetNodeByID(existing.NodeID)
+		if err != nil {
+			response.WriteJSON(w, response.Err(-2, err.Error()))
+			return
+		}
+		if n == nil {
+			response.WriteJSON(w, response.ErrDefault("节点不存在"))
+			return
+		}
+	}
+
+	if err := h.repo.UpdateServiceMonitor(existing); err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(existing))
+}
+
+type deleteServiceMonitorRequest struct {
+	ID int64 `json:"id"`
+}
+
+func (h *Handler) monitorServiceDelete(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	var req deleteServiceMonitorRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+
+	if req.ID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的监控ID"))
+		return
+	}
+
+	if err := h.repo.DeleteServiceMonitor(req.ID); err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OKEmpty())
+}
+
+func (h *Handler) monitorServiceRun(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+	if h.healthCheck == nil {
+		response.WriteJSON(w, response.ErrDefault("监控服务不可用"))
+		return
+	}
+
+	var req deleteServiceMonitorRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+	if req.ID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的监控ID"))
+		return
+	}
+
+	m, err := h.repo.GetServiceMonitor(req.ID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	if m == nil {
+		response.WriteJSON(w, response.ErrDefault("监控不存在"))
+		return
+	}
+
+	res, err := h.healthCheck.RunOnce(m)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	if err := h.repo.InsertServiceMonitorResult(res); err != nil {
+		log.Printf("monitoring write failed op=service_monitor_result.manual_insert monitor_id=%d err=%v", res.MonitorID, err)
+	}
+	response.WriteJSON(w, response.OK(res))
+}
+
+func (h *Handler) monitorServiceResultsHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	monitorIDStr := extractPathParam(r.URL.Path, "/api/v1/monitor/services/", "/results")
+	monitorID, err := strconv.ParseInt(monitorIDStr, 10, 64)
+	if err != nil || monitorID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的监控ID"))
+		return
+	}
+
+	// If start/end time range is provided, use time-based query (mirrors node metrics / tunnel quality pattern).
+	startStr := r.URL.Query().Get("start")
+	endStr := r.URL.Query().Get("end")
+	if startStr != "" && endStr != "" {
+		startMs, err1 := strconv.ParseInt(startStr, 10, 64)
+		endMs, err2 := strconv.ParseInt(endStr, 10, 64)
+		if err1 != nil || err2 != nil || startMs <= 0 || endMs <= 0 || endMs < startMs {
+			response.WriteJSON(w, response.ErrDefault("无效的时间范围"))
+			return
+		}
+		if endMs-startMs > maxMetricsRangeMs {
+			response.WriteJSON(w, response.ErrDefault("时间范围过大"))
+			return
+		}
+		results, err := h.repo.GetServiceMonitorResultsByTimeRange(monitorID, startMs, endMs)
+		if err != nil {
+			response.WriteJSON(w, response.Err(-2, err.Error()))
+			return
+		}
+		response.WriteJSON(w, response.OK(results))
+		return
+	}
+
+	// Fallback: count-based limit query (backward compat).
+	limit := 100
+	if l := r.URL.Query().Get("limit"); l != "" {
+		if v, err := strconv.Atoi(l); err == nil && v > 0 && v <= 1000 {
+			limit = v
+		}
+	}
+
+	results, err := h.repo.GetServiceMonitorResults(monitorID, limit)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+
+	response.WriteJSON(w, response.OK(results))
+}
+
+func (h *Handler) monitorServiceLatestResultsHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+
+	// Try in-memory cache first (updated every 1s)
+	if h.healthCheck != nil {
+		cached := h.healthCheck.GetLatestCached()
+		if len(cached) > 0 {
+			response.WriteJSON(w, response.OK(cached))
+			return
+		}
+	}
+
+	// Fallback to database
+	results, err := h.repo.GetLatestServiceMonitorResults()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	response.WriteJSON(w, response.OK(results))
+}
+
+func (h *Handler) monitorServiceLimitsHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureMonitoringAccess(w, r) {
+		return
+	}
+	response.WriteJSON(w, response.OK(h.resolveServiceMonitorLimits()))
+}
+
+func extractPathParam(path, prefix, suffix string) string {
+	if !strings.HasPrefix(path, prefix) {
+		return ""
+	}
+	rest := strings.TrimPrefix(path, prefix)
+	if suffix != "" {
+		rest = strings.TrimSuffix(rest, suffix)
+	}
+	return rest
+}
+
+type monitorAccessData struct {
+	Allowed bool   `json:"allowed"`
+	Reason  string `json:"reason,omitempty"`
+}
+
+// monitorAccessHandler is a lightweight capability check for frontend navigation.
+// It does NOT replace authorization on the actual monitoring endpoints.
+func (h *Handler) monitorAccessHandler(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+
+	userID, roleID, err := userRoleFromRequest(r)
+	if err != nil {
+		response.WriteJSON(w, response.Err(401, "未登录或token已过期"))
+		return
+	}
+	if roleID == 0 {
+		response.WriteJSON(w, response.OK(monitorAccessData{Allowed: true}))
+		return
+	}
+
+	allowed, err := h.repo.HasMonitorPermission(userID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	data := monitorAccessData{Allowed: allowed}
+	if !allowed {
+		data.Reason = "need_admin_grant"
+	}
+	response.WriteJSON(w, response.OK(data))
+}
+
+func (h *Handler) ensureAdminAccess(w http.ResponseWriter, r *http.Request) bool {
+	_, roleID, err := userRoleFromRequest(r)
+	if err != nil {
+		response.WriteJSON(w, response.Err(401, "未登录或token已过期"))
+		return false
+	}
+	if roleID != 0 {
+		response.WriteJSON(w, response.Err(403, "权限不足，仅管理员可操作"))
+		return false
+	}
+	return true
+}
+
+func (h *Handler) ensureMonitoringAccess(w http.ResponseWriter, r *http.Request) bool {
+	userID, roleID, err := userRoleFromRequest(r)
+	if err != nil {
+		response.WriteJSON(w, response.Err(401, "未登录或token已过期"))
+		return false
+	}
+	if roleID == 0 {
+		return true
+	}
+	allowed, err := h.repo.HasMonitorPermission(userID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return false
+	}
+	if !allowed {
+		response.WriteJSON(w, response.Err(403, "权限不足：当前账户非管理员，且未被授予监控权限。请联系管理员在用户管理中授权监控权限。"))
+		return false
+	}
+	return true
+}
+
+func (h *Handler) monitorPermissionList(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodGet {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureAdminAccess(w, r) {
+		return
+	}
+
+	items, err := h.repo.ListMonitorPermissions()
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	response.WriteJSON(w, response.OK(items))
+}
+
+type monitorPermissionMutationRequest struct {
+	UserID int64 `json:"userId"`
+}
+
+func (h *Handler) monitorPermissionAssign(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureAdminAccess(w, r) {
+		return
+	}
+
+	var req monitorPermissionMutationRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+	if req.UserID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的用户ID"))
+		return
+	}
+
+	u, err := h.repo.GetUserByID(req.UserID)
+	if err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	if u == nil {
+		response.WriteJSON(w, response.ErrDefault("用户不存在"))
+		return
+	}
+
+	if err := h.repo.InsertMonitorPermission(req.UserID, time.Now().UnixMilli()); err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	response.WriteJSON(w, response.OKEmpty())
+}
+
+func (h *Handler) monitorPermissionRemove(w http.ResponseWriter, r *http.Request) {
+	if r.Method != http.MethodPost {
+		response.WriteJSON(w, response.ErrDefault("请求失败"))
+		return
+	}
+	if !h.ensureAdminAccess(w, r) {
+		return
+	}
+
+	var req monitorPermissionMutationRequest
+	if err := decodeJSON(r.Body, &req); err != nil {
+		response.WriteJSON(w, response.ErrDefault("请求参数错误"))
+		return
+	}
+	if req.UserID <= 0 {
+		response.WriteJSON(w, response.ErrDefault("无效的用户ID"))
+		return
+	}
+
+	if err := h.repo.DeleteMonitorPermission(req.UserID); err != nil {
+		response.WriteJSON(w, response.Err(-2, err.Error()))
+		return
+	}
+	response.WriteJSON(w, response.OKEmpty())
+}

go-backend/internal/http/handler/mutations.go

@@ -832,7 +832,7 @@ func (h *Handler) tunnelUpdate(w http.ResponseWriter, r *http.Request) {
 			newEntryNodeIDs = append(newEntryNodeIDs, in.NodeID)
 		}
 	}
-	if err := h.validateTunnelEntryPortConflictsForNewEntries(tx, id, oldEntryNodeIDs, newEntryNodeIDs); err != nil {
+	if err := h.validateTunnelEntryPortConflictsForNewEntriesTx(tx, id, oldEntryNodeIDs, newEntryNodeIDs); err != nil {
 		response.WriteJSON(w, response.ErrDefault(err.Error()))
 		return
 	}
@@ -1020,7 +1020,21 @@ func (h *Handler) cleanupTunnelForwardRuntimesOnRemovedEntryNodes(tunnelID int64
 	}
 }
 
-func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tx *gorm.DB, tunnelID int64, oldEntryNodeIDs, newEntryNodeIDs []int64) error {
+func (h *Handler) validateForwardPortAvailabilityTx(tx *gorm.DB, node *nodeRecord, port int, currentForwardID int64) error {
+	if h == nil || h.repo == nil || tx == nil || node == nil || port <= 0 {
+		return nil
+	}
+	occupied, err := h.repo.HasOtherForwardOnNodePortTx(tx, node.ID, port, currentForwardID)
+	if err != nil {
+		return err
+	}
+	if occupied {
+		return fmt.Errorf("节点 %s 端口 %d 已被其他转发占用", node.Name, port)
+	}
+	return nil
+}
+
+func (h *Handler) validateTunnelEntryPortConflictsForNewEntriesTx(tx *gorm.DB, tunnelID int64, oldEntryNodeIDs, newEntryNodeIDs []int64) error {
 	if h == nil || h.repo == nil || tx == nil || tunnelID <= 0 {
 		return nil
 	}
@@ -1054,9 +1068,7 @@ func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tx *gorm.DB, tun
 			if nodeErr != nil {
 				continue
 			}
-			if err := validateLocalNodePort(node, port); err != nil {
-				return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err)
-			}
+
 			if err := h.validateForwardPortAvailabilityTx(tx, node, port, f.ID); err != nil {
 				return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err)
 			}
@@ -1090,24 +1102,113 @@ func (h *Handler) syncTunnelForwardsEntryPorts(tunnelID int64, entryNodeIDs []in
 		if err != nil {
 			continue
 		}
-		port := pickForwardPortFromRecords(oldPorts)
-		if port <= 0 {
+		referencePort := pickForwardPortFromRecords(oldPorts)
+		if referencePort <= 0 {
 			continue
 		}
 
-		var entries []forwardPortReplaceEntry
-		if allowInIP {
-			entries = buildForwardPortEntriesWithPreservedInIP(entryNodeIDs, oldPorts, port)
-		} else {
-			entries = make([]forwardPortReplaceEntry, 0, len(entryNodeIDs))
-			for _, nid := range entryNodeIDs {
-				entries = append(entries, forwardPortReplaceEntry{NodeID: nid, Port: port, InIP: ""})
+		// Build a map of existing node → port/inIP from old records.
+		oldPortByNode := make(map[int64]forwardPortRecord)
+		for _, fp := range oldPorts {
+			if fp.NodeID > 0 {
+				oldPortByNode[fp.NodeID] = fp
 			}
 		}
+
+		entries := make([]forwardPortReplaceEntry, 0, len(entryNodeIDs))
+		for _, nid := range entryNodeIDs {
+			if existing, ok := oldPortByNode[nid]; ok && existing.Port > 0 {
+				// Existing entry node: keep its current port.
+				inIP := existing.InIP
+				if !allowInIP {
+					inIP = ""
+				}
+				entries = append(entries, forwardPortReplaceEntry{NodeID: nid, Port: existing.Port, InIP: inIP})
+				continue
+			}
+
+			// New entry node: try to follow the reference port.
+			port := h.resolvePortForNewEntryNode(nid, referencePort, f.ID)
+			inIP := ""
+			if allowInIP {
+				// For single-entry tunnels, try to preserve inIP from old records.
+				for _, fp := range oldPorts {
+					if strings.TrimSpace(fp.InIP) != "" {
+						inIP = fp.InIP
+						break
+					}
+				}
+			}
+			entries = append(entries, forwardPortReplaceEntry{NodeID: nid, Port: port, InIP: inIP})
+		}
 		_ = h.repo.ReplaceForwardPorts(f.ID, entries)
 	}
 }
 
+// resolvePortForNewEntryNode determines the port for a forward on a newly added
+// entry node. It tries to reuse referencePort (from existing entries); if that
+// port is out of range or already occupied, it picks a random available port
+// for this specific node.
+func (h *Handler) resolvePortForNewEntryNode(nodeID int64, referencePort int, forwardID int64) int {
+	node, err := h.getNodeRecord(nodeID)
+	if err != nil {
+		return referencePort
+	}
+
+	// Check if referencePort is within the node's allowed range.
+	if validateLocalNodePort(node, referencePort) == nil &&
+		validateRemoteNodePort(node, referencePort) == nil {
+		// In range — check availability.
+		occupied, occErr := h.repo.HasOtherForwardOnNodePort(nodeID, referencePort, forwardID)
+		if occErr == nil && !occupied {
+			return referencePort
+		}
+	}
+
+	// referencePort doesn't work for this node; pick a random one.
+	newPort := h.pickRandomPortForNode(nodeID)
+	if newPort > 0 {
+		return newPort
+	}
+	return referencePort // last resort fallback
+}
+
+// pickRandomPortForNode picks a random available port from a single node's
+// port range, excluding ports already occupied by other forwards or chains.
+func (h *Handler) pickRandomPortForNode(nodeID int64) int {
+	portRange, err := h.repo.GetNodePortRange(nodeID)
+	if err != nil {
+		return 0
+	}
+	if portRange == "" {
+		portRange = "1000-65535"
+	}
+
+	nodePorts, err := parsePorts(portRange)
+	if err != nil || len(nodePorts) == 0 {
+		return 0
+	}
+
+	used, err := h.getUsedPorts(nodeID)
+	if err != nil {
+		return 0
+	}
+
+	var available []int
+	for _, p := range nodePorts {
+		if !used[p] {
+			available = append(available, p)
+		}
+	}
+
+	if len(available) == 0 {
+		return 0
+	}
+
+	idx, _ := rand.Int(rand.Reader, big.NewInt(int64(len(available))))
+	return available[idx.Int64()]
+}
+
 func (h *Handler) tunnelDelete(w http.ResponseWriter, r *http.Request) {
 	if r.Method != http.MethodPost {
 		response.WriteJSON(w, response.ErrDefault("请求失败"))
@@ -1456,6 +1557,12 @@ func (h *Handler) userTunnelRemove(w http.ResponseWriter, r *http.Request) {
 	if id <= 0 {
 		return
 	}
+	userID, tunnelID, lookupErr := h.repo.GetUserTunnelUserAndTunnel(id)
+	if lookupErr != nil {
+		response.WriteJSON(w, response.Err(-2, lookupErr.Error()))
+		return
+	}
+	h.cleanupForwardsForUserTunnel(userID, tunnelID)
 	if err := h.repo.DeleteUserTunnel(id); err != nil {
 		response.WriteJSON(w, response.Err(-2, err.Error()))
 		return
@@ -2382,14 +2489,18 @@ func (h *Handler) groupUserAssign(w http.ResponseWriter, r *http.Request) {
 		response.WriteJSON(w, response.Err(-2, err.Error()))
 		return
 	}
-	if err := h.repo.RevokeGroupGrantsForRemovedUsersTx(tx, req.GroupID, previousUserIDs, req.UserIDs); err != nil {
-		response.WriteJSON(w, response.Err(-2, err.Error()))
+	revokedPairs, revokeErr := h.repo.RevokeGroupGrantsForRemovedUsersTx(tx, req.GroupID, previousUserIDs, req.UserIDs)
+	if revokeErr != nil {
+		response.WriteJSON(w, response.Err(-2, revokeErr.Error()))
 		return
 	}
 	if err := tx.Commit().Error; err != nil {
 		response.WriteJSON(w, response.Err(-2, err.Error()))
 		return
 	}
+	for _, pair := range revokedPairs {
+		h.cleanupForwardsForUserTunnel(pair.UserID, pair.TunnelID)
+	}
 	_ = h.syncPermissionsByUserGroup(req.GroupID)
 	response.WriteJSON(w, response.OKEmpty())
 }
@@ -2433,9 +2544,12 @@ func (h *Handler) groupPermissionRemove(w http.ResponseWriter, r *http.Request)
 		response.WriteJSON(w, response.Err(-2, err.Error()))
 		return
 	}
+	var revokedPairs []repo.RevokedUserTunnelPair
 	if exists {
-		if err := h.repo.RevokeGroupPermissionPairTx(tx, ug, tg); err != nil {
-			response.WriteJSON(w, response.Err(-2, err.Error()))
+		var revokeErr error
+		revokedPairs, revokeErr = h.repo.RevokeGroupPermissionPairTx(tx, ug, tg)
+		if revokeErr != nil {
+			response.WriteJSON(w, response.Err(-2, revokeErr.Error()))
 			return
 		}
 	}
@@ -2444,6 +2558,9 @@ func (h *Handler) groupPermissionRemove(w http.ResponseWriter, r *http.Request)
 		response.WriteJSON(w, response.Err(-2, err.Error()))
 		return
 	}
+	for _, pair := range revokedPairs {
+		h.cleanupForwardsForUserTunnel(pair.UserID, pair.TunnelID)
+	}
 	response.WriteJSON(w, response.OKEmpty())
 }
 
@@ -2663,6 +2780,22 @@ func (h *Handler) prepareTunnelCreateState(tx *gorm.DB, req map[string]interface
 		}
 	}
 
+	// When updating an existing tunnel (excludeTunnelID > 0), build a set of
+	// node IDs that already belong to the tunnel so we can tolerate offline
+	// nodes that the user is keeping or removing, while still rejecting newly
+	// added offline nodes.
+	existingNodeIDs := make(map[int64]struct{})
+	if excludeTunnelID > 0 {
+		var existIDs []int64
+		if err := tx.Model(&model.ChainTunnel{}).
+			Where("tunnel_id = ?", excludeTunnelID).
+			Pluck("node_id", &existIDs).Error; err == nil {
+			for _, eid := range existIDs {
+				existingNodeIDs[eid] = struct{}{}
+			}
+		}
+	}
+
 	seen := make(map[int64]struct{}, len(nodeIDs))
 	for _, nodeID := range nodeIDs {
 		if _, ok := seen[nodeID]; ok {
@@ -2681,7 +2814,12 @@ func (h *Handler) prepareTunnelCreateState(tx *gorm.DB, req map[string]interface
 			return nil, errors.New("节点不存在")
 		}
 		if node.IsRemote != 1 && node.Status != 1 {
-			return nil, errors.New("部分节点不在线")
+			// For tunnel updates, allow offline nodes that already belong to the
+			// tunnel (user may be removing them). Only reject genuinely new offline nodes.
+			_, isExisting := existingNodeIDs[nodeID]
+			if excludeTunnelID <= 0 || !isExisting {
+				return nil, errors.New("部分节点不在线")
+			}
 		}
 		state.Nodes[nodeID] = node
 	}
@@ -3081,7 +3219,6 @@ func (h *Handler) applyTunnelRuntime(state *tunnelCreateState) ([]int64, []int64
 	}
 
 	for _, inNode := range state.InNodes {
-		node := state.Nodes[inNode.NodeID]
 		targets := state.OutNodes
 		if len(state.ChainHops) > 0 {
 			targets = state.ChainHops[0]
@@ -3091,7 +3228,7 @@ func (h *Handler) applyTunnelRuntime(state *tunnelCreateState) ([]int64, []int64
 			return createdChains, createdServices, err
 		}
 		if _, err := h.sendNodeCommand(inNode.NodeID, "AddChains", chainData, true, false); err != nil {
-			if node != nil && node.IsRemote == 1 && shouldDeferTunnelRuntimeApplyError(err) {
+			if shouldDeferTunnelRuntimeApplyError(err) {
 				continue
 			}
 			return createdChains, createdServices, fmt.Errorf("入口节点 %s 下发转发链失败: %w", nodeDisplayName(state.Nodes[inNode.NodeID]), err)
@@ -3105,7 +3242,8 @@ func (h *Handler) applyTunnelRuntime(state *tunnelCreateState) ([]int64, []int64
 			nextTargets = state.ChainHops[i+1]
 		}
 		for _, chainNode := range hop {
-			if node := state.Nodes[chainNode.NodeID]; node != nil && node.IsRemote == 1 {
+			node := state.Nodes[chainNode.NodeID]
+			if node != nil && (node.IsRemote == 1 || node.Status != 1) {
 				continue
 			}
 			chainData, err := buildTunnelChainConfig(state.TunnelID, chainNode.NodeID, nextTargets, state.Nodes, state.IPPreference)
@@ -3113,12 +3251,18 @@ func (h *Handler) applyTunnelRuntime(state *tunnelCreateState) ([]int64, []int64
 				return createdChains, createdServices, err
 			}
 			if _, err := h.sendNodeCommand(chainNode.NodeID, "AddChains", chainData, true, false); err != nil {
+				if shouldDeferTunnelRuntimeApplyError(err) {
+					continue
+				}
 				return createdChains, createdServices, fmt.Errorf("转发链节点 %s 下发转发链失败: %w", nodeDisplayName(state.Nodes[chainNode.NodeID]), err)
 			}
 			createdChains = append(createdChains, chainNode.NodeID)
 
 			serviceData := buildTunnelChainServiceConfig(state.TunnelID, chainNode, state.Nodes[chainNode.NodeID], len(nextTargets))
 			if err := h.addTunnelServiceOnNode(chainNode.NodeID, state.TunnelID, serviceData); err != nil {
+				if shouldDeferTunnelRuntimeApplyError(err) {
+					continue
+				}
 				return createdChains, createdServices, fmt.Errorf("转发链节点 %s 下发服务失败: %w", nodeDisplayName(state.Nodes[chainNode.NodeID]), err)
 			}
 			createdServices = append(createdServices, chainNode.NodeID)
@@ -3126,11 +3270,15 @@ func (h *Handler) applyTunnelRuntime(state *tunnelCreateState) ([]int64, []int64
 	}
 
 	for _, outNode := range state.OutNodes {
-		if node := state.Nodes[outNode.NodeID]; node != nil && node.IsRemote == 1 {
+		node := state.Nodes[outNode.NodeID]
+		if node != nil && (node.IsRemote == 1 || node.Status != 1) {
 			continue
 		}
 		serviceData := buildTunnelChainServiceConfig(state.TunnelID, outNode, state.Nodes[outNode.NodeID], 1)
 		if err := h.addTunnelServiceOnNode(outNode.NodeID, state.TunnelID, serviceData); err != nil {
+			if shouldDeferTunnelRuntimeApplyError(err) {
+				continue
+			}
 			return createdChains, createdServices, fmt.Errorf("出口节点 %s 下发服务失败: %w", nodeDisplayName(state.Nodes[outNode.NodeID]), err)
 		}
 		createdServices = append(createdServices, outNode.NodeID)
@@ -3221,6 +3369,13 @@ func shouldDeferTunnelRuntimeApplyError(err error) bool {
 	return false
 }
 
+// isNodeOfflineOrTimeoutError returns true when the error indicates a node
+// is unreachable (offline or timed out), matching the same patterns used by
+// shouldDeferTunnelRuntimeApplyError.
+func isNodeOfflineOrTimeoutError(err error) bool {
+	return shouldDeferTunnelRuntimeApplyError(err)
+}
+
 func buildTunnelChainConfig(tunnelID int64, fromNodeID int64, targets []tunnelRuntimeNode, nodes map[int64]*nodeRecord, ipPreference string) (map[string]interface{}, error) {
 	fromNode := nodes[fromNodeID]
 	if fromNode == nil {
@@ -3347,6 +3502,7 @@ func selectTunnelDialHost(fromNode, toNode *nodeRecord, ipPreference string, con
 			}
 		}
 	default:
+		// 同版本优先
 		if fromV4 && toV4 {
 			if host := pickNodeAddressV4(toNode); host != "" {
 				return host, nil
@@ -3357,6 +3513,17 @@ func selectTunnelDialHost(fromNode, toNode *nodeRecord, ipPreference string, con
 				return host, nil
 			}
 		}
+		// 跨版本支持：v6入v4出 / v4入v6出
+		if fromV6 && toV4 {
+			if host := pickNodeAddressV4(toNode); host != "" {
+				return host, nil
+			}
+		}
+		if fromV4 && toV6 {
+			if host := pickNodeAddressV6(toNode); host != "" {
+				return host, nil
+			}
+		}
 	}
 	return "", fmt.Errorf("节点链路不兼容：%s(v4=%t,v6=%t) -> %s(v4=%t,v6=%t)", nodeDisplayName(fromNode), fromV4, fromV6, nodeDisplayName(toNode), toV4, toV6)
 }
@@ -3899,6 +4066,27 @@ func (h *Handler) syncUserTunnelForwards(userID, tunnelID int64) error {
 	return nil
 }
 
+// cleanupForwardsForUserTunnel deletes all forwarding rules belonging to a
+// specific user+tunnel pair. It notifies nodes to remove the runtime services
+// first, then deletes the DB records. This is best-effort: individual failures
+// do not abort the overall cleanup so that remaining forwards are still cleaned.
+func (h *Handler) cleanupForwardsForUserTunnel(userID, tunnelID int64) {
+	if userID <= 0 || tunnelID <= 0 {
+		return
+	}
+	forwards, err := h.repo.ListForwardsByUserAndTunnel(userID, tunnelID)
+	if err != nil || len(forwards) == 0 {
+		return
+	}
+	for i := range forwards {
+		f := &forwards[i]
+		if f.Status == 1 {
+			_ = h.controlForwardServices(f, "DeleteService", true)
+		}
+		_ = h.deleteForwardByID(f.ID)
+	}
+}
+
 func (h *Handler) normalizeSpeedLimitReference(speedID *int64) (*int64, error) {
 	if speedID == nil {
 		return nil, nil
@@ -4153,20 +4341,6 @@ func (h *Handler) validateForwardPortAvailability(node *nodeRecord, port int, cu
 	return nil
 }
 
-func (h *Handler) validateForwardPortAvailabilityTx(tx *gorm.DB, node *nodeRecord, port int, currentForwardID int64) error {
-	if h == nil || h.repo == nil || tx == nil || node == nil || port <= 0 {
-		return nil
-	}
-	occupied, err := h.repo.HasOtherForwardOnNodePortTx(tx, node.ID, port, currentForwardID)
-	if err != nil {
-		return err
-	}
-	if occupied {
-		return fmt.Errorf("节点 %s 端口 %d 已被其他转发占用", node.Name, port)
-	}
-	return nil
-}
-
 func parsePortRangeMinMax(input string) (int, int) {
 	input = strings.TrimSpace(input)
 	if input == "" {

go-backend/internal/http/handler/tunnel_entry_sqlite_test.go

@@ -83,7 +83,7 @@ func TestValidateTunnelEntryPortConflictsForNewEntriesDoesNotBlockOnSQLiteTx(t *
 	doneCh := make(chan struct{})
 	go func() {
 		defer close(doneCh)
-		errCh <- h.validateTunnelEntryPortConflictsForNewEntries(tx, tunnelID, []int64{oldEntryID}, []int64{oldEntryID, newEntryID})
+		errCh <- h.validateTunnelEntryPortConflictsForNewEntriesTx(tx, tunnelID, []int64{oldEntryID}, []int64{oldEntryID, newEntryID})
 	}()
 
 	select {

go-backend/internal/http/handler/tunnel_metrics_ingestion.go

@@ -0,0 +1,120 @@
+package handler
+
+import (
+	"log"
+	"strings"
+	"time"
+
+	"go-backend/internal/store/model"
+)
+
+type tunnelTrafficDelta struct {
+	bytesIn  int64
+	bytesOut int64
+}
+
+func unixMilliBucketMinute(nowMs int64) int64 {
+	if nowMs <= 0 {
+		return 0
+	}
+	const minuteMs = int64(time.Minute / time.Millisecond)
+	return nowMs - (nowMs % minuteMs)
+}
+
+func (h *Handler) recordTunnelMetricsFromFlowItems(nodeID int64, items []flowItem, nowMs int64) {
+	if h == nil || h.repo == nil {
+		return
+	}
+	if nodeID <= 0 || len(items) == 0 {
+		return
+	}
+
+	bucketTs := unixMilliBucketMinute(nowMs)
+	if bucketTs <= 0 {
+		return
+	}
+
+	forwardDeltas := make(map[int64]tunnelTrafficDelta)
+	var skippedParse, skippedZero int
+	for _, item := range items {
+		name := strings.TrimSpace(item.N)
+		if name == "" || name == "web_api" {
+			continue
+		}
+		forwardID, _, _, ok := parseFlowServiceIDs(name)
+		if !ok {
+			skippedParse++
+			continue
+		}
+		if item.D == 0 && item.U == 0 {
+			skippedZero++
+			continue
+		}
+		d := forwardDeltas[forwardID]
+		d.bytesIn += item.D
+		d.bytesOut += item.U
+		forwardDeltas[forwardID] = d
+	}
+	if len(forwardDeltas) == 0 {
+		if len(items) > 0 {
+			log.Printf("monitoring debug op=tunnel_metric.no_forward_deltas node_id=%d items=%d skipped_parse=%d skipped_zero=%d", nodeID, len(items), skippedParse, skippedZero)
+		}
+		return
+	}
+
+	forwardIDs := make([]int64, 0, len(forwardDeltas))
+	for id := range forwardDeltas {
+		forwardIDs = append(forwardIDs, id)
+	}
+
+	forwardTunnelMap, err := h.repo.MapForwardIDsToTunnelIDs(forwardIDs)
+	if err != nil {
+		log.Printf("monitoring write skipped op=tunnel_metric.map_forward_to_tunnel node_id=%d err=%v", nodeID, err)
+		return
+	}
+	if len(forwardTunnelMap) == 0 {
+		log.Printf("monitoring debug op=tunnel_metric.no_tunnel_map node_id=%d forward_ids=%v", nodeID, forwardIDs)
+		return
+	}
+
+	tunnelAgg := make(map[int64]tunnelTrafficDelta)
+	for forwardID, delta := range forwardDeltas {
+		tunnelID := forwardTunnelMap[forwardID]
+		if tunnelID <= 0 {
+			continue
+		}
+		a := tunnelAgg[tunnelID]
+		a.bytesIn += delta.bytesIn
+		a.bytesOut += delta.bytesOut
+		tunnelAgg[tunnelID] = a
+	}
+	if len(tunnelAgg) == 0 {
+		return
+	}
+
+	metrics := make([]*model.TunnelMetric, 0, len(tunnelAgg))
+	for tunnelID, delta := range tunnelAgg {
+		if delta.bytesIn == 0 && delta.bytesOut == 0 {
+			continue
+		}
+		metrics = append(metrics, &model.TunnelMetric{
+			TunnelID:     tunnelID,
+			NodeID:       nodeID,
+			Timestamp:    bucketTs,
+			BytesIn:      delta.bytesIn,
+			BytesOut:     delta.bytesOut,
+			Connections:  0,
+			Errors:       0,
+			AvgLatencyMs: 0,
+		})
+	}
+	if len(metrics) == 0 {
+		return
+	}
+
+	if err := h.repo.UpsertTunnelMetricBuckets(metrics); err != nil {
+		log.Printf("monitoring write failed op=tunnel_metric.upsert_buckets node_id=%d bucket_ts=%d count=%d err=%v", nodeID, bucketTs, len(metrics), err)
+	} else {
+		log.Printf("monitoring ok op=tunnel_metric.upsert_buckets node_id=%d bucket_ts=%d count=%d", nodeID, bucketTs, len(metrics))
+	}
+}

go-backend/internal/http/handler/tunnel_quality_prober.go

@@ -0,0 +1,452 @@
+package handler
+
+import (
+	"context"
+	"encoding/json"
+	"log"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"go-backend/internal/store/model"
+)
+
+const (
+	tunnelQualityProbeInterval  = 1 * time.Second
+	tunnelQualityProbeTimeout   = 8 * time.Second
+	tunnelQualityPingTimeoutMs  = 5000
+	tunnelQualityRetention      = 24 * time.Hour // keep 24h of history
+	tunnelQualityPruneInterval  = 10 * time.Minute
+	tunnelQualityReportInterval = 30 * time.Second // DB save interval
+)
+
+type TunnelQualityHop struct {
+	FromNodeID   int64   `json:"fromNodeId"`
+	FromNodeName string  `json:"fromNodeName"`
+	ToNodeID     int64   `json:"toNodeId"`
+	ToNodeName   string  `json:"toNodeName"`
+	Latency      float64 `json:"latency"`
+	Loss         float64 `json:"loss"`
+	TargetIP     string  `json:"targetIp,omitempty"`
+	TargetPort   int     `json:"targetPort,omitempty"`
+}
+
+// tunnelQualitySnapshot is the in-memory latest probe result for a tunnel.
+type tunnelQualitySnapshot struct {
+	TunnelID           int64   `json:"tunnelId"`
+	EntryToExitLatency float64 `json:"entryToExitLatency"`
+	ExitToBingLatency  float64 `json:"exitToBingLatency"`
+	EntryToExitLoss    float64 `json:"entryToExitLoss"`
+	ExitToBingLoss     float64 `json:"exitToBingLoss"`
+	Success            bool    `json:"success"`
+	ErrorMessage       string  `json:"errorMessage,omitempty"`
+	Timestamp          int64   `json:"timestamp"`
+	ChainDetails       string  `json:"chainDetails,omitempty"`
+
+	// internal fields for db reporting
+	lastDBWrite int64 `json:"-"`
+}
+
+// tunnelQualityProber runs periodic TCP ping probes against all enabled tunnels.
+// Design mirrors health.Checker: background goroutine with worker pool + scheduled cleanup.
+type tunnelQualityProber struct {
+	handler   *Handler
+	cache     sync.Map // tunnelID (int64) → *tunnelQualitySnapshot
+	ctx       context.Context
+	cancel    context.CancelFunc
+	interval  time.Duration
+	lastPrune int64
+	probing   int32 // atomic flag: 1 = probeAll running, 0 = idle
+}
+
+// newTunnelQualityProber creates a new prober (not yet running).
+func newTunnelQualityProber(h *Handler) *tunnelQualityProber {
+	return &tunnelQualityProber{
+		handler:  h,
+		interval: tunnelQualityProbeInterval,
+	}
+}
+
+// Start launches the background probe loop (call from jobs.go).
+func (p *tunnelQualityProber) Start(ctx context.Context) {
+	// Use the provided context so we stop with other background jobs.
+	p.ctx, p.cancel = context.WithCancel(ctx)
+	p.loop()
+}
+
+// Stop halts the background probe loop.
+func (p *tunnelQualityProber) Stop() {
+	if p == nil || p.cancel == nil {
+		return
+	}
+
+	p.cancel()
+}
+
+// GetAll returns all cached quality snapshots (latest per tunnel).
+func (p *tunnelQualityProber) GetAll() []tunnelQualitySnapshot {
+	var items []tunnelQualitySnapshot
+	p.cache.Range(func(_, value interface{}) bool {
+		if snap, ok := value.(*tunnelQualitySnapshot); ok {
+			items = append(items, *snap)
+		}
+		return true
+	})
+	return items
+}
+
+func (p *tunnelQualityProber) loop() {
+	// Initial delay to let the system boot up
+	select {
+	case <-time.After(5 * time.Second):
+	case <-p.ctx.Done():
+		return
+	}
+
+	// Run once immediately
+	p.probeAll()
+
+	ticker := time.NewTicker(p.interval)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-p.ctx.Done():
+			return
+		case <-ticker.C:
+			p.probeAll()
+			p.maybePrune()
+		}
+	}
+}
+
+func (p *tunnelQualityProber) isEnabled() bool {
+	if p == nil || p.handler == nil {
+		return true
+	}
+
+	return p.handler.isTunnelQualityMonitoringEnabled()
+}
+
+// maybePrune deletes old quality rows periodically (mirrors PruneServiceMonitorResults).
+func (p *tunnelQualityProber) maybePrune() {
+	if !p.isEnabled() {
+		return
+	}
+
+	now := time.Now().UnixMilli()
+	if p.lastPrune > 0 && now-p.lastPrune < int64(tunnelQualityPruneInterval/time.Millisecond) {
+		return
+	}
+	p.lastPrune = now
+
+	h := p.handler
+	if h == nil || h.repo == nil {
+		return
+	}
+
+	cutoff := now - int64(tunnelQualityRetention/time.Millisecond)
+	if err := h.repo.PruneTunnelQualityResults(cutoff); err != nil {
+		log.Printf("tunnel_quality_prober: prune err=%v", err)
+	}
+}
+
+func (p *tunnelQualityProber) probeAll() {
+	if !p.isEnabled() {
+		return
+	}
+
+	// Skip if previous probe round is still running (interval < timeout guard)
+	if !atomic.CompareAndSwapInt32(&p.probing, 0, 1) {
+		return
+	}
+	defer atomic.StoreInt32(&p.probing, 0)
+
+	h := p.handler
+	if h == nil || h.repo == nil {
+		return
+	}
+
+	tunnelIDs, err := h.repo.ListEnabledTunnelIDs()
+	if err != nil {
+		log.Printf("tunnel_quality_prober: list enabled tunnels err=%v", err)
+		return
+	}
+	if len(tunnelIDs) == 0 {
+		return
+	}
+
+	// Probe tunnels concurrently with a worker limit
+	// (mirrors health.Checker worker pool pattern)
+	const maxWorkers = 20
+	sem := make(chan struct{}, maxWorkers)
+	var wg sync.WaitGroup
+
+	for _, tunnelID := range tunnelIDs {
+		select {
+		case <-p.ctx.Done():
+			return
+		default:
+		}
+
+		wg.Add(1)
+		sem <- struct{}{}
+		go func(tid int64) {
+			defer wg.Done()
+			defer func() { <-sem }()
+			p.probeTunnel(tid)
+		}(tunnelID)
+	}
+	wg.Wait()
+}
+
+func (p *tunnelQualityProber) probeTunnel(tunnelID int64) {
+	h := p.handler
+	if h == nil || h.repo == nil {
+		return
+	}
+
+	now := time.Now().UnixMilli()
+	snap := &tunnelQualitySnapshot{
+		TunnelID:  tunnelID,
+		Timestamp: now,
+	}
+
+	// Get tunnel chain info
+	tunnel, err := h.getTunnelRecord(tunnelID)
+	if err != nil {
+		snap.ErrorMessage = "隧道不存在"
+		p.storeResult(snap)
+		return
+	}
+
+	chainRows, err := h.listChainNodesForTunnel(tunnelID)
+	if err != nil || len(chainRows) == 0 {
+		snap.ErrorMessage = "隧道配置不完整"
+		p.storeResult(snap)
+		return
+	}
+
+	ipPreference := h.repo.GetTunnelIPPreference(tunnelID)
+	inNodes, midNodesGrouped, outNodes := splitChainNodeGroups(chainRows)
+
+	options := diagnosisExecOptions{
+		commandTimeout: tunnelQualityProbeTimeout,
+		pingTimeoutMS:  tunnelQualityPingTimeoutMs,
+		timeoutMessage: "探测超时",
+	}
+
+	switch tunnel.Type {
+	case 1:
+		// Port forwarding: entry → Bing only
+		if len(inNodes) > 0 {
+			lat, loss, err := p.tcpPingNode(inNodes[0].NodeID, "www.bing.com", 443, options)
+			if err == nil {
+				snap.ExitToBingLatency = lat
+				snap.ExitToBingLoss = loss
+				snap.Success = true
+			} else {
+				snap.ErrorMessage = err.Error()
+			}
+		}
+	case 2:
+		// Tunnel forwarding: entry → exit + exit → Bing
+		probeOK := true
+
+		if len(inNodes) > 0 && len(outNodes) > 0 {
+			var hops []TunnelQualityHop
+			var totalLat float64
+			remainingSuccessProb := 1.0
+
+			nodesInPath := make([]chainNodeRecord, 0, 2+len(midNodesGrouped))
+			nodesInPath = append(nodesInPath, inNodes[0])
+			for _, midGroup := range midNodesGrouped {
+				if len(midGroup) > 0 {
+					nodesInPath = append(nodesInPath, midGroup[0])
+				}
+			}
+			nodesInPath = append(nodesInPath, outNodes[0])
+
+			for i := 0; i < len(nodesInPath)-1; i++ {
+				source := nodesInPath[i]
+				target := nodesInPath[i+1]
+
+				hop := TunnelQualityHop{
+					FromNodeID:   source.NodeID,
+					FromNodeName: source.NodeName,
+					ToNodeID:     target.NodeID,
+					ToNodeName:   target.NodeName,
+				}
+
+				targetNode, nodeErr := h.getNodeRecord(target.NodeID)
+				if nodeErr != nil || targetNode == nil {
+					snap.ErrorMessage = "节点 " + target.NodeName + " 不可用"
+					probeOK = false
+					hop.Latency = -1
+					hop.Loss = 100
+					hops = append(hops, hop)
+					break
+				}
+				
+				fromNode, _ := h.getNodeRecord(source.NodeID)
+				targetIP, targetPort, resolveErr := resolveChainProbeTarget(fromNode, targetNode, target.Port, ipPreference, target.ConnectIP)
+				if resolveErr != nil {
+					snap.ErrorMessage = "解析节点 " + target.NodeName + " 失败: " + resolveErr.Error()
+					probeOK = false
+					hop.Latency = -1
+					hop.Loss = 100
+					hops = append(hops, hop)
+					break
+				}
+
+				hop.TargetIP = targetIP
+				hop.TargetPort = targetPort
+
+				lat, loss, err := p.tcpPingNode(source.NodeID, targetIP, targetPort, options)
+				if err == nil {
+					hop.Latency = lat
+					hop.Loss = loss
+					totalLat += lat
+					remainingSuccessProb *= (1.0 - loss/100.0)
+					hops = append(hops, hop)
+				} else {
+					probeOK = false
+					hop.Latency = -1
+					hop.Loss = 100
+					hops = append(hops, hop)
+					if snap.ErrorMessage == "" {
+						snap.ErrorMessage = err.Error()
+					}
+					break
+				}
+			}
+
+			if probeOK {
+				snap.EntryToExitLatency = totalLat
+				snap.EntryToExitLoss = (1.0 - remainingSuccessProb) * 100.0
+			} else {
+				snap.EntryToExitLatency = -1
+				snap.EntryToExitLoss = 100
+			}
+
+			if len(hops) > 0 {
+				if b, err := json.Marshal(hops); err == nil {
+					snap.ChainDetails = string(b)
+				}
+			}
+		}
+
+		// Exit → Bing
+		if len(outNodes) > 0 {
+			lat, loss, err := p.tcpPingNode(outNodes[0].NodeID, "www.bing.com", 443, options)
+			if err == nil {
+				snap.ExitToBingLatency = lat
+				snap.ExitToBingLoss = loss
+			} else {
+				if snap.ErrorMessage == "" {
+					snap.ErrorMessage = err.Error()
+				}
+				probeOK = false
+			}
+		}
+
+		snap.Success = probeOK
+	default:
+		// Unknown type: entry → Bing
+		if len(inNodes) > 0 {
+			lat, loss, err := p.tcpPingNode(inNodes[0].NodeID, "www.bing.com", 443, options)
+			if err == nil {
+				snap.ExitToBingLatency = lat
+				snap.ExitToBingLoss = loss
+				snap.Success = true
+			} else {
+				snap.ErrorMessage = err.Error()
+			}
+		}
+	}
+
+	p.storeResult(snap)
+}
+
+func (p *tunnelQualityProber) tcpPingNode(nodeID int64, ip string, port int, options diagnosisExecOptions) (latency float64, loss float64, err error) {
+	h := p.handler
+	if h == nil {
+		return 0, 100, nil
+	}
+
+	node, nodeErr := h.getNodeRecord(nodeID)
+	if nodeErr != nil {
+		return 0, 100, nodeErr
+	}
+
+	var pingData map[string]interface{}
+	var pingErr error
+	if node != nil && node.IsRemote == 1 {
+		pingData, pingErr = h.tcpPingViaRemoteNode(node, ip, port, options)
+	} else {
+		pingData, pingErr = h.tcpPingViaNode(nodeID, ip, port, options)
+	}
+	if pingErr != nil {
+		return 0, 100, pingErr
+	}
+
+	avgTime := asFloat(pingData["averageTime"], 0)
+	packetLoss := asFloat(pingData["packetLoss"], 100)
+
+	return avgTime, packetLoss, nil
+}
+
+func (p *tunnelQualityProber) storeResult(snap *tunnelQualitySnapshot) {
+	if snap == nil {
+		return
+	}
+
+	// Update in-memory cache (latest per tunnel)
+	// Retain the lastDBWrite timestamp if it exists, so we only DB write every 30s
+	var lastWrite int64
+	if existing, ok := p.cache.Load(snap.TunnelID); ok {
+		if eg, ok := existing.(*tunnelQualitySnapshot); ok {
+			lastWrite = eg.lastDBWrite
+		}
+	}
+	snap.lastDBWrite = lastWrite
+
+	now := time.Now().UnixMilli()
+	writeToDB := false
+	if now-snap.lastDBWrite >= int64(tunnelQualityReportInterval/time.Millisecond) {
+		writeToDB = true
+		snap.lastDBWrite = now
+	}
+
+	p.cache.Store(snap.TunnelID, snap)
+
+	if !writeToDB {
+		return
+	}
+
+	// Persist to database (history)
+	h := p.handler
+	if h == nil || h.repo == nil {
+		return
+	}
+
+	successInt := 0
+	if snap.Success {
+		successInt = 1
+	}
+
+	q := &model.TunnelQuality{
+		TunnelID:           snap.TunnelID,
+		EntryToExitLatency: snap.EntryToExitLatency,
+		ExitToBingLatency:  snap.ExitToBingLatency,
+		EntryToExitLoss:    snap.EntryToExitLoss,
+		ExitToBingLoss:     snap.ExitToBingLoss,
+		Success:            successInt,
+		ErrorMessage:       snap.ErrorMessage,
+		Timestamp:          snap.Timestamp,
+		ChainDetails:       snap.ChainDetails,
+	}
+	if err := h.repo.InsertTunnelQuality(q); err != nil {
+		log.Printf("tunnel_quality_prober: insert db err=%v tunnel_id=%d", err, snap.TunnelID)
+	}
+}

go-backend/internal/http/middleware/auth.go

@@ -101,6 +101,10 @@ func shouldSkip(path string) bool {
 }
 
 func requiresAdmin(path string) bool {
+	if strings.HasPrefix(path, "/api/v1/monitor/permission/") {
+		return true
+	}
+
 	if strings.HasPrefix(path, "/api/v1/group/") {
 		return true
 	}

go-backend/internal/metrics/ingestion.go

@@ -0,0 +1,135 @@
+package metrics
+
+import (
+	"context"
+	"log"
+	"sync"
+	"time"
+
+	"go-backend/internal/store/model"
+	"go-backend/internal/store/repo"
+)
+
+type SystemInfo struct {
+	Uptime           uint64  `json:"uptime"`
+	BytesReceived    uint64  `json:"bytes_received"`
+	BytesTransmitted uint64  `json:"bytes_transmitted"`
+	CPUUsage         float64 `json:"cpu_usage"`
+	MemoryUsage      float64 `json:"memory_usage"`
+	DiskUsage        float64 `json:"disk_usage"`
+	Load1            float64 `json:"load1"`
+	Load5            float64 `json:"load5"`
+	Load15           float64 `json:"load15"`
+	TCPConns         int64   `json:"tcp_conns"`
+	UDPConns         int64   `json:"udp_conns"`
+	NetInSpeed       int64   `json:"net_in_speed"`
+	NetOutSpeed      int64   `json:"net_out_speed"`
+}
+
+type IngestionService struct {
+	repo          *repo.Repository
+	nodeBuffer    []*model.NodeMetric
+	nodeBufferMu  sync.Mutex
+	flushInterval time.Duration
+	retentionDays int
+}
+
+func NewIngestionService(repo *repo.Repository) *IngestionService {
+	return &IngestionService{
+		repo:          repo,
+		nodeBuffer:    make([]*model.NodeMetric, 0, 500),
+		flushInterval: 30 * time.Second,
+		retentionDays: 7,
+	}
+}
+
+func (s *IngestionService) Start(ctx context.Context) {
+	flushTicker := time.NewTicker(s.flushInterval)
+	defer flushTicker.Stop()
+
+	pruneTicker := time.NewTicker(1 * time.Hour)
+	defer pruneTicker.Stop()
+
+	for {
+		select {
+		case <-ctx.Done():
+			s.flushNodeMetrics()
+			return
+		case <-flushTicker.C:
+			s.flushNodeMetrics()
+		case <-pruneTicker.C:
+			s.pruneMetrics()
+		}
+	}
+}
+
+func (s *IngestionService) RecordNodeMetric(nodeID int64, info SystemInfo) {
+	m := &model.NodeMetric{
+		NodeID:      nodeID,
+		Timestamp:   time.Now().UnixMilli(),
+		CPUUsage:    info.CPUUsage,
+		MemUsage:    info.MemoryUsage,
+		DiskUsage:   info.DiskUsage,
+		NetInBytes:  int64(info.BytesReceived),
+		NetOutBytes: int64(info.BytesTransmitted),
+		NetInSpeed:  info.NetInSpeed,
+		NetOutSpeed: info.NetOutSpeed,
+		Load1:       info.Load1,
+		Load5:       info.Load5,
+		Load15:      info.Load15,
+		TCPConns:    info.TCPConns,
+		UDPConns:    info.UDPConns,
+		Uptime:      int64(info.Uptime),
+	}
+
+	s.nodeBufferMu.Lock()
+	s.nodeBuffer = append(s.nodeBuffer, m)
+	shouldFlush := len(s.nodeBuffer) >= 200
+	s.nodeBufferMu.Unlock()
+
+	if shouldFlush {
+		go s.flushNodeMetrics()
+	}
+}
+
+func (s *IngestionService) flushNodeMetrics() {
+	s.nodeBufferMu.Lock()
+	if len(s.nodeBuffer) == 0 {
+		s.nodeBufferMu.Unlock()
+		return
+	}
+	buffer := s.nodeBuffer
+	s.nodeBuffer = make([]*model.NodeMetric, 0, 500)
+	s.nodeBufferMu.Unlock()
+
+	if s.repo == nil {
+		return
+	}
+	if err := s.repo.InsertNodeMetricBatch(buffer); err != nil {
+		log.Printf("monitoring write failed op=node_metric.flush count=%d err=%v", len(buffer), err)
+	}
+}
+
+func (s *IngestionService) pruneMetrics() {
+	cutoff := time.Now().Add(-time.Duration(s.retentionDays) * 24 * time.Hour).UnixMilli()
+	if s.repo == nil {
+		return
+	}
+	if err := s.repo.PruneNodeMetrics(cutoff); err != nil {
+		log.Printf("monitoring prune failed op=node_metric cutoff=%d err=%v", cutoff, err)
+	}
+	if err := s.repo.PruneTunnelMetrics(cutoff); err != nil {
+		log.Printf("monitoring prune failed op=tunnel_metric cutoff=%d err=%v", cutoff, err)
+	}
+	if err := s.repo.PruneServiceMonitorResults(cutoff); err != nil {
+		log.Printf("monitoring prune failed op=service_monitor_result cutoff=%d err=%v", cutoff, err)
+	}
+}
+
+func (s *IngestionService) GetLatestMetric(nodeID int64) (*model.NodeMetric, error) {
+	return s.repo.GetLatestNodeMetric(nodeID)
+}
+
+func (s *IngestionService) GetMetrics(nodeID int64, startMs, endMs int64) ([]model.NodeMetric, error) {
+	return s.repo.GetNodeMetrics(nodeID, startMs, endMs)
+}

go-backend/internal/metrics/ingestion_test.go

@@ -0,0 +1,294 @@
+package metrics
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"go-backend/internal/store/repo"
+)
+
+func TestRecordNodeMetric(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	info := SystemInfo{
+		Uptime:           86400,
+		BytesReceived:    1024000,
+		BytesTransmitted: 2048000,
+		CPUUsage:         45.5,
+		MemoryUsage:      60.2,
+		DiskUsage:        30.1,
+		Load1:            1.5,
+		Load5:            1.2,
+		Load15:           0.9,
+		TCPConns:         100,
+		UDPConns:         50,
+		NetInSpeed:       51200,
+		NetOutSpeed:      102400,
+	}
+
+	svc.RecordNodeMetric(1, info)
+	svc.flushNodeMetrics()
+
+	metrics, err := r.GetNodeMetrics(1, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) != 1 {
+		t.Fatalf("expected 1 metric, got %d", len(metrics))
+	}
+
+	m := metrics[0]
+	if m.CPUUsage != 45.5 {
+		t.Fatalf("expected CPUUsage 45.5, got %f", m.CPUUsage)
+	}
+	if m.MemUsage != 60.2 {
+		t.Fatalf("expected MemUsage 60.2, got %f", m.MemUsage)
+	}
+	if m.DiskUsage != 30.1 {
+		t.Fatalf("expected DiskUsage 30.1, got %f", m.DiskUsage)
+	}
+	if m.Load1 != 1.5 {
+		t.Fatalf("expected Load1 1.5, got %f", m.Load1)
+	}
+	if m.TCPConns != 100 {
+		t.Fatalf("expected TCPConns 100, got %d", m.TCPConns)
+	}
+	if m.UDPConns != 50 {
+		t.Fatalf("expected UDPConns 50, got %d", m.UDPConns)
+	}
+}
+
+func TestRecordNodeMetricAutoFlush(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	info := SystemInfo{
+		CPUUsage:    50.0,
+		MemoryUsage: 60.0,
+		DiskUsage:   30.0,
+	}
+
+	for i := 0; i < 250; i++ {
+		svc.RecordNodeMetric(1, info)
+	}
+
+	time.Sleep(100 * time.Millisecond)
+
+	metrics, err := r.GetNodeMetrics(1, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) < 200 {
+		t.Fatalf("expected at least 200 metrics after auto-flush, got %d", len(metrics))
+	}
+}
+
+func TestIngestionServiceStart(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+	svc.flushInterval = 100 * time.Millisecond
+
+	ctx, cancel := context.WithTimeout(context.Background(), 500*time.Millisecond)
+	defer cancel()
+
+	info := SystemInfo{
+		CPUUsage:    45.0,
+		MemoryUsage: 55.0,
+		DiskUsage:   35.0,
+	}
+
+	go svc.Start(ctx)
+
+	for i := 0; i < 10; i++ {
+		svc.RecordNodeMetric(1, info)
+		time.Sleep(50 * time.Millisecond)
+	}
+
+	<-ctx.Done()
+
+	metrics, err := r.GetNodeMetrics(1, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) == 0 {
+		t.Fatalf("expected metrics after service run")
+	}
+}
+
+func TestGetLatestMetric(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	now := time.Now().UnixMilli()
+
+	info1 := SystemInfo{CPUUsage: 40.0, MemoryUsage: 50.0, DiskUsage: 30.0}
+	svc.RecordNodeMetric(1, info1)
+
+	time.Sleep(5 * time.Millisecond)
+
+	info2 := SystemInfo{CPUUsage: 60.0, MemoryUsage: 70.0, DiskUsage: 40.0}
+	svc.RecordNodeMetric(1, info2)
+
+	svc.flushNodeMetrics()
+
+	latest, err := svc.GetLatestMetric(1)
+	if err != nil {
+		t.Fatalf("get latest: %v", err)
+	}
+	if latest == nil {
+		t.Fatalf("expected latest metric")
+	}
+	if latest.CPUUsage != 60.0 {
+		t.Fatalf("expected latest CPUUsage 60.0, got %f", latest.CPUUsage)
+	}
+
+	_ = now
+
+	latestNone, err := svc.GetLatestMetric(999)
+	if err != nil {
+		t.Fatalf("get latest for non-existent: %v", err)
+	}
+	if latestNone != nil {
+		t.Fatalf("expected nil for non-existent node")
+	}
+}
+
+func TestGetMetricsWithTimeRange(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	now := time.Now().UnixMilli()
+
+	for i := 0; i < 5; i++ {
+		info := SystemInfo{
+			CPUUsage:    float64(40 + i*5),
+			MemoryUsage: 50.0,
+			DiskUsage:   30.0,
+		}
+		svc.RecordNodeMetric(1, info)
+		time.Sleep(10 * time.Millisecond)
+	}
+
+	svc.flushNodeMetrics()
+
+	metrics, err := svc.GetMetrics(1, now-60000, now+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) != 5 {
+		t.Fatalf("expected 5 metrics, got %d", len(metrics))
+	}
+}
+
+func TestPruneMetrics(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+	svc.retentionDays = 1
+
+	info := SystemInfo{CPUUsage: 50.0, MemoryUsage: 60.0, DiskUsage: 30.0}
+
+	svc.RecordNodeMetric(1, info)
+	svc.flushNodeMetrics()
+
+	svc.pruneMetrics()
+
+	metrics, err := r.GetNodeMetrics(1, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) != 1 {
+		t.Fatalf("expected 1 metric (not pruned), got %d", len(metrics))
+	}
+}
+
+func TestMultipleNodes(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	info := SystemInfo{
+		CPUUsage:    50.0,
+		MemoryUsage: 60.0,
+		DiskUsage:   30.0,
+	}
+
+	svc.RecordNodeMetric(1, info)
+	svc.RecordNodeMetric(2, info)
+	svc.RecordNodeMetric(3, info)
+
+	svc.flushNodeMetrics()
+
+	for nodeID := int64(1); nodeID <= 3; nodeID++ {
+		metrics, err := r.GetNodeMetrics(nodeID, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+		if err != nil {
+			t.Fatalf("get metrics for node %d: %v", nodeID, err)
+		}
+		if len(metrics) != 1 {
+			t.Fatalf("expected 1 metric for node %d, got %d", nodeID, len(metrics))
+		}
+	}
+}
+
+func TestZeroValues(t *testing.T) {
+	r, err := repo.Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	svc := NewIngestionService(r)
+
+	info := SystemInfo{}
+
+	svc.RecordNodeMetric(1, info)
+	svc.flushNodeMetrics()
+
+	metrics, err := r.GetNodeMetrics(1, time.Now().UnixMilli()-60000, time.Now().UnixMilli()+1000)
+	if err != nil {
+		t.Fatalf("get metrics: %v", err)
+	}
+	if len(metrics) != 1 {
+		t.Fatalf("expected 1 metric, got %d", len(metrics))
+	}
+
+	m := metrics[0]
+	if m.CPUUsage != 0 || m.MemUsage != 0 || m.DiskUsage != 0 {
+		t.Fatalf("expected zero values, got CPU=%f Mem=%f Disk=%f", m.CPUUsage, m.MemUsage, m.DiskUsage)
+	}
+}

go-backend/internal/monitoring/limits.go

@@ -0,0 +1,114 @@
+package monitoring
+
+import (
+	"strconv"
+	"strings"
+)
+
+type ServiceMonitorLimits struct {
+	CheckerScanIntervalSec int `json:"checkerScanIntervalSec"`
+	WorkerLimit            int `json:"workerLimit"`
+
+	MinIntervalSec     int `json:"minIntervalSec"`
+	DefaultIntervalSec int `json:"defaultIntervalSec"`
+
+	MinTimeoutSec     int `json:"minTimeoutSec"`
+	DefaultTimeoutSec int `json:"defaultTimeoutSec"`
+	MaxTimeoutSec     int `json:"maxTimeoutSec"`
+}
+
+const (
+	ConfigServiceMonitorCheckerScanIntervalSec = "service_monitor_checker_scan_interval_sec"
+	ConfigServiceMonitorWorkerLimit            = "service_monitor_worker_limit"
+	ConfigServiceMonitorMinIntervalSec         = "service_monitor_min_interval_sec"
+	ConfigServiceMonitorDefaultIntervalSec     = "service_monitor_default_interval_sec"
+	ConfigServiceMonitorMinTimeoutSec          = "service_monitor_min_timeout_sec"
+	ConfigServiceMonitorDefaultTimeoutSec      = "service_monitor_default_timeout_sec"
+	ConfigServiceMonitorMaxTimeoutSec          = "service_monitor_max_timeout_sec"
+)
+
+func DefaultServiceMonitorLimits() ServiceMonitorLimits {
+	return ServiceMonitorLimits{
+		CheckerScanIntervalSec: 1,
+		WorkerLimit:            20,
+		MinIntervalSec:         1,
+		DefaultIntervalSec:     1,
+		MinTimeoutSec:          1,
+		DefaultTimeoutSec:      5,
+		MaxTimeoutSec:          60,
+	}
+}
+
+// ServiceMonitorLimitsFromConfigMap parses limits from vite_config values.
+// Missing/invalid values fall back to defaults.
+func ServiceMonitorLimitsFromConfigMap(cfg map[string]string) ServiceMonitorLimits {
+	limits := DefaultServiceMonitorLimits()
+	if cfg == nil {
+		return limits
+	}
+
+	limits.CheckerScanIntervalSec = parseConfigInt(cfg, ConfigServiceMonitorCheckerScanIntervalSec, limits.CheckerScanIntervalSec)
+	limits.WorkerLimit = parseConfigInt(cfg, ConfigServiceMonitorWorkerLimit, limits.WorkerLimit)
+	limits.MinIntervalSec = parseConfigInt(cfg, ConfigServiceMonitorMinIntervalSec, limits.MinIntervalSec)
+	limits.DefaultIntervalSec = parseConfigInt(cfg, ConfigServiceMonitorDefaultIntervalSec, limits.DefaultIntervalSec)
+	limits.MinTimeoutSec = parseConfigInt(cfg, ConfigServiceMonitorMinTimeoutSec, limits.MinTimeoutSec)
+	limits.DefaultTimeoutSec = parseConfigInt(cfg, ConfigServiceMonitorDefaultTimeoutSec, limits.DefaultTimeoutSec)
+	limits.MaxTimeoutSec = parseConfigInt(cfg, ConfigServiceMonitorMaxTimeoutSec, limits.MaxTimeoutSec)
+
+	return normalizeServiceMonitorLimits(limits)
+}
+
+func normalizeServiceMonitorLimits(limits ServiceMonitorLimits) ServiceMonitorLimits {
+	if limits.CheckerScanIntervalSec <= 0 {
+		limits.CheckerScanIntervalSec = 30
+	}
+	if limits.WorkerLimit <= 0 {
+		limits.WorkerLimit = 5
+	}
+	if limits.WorkerLimit > 50 {
+		limits.WorkerLimit = 50
+	}
+
+	if limits.MinIntervalSec <= 0 {
+		limits.MinIntervalSec = limits.CheckerScanIntervalSec
+	}
+	if limits.MinIntervalSec < limits.CheckerScanIntervalSec {
+		limits.MinIntervalSec = limits.CheckerScanIntervalSec
+	}
+	if limits.DefaultIntervalSec <= 0 {
+		limits.DefaultIntervalSec = 60
+	}
+	if limits.DefaultIntervalSec < limits.MinIntervalSec {
+		limits.DefaultIntervalSec = limits.MinIntervalSec
+	}
+
+	if limits.MinTimeoutSec <= 0 {
+		limits.MinTimeoutSec = 1
+	}
+	if limits.DefaultTimeoutSec <= 0 {
+		limits.DefaultTimeoutSec = 5
+	}
+	if limits.DefaultTimeoutSec < limits.MinTimeoutSec {
+		limits.DefaultTimeoutSec = limits.MinTimeoutSec
+	}
+	if limits.MaxTimeoutSec <= 0 {
+		limits.MaxTimeoutSec = 60
+	}
+	if limits.MaxTimeoutSec < limits.DefaultTimeoutSec {
+		limits.MaxTimeoutSec = limits.DefaultTimeoutSec
+	}
+
+	return limits
+}
+
+func parseConfigInt(cfg map[string]string, key string, fallback int) int {
+	v := strings.TrimSpace(cfg[key])
+	if v == "" {
+		return fallback
+	}
+	n, err := strconv.Atoi(v)
+	if err != nil {
+		return fallback
+	}
+	return n
+}

go-backend/internal/store/model/model.go

@@ -235,6 +235,16 @@ type GroupPermissionGrant struct {
 
 func (GroupPermissionGrant) TableName() string { return "group_permission_grant" }
 
+// MonitorPermission grants a non-admin user access to monitoring endpoints.
+// One row per user_id.
+type MonitorPermission struct {
+	ID          int64 `gorm:"primaryKey;autoIncrement" json:"id"`
+	UserID      int64 `gorm:"column:user_id;not null;uniqueIndex:idx_monitor_permission_user" json:"userId"`
+	CreatedTime int64 `gorm:"column:created_time;not null" json:"createdTime"`
+}
+
+func (MonitorPermission) TableName() string { return "monitor_permission" }
+
 type ViteConfig struct {
 	ID    int64  `gorm:"primaryKey;autoIncrement" json:"id"`
 	Name  string `gorm:"type:varchar(200);not null;uniqueIndex" json:"name"`
@@ -641,3 +651,84 @@ type UserForwardDetail struct {
 	Status     int
 	CreatedAt  int64
 }
+
+type NodeMetric struct {
+	ID          int64   `gorm:"primaryKey;autoIncrement" json:"id"`
+	NodeID      int64   `gorm:"column:node_id;not null;index:idx_node_metric_node_time,priority:1" json:"nodeId"`
+	Timestamp   int64   `gorm:"not null;index:idx_node_metric_node_time,priority:2;index:idx_node_metric_time" json:"timestamp"`
+	CPUUsage    float64 `gorm:"column:cpu_usage" json:"cpuUsage"`
+	MemUsage    float64 `gorm:"column:mem_usage" json:"memoryUsage"`
+	DiskUsage   float64 `gorm:"column:disk_usage" json:"diskUsage"`
+	NetInBytes  int64   `gorm:"column:net_in_bytes" json:"netInBytes"`
+	NetOutBytes int64   `gorm:"column:net_out_bytes" json:"netOutBytes"`
+	NetInSpeed  int64   `gorm:"column:net_in_speed" json:"netInSpeed"`
+	NetOutSpeed int64   `gorm:"column:net_out_speed" json:"netOutSpeed"`
+	Load1       float64 `gorm:"column:load1" json:"load1"`
+	Load5       float64 `gorm:"column:load5" json:"load5"`
+	Load15      float64 `gorm:"column:load15" json:"load15"`
+	TCPConns    int64   `gorm:"column:tcp_conns" json:"tcpConns"`
+	UDPConns    int64   `gorm:"column:udp_conns" json:"udpConns"`
+	Uptime      int64   `gorm:"column:uptime" json:"uptime"`
+}
+
+func (NodeMetric) TableName() string { return "node_metric" }
+
+type TunnelMetric struct {
+	ID           int64   `gorm:"primaryKey;autoIncrement" json:"id"`
+	TunnelID     int64   `gorm:"column:tunnel_id;not null;uniqueIndex:idx_tunnel_metric_tunnel_time,priority:1" json:"tunnelId"`
+	NodeID       int64   `gorm:"column:node_id;not null;uniqueIndex:idx_tunnel_metric_tunnel_time,priority:2" json:"nodeId"`
+	Timestamp    int64   `gorm:"not null;uniqueIndex:idx_tunnel_metric_tunnel_time,priority:3;index:idx_tunnel_metric_time" json:"timestamp"`
+	BytesIn      int64   `gorm:"column:bytes_in" json:"bytesIn"`
+	BytesOut     int64   `gorm:"column:bytes_out" json:"bytesOut"`
+	Connections  int64   `gorm:"column:connections" json:"connections"`
+	Errors       int64   `gorm:"column:errors" json:"errors"`
+	AvgLatencyMs float64 `gorm:"column:avg_latency_ms" json:"avgLatencyMs"`
+}
+
+func (TunnelMetric) TableName() string { return "tunnel_metric" }
+
+type ServiceMonitor struct {
+	ID          int64  `gorm:"primaryKey;autoIncrement" json:"id"`
+	Name        string `gorm:"type:varchar(100);not null" json:"name"`
+	Type        string `gorm:"type:varchar(20);not null" json:"type"`
+	Target      string `gorm:"type:text;not null" json:"target"`
+	IntervalSec int    `gorm:"column:interval_sec;not null;default:60" json:"intervalSec"`
+	TimeoutSec  int    `gorm:"column:timeout_sec;not null;default:5" json:"timeoutSec"`
+	NodeID      int64  `gorm:"column:node_id;index" json:"nodeId"`
+	Enabled     int    `gorm:"not null;default:1" json:"enabled"`
+	CreatedTime int64  `gorm:"column:created_time;not null" json:"createdTime"`
+	UpdatedTime int64  `gorm:"column:updated_time;not null" json:"updatedTime"`
+}
+
+func (ServiceMonitor) TableName() string { return "service_monitor" }
+
+type ServiceMonitorResult struct {
+	ID           int64   `gorm:"primaryKey;autoIncrement" json:"id"`
+	MonitorID    int64   `gorm:"column:monitor_id;not null;index:idx_monitor_result_monitor_time,priority:1" json:"monitorId"`
+	NodeID       int64   `gorm:"column:node_id;not null;index" json:"nodeId"`
+	Timestamp    int64   `gorm:"not null;index:idx_monitor_result_monitor_time,priority:2" json:"timestamp"`
+	Success      int     `gorm:"not null" json:"success"`
+	LatencyMs    float64 `gorm:"column:latency_ms" json:"latencyMs"`
+	StatusCode   int     `gorm:"column:status_code" json:"statusCode"`
+	ErrorMessage string  `gorm:"column:error_message;type:text" json:"errorMessage"`
+}
+
+func (ServiceMonitorResult) TableName() string { return "service_monitor_result" }
+
+// TunnelQuality stores periodic probe results for a tunnel.
+// Unlike the old upsert model, rows accumulate for history/charting.
+// Old rows are pruned periodically (default: keep 24h).
+type TunnelQuality struct {
+	ID                 int64   `gorm:"primaryKey;autoIncrement" json:"id"`
+	TunnelID           int64   `gorm:"column:tunnel_id;not null;index:idx_tunnel_quality_tunnel_time,priority:1" json:"tunnelId"`
+	EntryToExitLatency float64 `gorm:"column:entry_to_exit_latency" json:"entryToExitLatency"`
+	ExitToBingLatency  float64 `gorm:"column:exit_to_bing_latency" json:"exitToBingLatency"`
+	EntryToExitLoss    float64 `gorm:"column:entry_to_exit_loss" json:"entryToExitLoss"`
+	ExitToBingLoss     float64 `gorm:"column:exit_to_bing_loss" json:"exitToBingLoss"`
+	Success            int     `gorm:"not null;default:1" json:"success"`
+	ErrorMessage       string  `gorm:"column:error_message;type:text" json:"errorMessage,omitempty"`
+	Timestamp          int64   `gorm:"not null;index:idx_tunnel_quality_tunnel_time,priority:2;index:idx_tunnel_quality_time" json:"timestamp"`
+	ChainDetails       string  `gorm:"column:chain_details;type:text" json:"chainDetails,omitempty"`
+}
+
+func (TunnelQuality) TableName() string { return "tunnel_quality" }

go-backend/internal/store/repo/repository.go

@@ -5,9 +5,11 @@ import (
 	"errors"
 	"fmt"
 	"log"
+	"net"
 	"os"
 	"path/filepath"
 	"sort"
+	"strconv"
 	"strings"
 	"time"
 
@@ -47,6 +49,11 @@ type UserGroupBackup = model.UserGroupBackup
 type PermissionBackup = model.PermissionBackup
 type PermissionGrantBackup = model.PermissionGrantBackup
 type ImportResult = model.ImportResult
+type NodeMetric = model.NodeMetric
+type TunnelMetric = model.TunnelMetric
+type ServiceMonitor = model.ServiceMonitor
+type ServiceMonitorResult = model.ServiceMonitorResult
+type TunnelQuality = model.TunnelQuality
 
 // ─── Repository ──────────────────────────────────────────────────────
 
@@ -176,12 +183,18 @@ func autoMigrateAll(db *gorm.DB) error {
 		&model.UserGroupUser{},
 		&model.GroupPermission{},
 		&model.GroupPermissionGrant{},
+		&model.MonitorPermission{},
 		&model.ViteConfig{},
 		&model.PeerShare{},
 		&model.PeerShareRuntime{},
 		&model.FederationTunnelBinding{},
 		&model.Announcement{},
 		&model.SchemaVersion{},
+		&model.NodeMetric{},
+		&model.TunnelMetric{},
+		&model.ServiceMonitor{},
+		&model.ServiceMonitorResult{},
+		&model.TunnelQuality{},
 	}
 
 	if db.Dialector.Name() != "sqlite" {
@@ -394,6 +407,24 @@ func (r *Repository) ListConfigs() (map[string]string, error) {
 	return result, nil
 }
 
+func (r *Repository) GetConfigsByNames(names []string) (map[string]string, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	if len(names) == 0 {
+		return map[string]string{}, nil
+	}
+	var configs []model.ViteConfig
+	if err := r.db.Select("name", "value").Where("name IN ?", names).Find(&configs).Error; err != nil {
+		return nil, err
+	}
+	result := make(map[string]string, len(configs))
+	for _, c := range configs {
+		result[c.Name] = c.Value
+	}
+	return result, nil
+}
+
 func (r *Repository) UpsertConfig(name, value string, now int64) error {
 	if r == nil || r.db == nil {
 		return errors.New("repository not initialized")
@@ -648,9 +679,11 @@ func (r *Repository) ListNodes() ([]map[string]interface{}, error) {
 			"version":       nullableString(n.Version),
 			"http":          n.HTTP, "tls": n.TLS, "socks": n.Socks,
 			"status": n.Status, "isRemote": n.IsRemote,
-			"remoteUrl":    nullableString(n.RemoteURL),
-			"remoteToken":  nullableString(n.RemoteToken),
-			"remoteConfig": nullableString(n.RemoteConfig),
+			"remoteUrl":                 nullableString(n.RemoteURL),
+			"remoteToken":               nullableString(n.RemoteToken),
+			"remoteConfig":              nullableString(n.RemoteConfig),
+			"expiryReminderDismissed":   n.ExpiryReminderDismissed,
+			"interfaceName":             nullableString(n.InterfaceName),
 		})
 	}
 	return items, nil
@@ -2689,12 +2722,13 @@ func (r *Repository) GetUserTunnelByID(id int64) (*model.UserTunnel, error) {
 
 // ─── Migration ───────────────────────────────────────────────────────
 
-const currentSchemaVersion = 5
+const currentSchemaVersion = 6
 
 var ensurePostgresIDDefaultsFn = ensurePostgresIDDefaults
 var migrateViteConfigValueColumnTypeFn = migrateViteConfigValueColumnType
 var migrateSpeedLimitTunnelBindingFn = migrateSpeedLimitTunnelBinding
 var migratePostgresTrafficInt64ColumnsFn = migratePostgresTrafficInt64Columns
+var migrateTunnelMetricBucketUniqueIndexFn = migrateTunnelMetricBucketUniqueIndex
 
 func getSchemaVersion(db *gorm.DB) int {
 	var v model.SchemaVersion
@@ -2764,6 +2798,12 @@ func migrateSchema(db *gorm.DB) error {
 		}
 	}
 
+	if ver < 6 {
+		if err := migrateTunnelMetricBucketUniqueIndexFn(db); err != nil {
+			return err
+		}
+	}
+
 	setSchemaVersion(db, currentSchemaVersion)
 	return nil
 }
@@ -2867,6 +2907,130 @@ func migratePostgresTrafficInt64Columns(db *gorm.DB) error {
 	return nil
 }
 
+func migrateTunnelMetricBucketUniqueIndex(db *gorm.DB) error {
+	if db == nil {
+		return errors.New("nil db")
+	}
+
+	if !db.Migrator().HasTable(&model.TunnelMetric{}) {
+		return nil
+	}
+
+	return db.Transaction(func(tx *gorm.DB) error {
+		// Only do the heavier dedupe work when needed.
+		var dupGroups int64
+		q := `
+			SELECT COUNT(1) AS cnt
+			FROM (
+				SELECT 1
+				FROM tunnel_metric
+				GROUP BY tunnel_id, node_id, timestamp
+				HAVING COUNT(*) > 1
+			) t
+		`
+		if err := tx.Raw(q).Scan(&dupGroups).Error; err != nil {
+			return fmt.Errorf("inspect tunnel_metric duplicates: %w", err)
+		}
+
+		if dupGroups > 0 {
+			switch tx.Dialector.Name() {
+			case "postgres":
+				sql := `
+					WITH agg AS (
+						SELECT MIN(id) AS keep_id,
+						       tunnel_id,
+						       node_id,
+						       timestamp,
+						       SUM(bytes_in) AS bytes_in,
+						       SUM(bytes_out) AS bytes_out,
+						       SUM(connections) AS connections,
+						       SUM(errors) AS errors,
+						       AVG(avg_latency_ms) AS avg_latency_ms
+						FROM tunnel_metric
+						GROUP BY tunnel_id, node_id, timestamp
+						HAVING COUNT(*) > 1
+					), updated AS (
+						UPDATE tunnel_metric tm
+						SET bytes_in = agg.bytes_in,
+						    bytes_out = agg.bytes_out,
+						    connections = agg.connections,
+						    errors = agg.errors,
+						    avg_latency_ms = agg.avg_latency_ms
+						FROM agg
+						WHERE tm.id = agg.keep_id
+						RETURNING tm.id
+					)
+					DELETE FROM tunnel_metric tm
+					USING agg
+					WHERE tm.tunnel_id = agg.tunnel_id
+					  AND tm.node_id = agg.node_id
+					  AND tm.timestamp = agg.timestamp
+					  AND tm.id <> agg.keep_id
+				`
+				if err := tx.Exec(sql).Error; err != nil {
+					return fmt.Errorf("dedupe tunnel_metric buckets: %w", err)
+				}
+			default:
+				// SQLite (and other) path.
+				if err := tx.Exec(`DROP TABLE IF EXISTS tunnel_metric_dedupe`).Error; err != nil {
+					return fmt.Errorf("prepare tunnel_metric dedupe table: %w", err)
+				}
+				if err := tx.Exec(`
+					CREATE TEMP TABLE tunnel_metric_dedupe AS
+					SELECT MIN(id) AS keep_id,
+					       tunnel_id,
+					       node_id,
+					       timestamp,
+					       SUM(bytes_in) AS bytes_in,
+					       SUM(bytes_out) AS bytes_out,
+					       SUM(connections) AS connections,
+					       SUM(errors) AS errors,
+					       AVG(avg_latency_ms) AS avg_latency_ms
+					FROM tunnel_metric
+					GROUP BY tunnel_id, node_id, timestamp
+					HAVING COUNT(*) > 1
+				`).Error; err != nil {
+					return fmt.Errorf("build tunnel_metric dedupe table: %w", err)
+				}
+				if err := tx.Exec(`
+					UPDATE tunnel_metric
+					SET bytes_in = (SELECT bytes_in FROM tunnel_metric_dedupe d WHERE d.keep_id = tunnel_metric.id),
+					    bytes_out = (SELECT bytes_out FROM tunnel_metric_dedupe d WHERE d.keep_id = tunnel_metric.id),
+					    connections = (SELECT connections FROM tunnel_metric_dedupe d WHERE d.keep_id = tunnel_metric.id),
+					    errors = (SELECT errors FROM tunnel_metric_dedupe d WHERE d.keep_id = tunnel_metric.id),
+					    avg_latency_ms = (SELECT avg_latency_ms FROM tunnel_metric_dedupe d WHERE d.keep_id = tunnel_metric.id)
+					WHERE id IN (SELECT keep_id FROM tunnel_metric_dedupe)
+				`).Error; err != nil {
+					return fmt.Errorf("update tunnel_metric deduped rows: %w", err)
+				}
+				if err := tx.Exec(`
+					DELETE FROM tunnel_metric
+					WHERE id IN (
+						SELECT tm.id
+						FROM tunnel_metric tm
+						JOIN tunnel_metric_dedupe d
+						  ON tm.tunnel_id = d.tunnel_id
+						 AND tm.node_id = d.node_id
+						 AND tm.timestamp = d.timestamp
+						WHERE tm.id <> d.keep_id
+					)
+				`).Error; err != nil {
+					return fmt.Errorf("delete tunnel_metric duplicates: %w", err)
+				}
+				_ = tx.Exec(`DROP TABLE IF EXISTS tunnel_metric_dedupe`).Error
+			}
+		}
+
+		// Uniqueness is required for safe upsert on (tunnel_id, node_id, timestamp).
+		if err := tx.Exec(
+			`CREATE UNIQUE INDEX IF NOT EXISTS uidx_tunnel_metric_bucket ON tunnel_metric(tunnel_id, node_id, timestamp)`,
+		).Error; err != nil {
+			return fmt.Errorf("create tunnel_metric unique index: %w", err)
+		}
+		return nil
+	})
+}
+
 func alterPostgresColumnToBigIntIfNeeded(db *gorm.DB, tableName, columnName string) error {
 	if db == nil {
 		return errors.New("nil db")
@@ -3080,7 +3244,7 @@ func resolveForwardIngress(db *gorm.DB, forwardID int64, tunnelID int64) (string
 		}
 
 		if ip != "" {
-			pair := fmt.Sprintf("%s:%d", ip, row.Port.Int64)
+			pair := formatForwardIngressAddress(ip, row.Port.Int64)
 			if _, ok := seenPairs[pair]; !ok {
 				seenPairs[pair] = struct{}{}
 				entries = append(entries, pair)
@@ -3097,6 +3261,17 @@ func resolveForwardIngress(db *gorm.DB, forwardID int64, tunnelID int64) (string
 	return strings.Join(entries, ","), inPort, nil
 }
 
+func formatForwardIngressAddress(host string, port int64) string {
+	host = strings.TrimSpace(host)
+	if host == "" || port <= 0 {
+		return ""
+	}
+	if strings.HasPrefix(host, "[") && strings.HasSuffix(host, "]") {
+		host = strings.TrimPrefix(strings.TrimSuffix(host, "]"), "[")
+	}
+	return net.JoinHostPort(host, strconv.FormatInt(port, 10))
+}
+
 func nullableString(v sql.NullString) interface{} {
 	if v.Valid {
 		return v.String
@@ -3140,3 +3315,395 @@ var osMkdirAll = func(path string) error {
 
 // Suppress unused import warning for log
 var _ = log.Printf
+
+func (r *Repository) InsertNodeMetric(m *model.NodeMetric) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Create(m).Error
+}
+
+func (r *Repository) InsertNodeMetricBatch(metrics []*model.NodeMetric) error {
+	if r == nil || r.db == nil || len(metrics) == 0 {
+		return nil
+	}
+	return r.db.CreateInBatches(metrics, 100).Error
+}
+
+func (r *Repository) GetNodeMetrics(nodeID int64, startMs, endMs int64) ([]model.NodeMetric, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+
+	rangeMs := endMs - startMs
+	const maxRawRangeMs = int64(60 * 60 * 1000) // 1 hour — return raw data for short ranges
+	const targetPoints = 500                     // target number of chart points for downsampled data
+
+	// For short ranges, return raw data (full resolution).
+	if rangeMs <= maxRawRangeMs {
+		var metrics []model.NodeMetric
+		err := r.db.Where("node_id = ? AND timestamp >= ? AND timestamp <= ?", nodeID, startMs, endMs).
+			Order("timestamp ASC").
+			Limit(5000).
+			Find(&metrics).Error
+		return metrics, err
+	}
+
+	// For longer ranges, downsample via SQL aggregation to keep the response small and fast.
+	bucketMs := rangeMs / targetPoints
+	if bucketMs < 1000 {
+		bucketMs = 1000 // minimum 1-second buckets
+	}
+
+	bucketExpr := fmt.Sprintf("(timestamp / %d * %d)", bucketMs, bucketMs)
+	groupExpr := fmt.Sprintf("timestamp / %d", bucketMs)
+
+	var metrics []model.NodeMetric
+	err := r.db.Model(&model.NodeMetric{}).
+		Select(
+			fmt.Sprintf(
+				"%d AS node_id, "+
+					"CAST(%s AS BIGINT) AS timestamp, "+
+					"AVG(cpu_usage) AS cpu_usage, "+
+					"AVG(mem_usage) AS mem_usage, "+
+					"AVG(disk_usage) AS disk_usage, "+
+					"CAST(AVG(net_in_bytes) AS BIGINT) AS net_in_bytes, "+
+					"CAST(AVG(net_out_bytes) AS BIGINT) AS net_out_bytes, "+
+					"CAST(AVG(net_in_speed) AS BIGINT) AS net_in_speed, "+
+					"CAST(AVG(net_out_speed) AS BIGINT) AS net_out_speed, "+
+					"AVG(load1) AS load1, "+
+					"AVG(load5) AS load5, "+
+					"AVG(load15) AS load15, "+
+					"CAST(AVG(tcp_conns) AS BIGINT) AS tcp_conns, "+
+					"CAST(AVG(udp_conns) AS BIGINT) AS udp_conns, "+
+					"CAST(MAX(uptime) AS BIGINT) AS uptime",
+				nodeID, bucketExpr,
+			),
+		).
+		Where("node_id = ? AND timestamp >= ? AND timestamp <= ?", nodeID, startMs, endMs).
+		Group(groupExpr).
+		Order("timestamp ASC").
+		Limit(targetPoints + 100). // safety margin
+		Scan(&metrics).Error
+
+	if metrics == nil {
+		metrics = make([]model.NodeMetric, 0)
+	}
+	return metrics, err
+}
+
+func (r *Repository) GetLatestNodeMetric(nodeID int64) (*model.NodeMetric, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var m model.NodeMetric
+	err := r.db.Where("node_id = ?", nodeID).Order("timestamp DESC").First(&m).Error
+	if err != nil {
+		if errors.Is(err, gorm.ErrRecordNotFound) {
+			return nil, nil
+		}
+		return nil, err
+	}
+	return &m, nil
+}
+
+func (r *Repository) PruneNodeMetrics(olderThanMs int64) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Where("timestamp < ?", olderThanMs).Delete(&model.NodeMetric{}).Error
+}
+
+func (r *Repository) InsertTunnelMetric(m *model.TunnelMetric) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Create(m).Error
+}
+
+func (r *Repository) InsertTunnelMetricBatch(metrics []*model.TunnelMetric) error {
+	if r == nil || r.db == nil || len(metrics) == 0 {
+		return nil
+	}
+	return r.db.CreateInBatches(metrics, 100).Error
+}
+
+// UpsertTunnelMetricBuckets adds the provided metric deltas into per-minute buckets.
+// Requires a unique index on (tunnel_id, node_id, timestamp) for safe upserts.
+func (r *Repository) UpsertTunnelMetricBuckets(metrics []*model.TunnelMetric) error {
+	if r == nil || r.db == nil || len(metrics) == 0 {
+		return nil
+	}
+
+	// Postgres rejects a single INSERT ... ON CONFLICT when the input contains
+	// duplicate conflict keys. Pre-aggregate within this batch to keep inserts safe.
+	type bucketKey struct {
+		tunnelID  int64
+		nodeID    int64
+		timestamp int64
+	}
+
+	agg := make(map[bucketKey]*model.TunnelMetric, len(metrics))
+	for _, m := range metrics {
+		if m == nil {
+			continue
+		}
+		if m.TunnelID <= 0 || m.NodeID <= 0 || m.Timestamp <= 0 {
+			continue
+		}
+		if m.BytesIn == 0 && m.BytesOut == 0 && m.Connections == 0 && m.Errors == 0 {
+			continue
+		}
+
+		k := bucketKey{tunnelID: m.TunnelID, nodeID: m.NodeID, timestamp: m.Timestamp}
+		if existing, ok := agg[k]; ok {
+			existing.BytesIn += m.BytesIn
+			existing.BytesOut += m.BytesOut
+			existing.Connections += m.Connections
+			existing.Errors += m.Errors
+			if existing.AvgLatencyMs == 0 && m.AvgLatencyMs != 0 {
+				existing.AvgLatencyMs = m.AvgLatencyMs
+			}
+			continue
+		}
+		cp := *m
+		agg[k] = &cp
+	}
+	if len(agg) == 0 {
+		return nil
+	}
+
+	rows := make([]*model.TunnelMetric, 0, len(agg))
+	for _, v := range agg {
+		rows = append(rows, v)
+	}
+
+	return r.db.Clauses(clause.OnConflict{
+		Columns: []clause.Column{{Name: "tunnel_id"}, {Name: "node_id"}, {Name: "timestamp"}},
+		DoUpdates: clause.Assignments(map[string]interface{}{
+			"bytes_in":    gorm.Expr("tunnel_metric.bytes_in + excluded.bytes_in"),
+			"bytes_out":   gorm.Expr("tunnel_metric.bytes_out + excluded.bytes_out"),
+			"connections": gorm.Expr("tunnel_metric.connections + excluded.connections"),
+			"errors":      gorm.Expr("tunnel_metric.errors + excluded.errors"),
+			// avg_latency_ms is not additive; keep the existing bucket value.
+		}),
+	}).CreateInBatches(rows, 100).Error
+}
+
+func (r *Repository) GetTunnelMetrics(tunnelID int64, startMs, endMs int64) ([]model.TunnelMetric, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var metrics []model.TunnelMetric
+	err := r.db.Where("tunnel_id = ? AND timestamp >= ? AND timestamp <= ?", tunnelID, startMs, endMs).
+		Order("timestamp DESC").
+		Limit(5000).
+		Find(&metrics).Error
+	if len(metrics) > 1 {
+		for i, j := 0, len(metrics)-1; i < j; i, j = i+1, j-1 {
+			metrics[i], metrics[j] = metrics[j], metrics[i]
+		}
+	}
+	return metrics, err
+}
+
+// GetTunnelMetricsAggregated returns tunnel-level aggregated series (one point per timestamp).
+// Storage remains per (tunnel_id, node_id, timestamp) for future drill-down.
+func (r *Repository) GetTunnelMetricsAggregated(tunnelID int64, startMs, endMs int64) ([]model.TunnelMetric, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+
+	var metrics []model.TunnelMetric
+	err := r.db.Model(&model.TunnelMetric{}).
+		Select(
+			"tunnel_id, 0 AS node_id, timestamp, "+
+				"SUM(bytes_in) AS bytes_in, "+
+				"SUM(bytes_out) AS bytes_out, "+
+				"SUM(connections) AS connections, "+
+				"SUM(errors) AS errors, "+
+				"AVG(avg_latency_ms) AS avg_latency_ms",
+		).
+		Where("tunnel_id = ? AND timestamp >= ? AND timestamp <= ?", tunnelID, startMs, endMs).
+		Group("tunnel_id, timestamp").
+		Order("timestamp ASC").
+		Limit(5000).
+		Scan(&metrics).Error
+	if metrics == nil {
+		metrics = make([]model.TunnelMetric, 0)
+	}
+	return metrics, err
+}
+
+func (r *Repository) PruneTunnelMetrics(olderThanMs int64) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Where("timestamp < ?", olderThanMs).Delete(&model.TunnelMetric{}).Error
+}
+
+func (r *Repository) ListServiceMonitors() ([]model.ServiceMonitor, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var monitors []model.ServiceMonitor
+	err := r.db.Order("id ASC").Find(&monitors).Error
+	return monitors, err
+}
+
+func (r *Repository) ListEnabledServiceMonitors() ([]model.ServiceMonitor, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var monitors []model.ServiceMonitor
+	err := r.db.Where("enabled = 1 AND type IN (?)", []string{"tcp", "icmp"}).Order("id ASC").Find(&monitors).Error
+	return monitors, err
+}
+
+func (r *Repository) GetServiceMonitor(id int64) (*model.ServiceMonitor, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var m model.ServiceMonitor
+	err := r.db.First(&m, id).Error
+	if err != nil {
+		if errors.Is(err, gorm.ErrRecordNotFound) {
+			return nil, nil
+		}
+		return nil, err
+	}
+	return &m, nil
+}
+
+func (r *Repository) CreateServiceMonitor(m *model.ServiceMonitor) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Create(m).Error
+}
+
+func (r *Repository) UpdateServiceMonitor(m *model.ServiceMonitor) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Save(m).Error
+}
+
+func (r *Repository) DeleteServiceMonitor(id int64) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	if id <= 0 {
+		return nil
+	}
+
+	// Keep API/UI semantics simple: deleting a monitor also deletes its history.
+	return r.db.Transaction(func(tx *gorm.DB) error {
+		if err := tx.Where("monitor_id = ?", id).Delete(&model.ServiceMonitorResult{}).Error; err != nil {
+			return err
+		}
+		return tx.Delete(&model.ServiceMonitor{}, id).Error
+	})
+}
+
+func (r *Repository) InsertServiceMonitorResult(result *model.ServiceMonitorResult) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Create(result).Error
+}
+
+func (r *Repository) GetServiceMonitorResults(monitorID int64, limit int) ([]model.ServiceMonitorResult, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	if limit <= 0 {
+		limit = 100
+	}
+	var results []model.ServiceMonitorResult
+	err := r.db.Where("monitor_id = ?", monitorID).
+		Order("timestamp DESC").
+		Limit(limit).
+		Find(&results).Error
+	return results, err
+}
+
+// GetServiceMonitorResultsByTimeRange returns results for a monitor within [startMs, endMs].
+// Mirrors GetNodeMetrics / GetTunnelMetrics pattern for time-range based charting.
+func (r *Repository) GetServiceMonitorResultsByTimeRange(monitorID int64, startMs, endMs int64) ([]model.ServiceMonitorResult, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+	var results []model.ServiceMonitorResult
+	err := r.db.Where("monitor_id = ? AND timestamp >= ? AND timestamp <= ?", monitorID, startMs, endMs).
+		Order("timestamp DESC").
+		Limit(5000).
+		Find(&results).Error
+	if len(results) > 1 {
+		for i, j := 0, len(results)-1; i < j; i, j = i+1, j-1 {
+			results[i], results[j] = results[j], results[i]
+		}
+	}
+	return results, err
+}
+
+// GetLatestServiceMonitorResults returns the newest result per monitor_id.
+// This is intended for list rendering (avoid N+1 queries).
+func (r *Repository) GetLatestServiceMonitorResults() ([]model.ServiceMonitorResult, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+
+	var results []model.ServiceMonitorResult
+
+	// Prefer a window-function query (works on modern SQLite + Postgres).
+	q1 := `
+		SELECT id, monitor_id, node_id, timestamp, success, latency_ms, status_code, error_message
+		FROM (
+			SELECT *, ROW_NUMBER() OVER (PARTITION BY monitor_id ORDER BY timestamp DESC, id DESC) AS rn
+			FROM service_monitor_result
+		) t
+		WHERE rn = 1
+		ORDER BY monitor_id ASC
+	`
+	if err := r.db.Raw(q1).Scan(&results).Error; err == nil {
+		return results, nil
+	}
+
+	// Fallback: just return newest rows (best-effort). This avoids hard failure on older SQLite builds.
+	// Note: This may not include all monitors if the table is extremely large and skewed.
+	results = nil
+	q2 := `
+		SELECT id, monitor_id, node_id, timestamp, success, latency_ms, status_code, error_message
+		FROM service_monitor_result
+		ORDER BY timestamp DESC, id DESC
+		LIMIT 5000
+	`
+	err := r.db.Raw(q2).Scan(&results).Error
+	if err != nil {
+		return nil, err
+	}
+
+	seen := make(map[int64]struct{}, len(results))
+	out := make([]model.ServiceMonitorResult, 0, len(results))
+	for _, row := range results {
+		if row.MonitorID <= 0 {
+			continue
+		}
+		if _, ok := seen[row.MonitorID]; ok {
+			continue
+		}
+		seen[row.MonitorID] = struct{}{}
+		out = append(out, row)
+	}
+	// Keep response stable for the frontend.
+	sort.Slice(out, func(i, j int) bool { return out[i].MonitorID < out[j].MonitorID })
+	return out, nil
+}
+
+func (r *Repository) PruneServiceMonitorResults(olderThanMs int64) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Where("timestamp < ?", olderThanMs).Delete(&model.ServiceMonitorResult{}).Error
+}

go-backend/internal/store/repo/repository_control.go

@@ -64,6 +64,7 @@ func (r *Repository) ListForwardsByTunnelTx(tx *gorm.DB, tunnelID int64) ([]mode
 	return rows, nil
 }
 
+
 func (r *Repository) ListActiveTunnelIDsByNode(nodeID int64) ([]int64, error) {
 	if r == nil || r.db == nil {
 		return nil, errors.New("repository not initialized")
@@ -125,6 +126,7 @@ func (r *Repository) ListForwardPortsTx(tx *gorm.DB, forwardID int64) ([]model.F
 	return rows, nil
 }
 
+
 func (r *Repository) HasOtherForwardOnNodePort(nodeID int64, port int, currentForwardID int64) (bool, error) {
 	if r == nil || r.db == nil {
 		return false, errors.New("repository not initialized")
@@ -151,6 +153,7 @@ func (r *Repository) HasOtherForwardOnNodePortTx(tx *gorm.DB, nodeID int64, port
 	return count > 0, nil
 }
 
+
 func (r *Repository) GetTunnelOutProtocol(tunnelID int64) (string, error) {
 	if r == nil || r.db == nil {
 		return "", errors.New("repository not initialized")

go-backend/internal/store/repo/repository_flow.go

@@ -80,6 +80,37 @@ func (r *Repository) ListActiveForwardsByUserTunnel(userID, tunnelID int64) ([]m
 	return rows, nil
 }
 
+func (r *Repository) ListForwardsByUserAndTunnel(userID, tunnelID int64) ([]model.ForwardRecord, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var forwards []model.Forward
+	err := r.db.Where("user_id = ? AND tunnel_id = ?", userID, tunnelID).Order("id ASC").Find(&forwards).Error
+	if err != nil {
+		return nil, err
+	}
+	rows := make([]model.ForwardRecord, 0, len(forwards))
+	for _, f := range forwards {
+		rows = append(rows, model.ForwardRecord{
+			ID:         f.ID,
+			UserID:     f.UserID,
+			UserName:   f.UserName,
+			Name:       f.Name,
+			TunnelID:   f.TunnelID,
+			RemoteAddr: f.RemoteAddr,
+			Strategy:   f.Strategy,
+			Status:     f.Status,
+			SpeedID:    f.SpeedID,
+		})
+	}
+	for i := range rows {
+		if strings.TrimSpace(rows[i].Strategy) == "" {
+			rows[i].Strategy = "fifo"
+		}
+	}
+	return rows, nil
+}
+
 func (r *Repository) GetForwardRecord(forwardID int64) (*model.ForwardRecord, error) {
 	if r == nil || r.db == nil {
 		return nil, errors.New("repository not initialized")
@@ -161,6 +192,64 @@ func (r *Repository) ForwardExists(forwardID int64) (bool, error) {
 	return count > 0, nil
 }
 
+// MapForwardIDsToTunnelIDs returns a mapping from forward.id to forward.tunnel_id.
+// Missing forward IDs are omitted from the returned map.
+func (r *Repository) MapForwardIDsToTunnelIDs(forwardIDs []int64) (map[int64]int64, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	if len(forwardIDs) == 0 {
+		return map[int64]int64{}, nil
+	}
+
+	// Deduplicate and filter invalid IDs.
+	ids := make([]int64, 0, len(forwardIDs))
+	seen := make(map[int64]struct{}, len(forwardIDs))
+	for _, id := range forwardIDs {
+		if id <= 0 {
+			continue
+		}
+		if _, ok := seen[id]; ok {
+			continue
+		}
+		seen[id] = struct{}{}
+		ids = append(ids, id)
+	}
+	if len(ids) == 0 {
+		return map[int64]int64{}, nil
+	}
+
+	type row struct {
+		ID       int64 `gorm:"column:id"`
+		TunnelID int64 `gorm:"column:tunnel_id"`
+	}
+
+	out := make(map[int64]int64, len(ids))
+	const chunkSize = 500
+	for start := 0; start < len(ids); start += chunkSize {
+		end := start + chunkSize
+		if end > len(ids) {
+			end = len(ids)
+		}
+
+		var rows []row
+		if err := r.db.Model(&model.Forward{}).
+			Select("id", "tunnel_id").
+			Where("id IN ?", ids[start:end]).
+			Find(&rows).Error; err != nil {
+			return nil, err
+		}
+		for _, r := range rows {
+			if r.ID <= 0 || r.TunnelID <= 0 {
+				continue
+			}
+			out[r.ID] = r.TunnelID
+		}
+	}
+
+	return out, nil
+}
+
 func (r *Repository) SpeedLimitExists(id int64) (bool, error) {
 	if r == nil || r.db == nil {
 		return false, errors.New("repository not initialized")
@@ -173,6 +262,24 @@ func (r *Repository) SpeedLimitExists(id int64) (bool, error) {
 	return count > 0, nil
 }
 
+func (r *Repository) CountActiveForwardsByUser(userID int64) (int64, error) {
+	if r == nil || r.db == nil {
+		return 0, errors.New("repository not initialized")
+	}
+	var count int64
+	err := r.db.Model(&model.Forward{}).Where("user_id = ? AND status = 1", userID).Count(&count).Error
+	return count, err
+}
+
+func (r *Repository) CountActiveForwardsByUserTunnel(userID, tunnelID int64) (int64, error) {
+	if r == nil || r.db == nil {
+		return 0, errors.New("repository not initialized")
+	}
+	var count int64
+	err := r.db.Model(&model.Forward{}).Where("user_id = ? AND tunnel_id = ? AND status = 1", userID, tunnelID).Count(&count).Error
+	return count, err
+}
+
 func (r *Repository) GetSpeedLimitSpeed(id int64) (int, error) {
 	if r == nil || r.db == nil {
 		return 0, errors.New("repository not initialized")

go-backend/internal/store/repo/repository_monitor_nodes.go

@@ -0,0 +1,19 @@
+package repo
+
+import (
+	"errors"
+
+	"go-backend/internal/store/model"
+)
+
+func (r *Repository) ListMonitorNodes() ([]model.Node, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var nodes []model.Node
+	err := r.db.Select("id", "inx", "name", "status", "version", "updated_time").
+		Where("is_remote = ?", 0).
+		Order("inx ASC, id ASC").
+		Find(&nodes).Error
+	return nodes, err
+}

go-backend/internal/store/repo/repository_monitor_permission.go

@@ -0,0 +1,54 @@
+package repo
+
+import (
+	"errors"
+
+	"go-backend/internal/store/model"
+
+	"gorm.io/gorm/clause"
+)
+
+func (r *Repository) InsertMonitorPermission(userID int64, now int64) error {
+	if r == nil || r.db == nil {
+		return errors.New("repository not initialized")
+	}
+	if userID <= 0 {
+		return nil
+	}
+	row := model.MonitorPermission{UserID: userID, CreatedTime: now}
+	return r.db.Clauses(clause.OnConflict{DoNothing: true}).Create(&row).Error
+}
+
+func (r *Repository) DeleteMonitorPermission(userID int64) error {
+	if r == nil || r.db == nil {
+		return errors.New("repository not initialized")
+	}
+	if userID <= 0 {
+		return nil
+	}
+	return r.db.Where("user_id = ?", userID).Delete(&model.MonitorPermission{}).Error
+}
+
+func (r *Repository) HasMonitorPermission(userID int64) (bool, error) {
+	if r == nil || r.db == nil {
+		return false, errors.New("repository not initialized")
+	}
+	if userID <= 0 {
+		return false, nil
+	}
+	var count int64
+	err := r.db.Model(&model.MonitorPermission{}).Where("user_id = ?", userID).Count(&count).Error
+	if err != nil {
+		return false, err
+	}
+	return count > 0, nil
+}
+
+func (r *Repository) ListMonitorPermissions() ([]model.MonitorPermission, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var items []model.MonitorPermission
+	err := r.db.Order("id ASC").Find(&items).Error
+	return items, err
+}

go-backend/internal/store/repo/repository_monitor_tunnels.go

@@ -0,0 +1,18 @@
+package repo
+
+import (
+	"errors"
+
+	"go-backend/internal/store/model"
+)
+
+func (r *Repository) ListMonitorTunnels() ([]model.Tunnel, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var tunnels []model.Tunnel
+	err := r.db.Select("id", "inx", "name", "status", "updated_time").
+		Order("inx ASC, id ASC").
+		Find(&tunnels).Error
+	return tunnels, err
+}

go-backend/internal/store/repo/repository_monitoring_test.go

@@ -0,0 +1,134 @@
+package repo
+
+import (
+	"sync"
+	"testing"
+	"time"
+
+	"go-backend/internal/store/model"
+)
+
+func TestGetTunnelMetricsAggregatedSumsAcrossNodes(t *testing.T) {
+	r, err := Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	ts := time.Now().UnixMilli()
+
+	if err := r.InsertTunnelMetric(&model.TunnelMetric{
+		TunnelID:  1,
+		NodeID:    1,
+		Timestamp: ts,
+		BytesIn:   100,
+		BytesOut:  200,
+	}); err != nil {
+		t.Fatalf("insert tunnel metric n1: %v", err)
+	}
+	if err := r.InsertTunnelMetric(&model.TunnelMetric{
+		TunnelID:  1,
+		NodeID:    2,
+		Timestamp: ts,
+		BytesIn:   300,
+		BytesOut:  400,
+	}); err != nil {
+		t.Fatalf("insert tunnel metric n2: %v", err)
+	}
+
+	metrics, err := r.GetTunnelMetricsAggregated(1, ts-1000, ts+1000)
+	if err != nil {
+		t.Fatalf("get aggregated tunnel metrics: %v", err)
+	}
+	if len(metrics) != 1 {
+		t.Fatalf("expected 1 aggregated point, got %d", len(metrics))
+	}
+	if metrics[0].Timestamp != ts {
+		t.Fatalf("expected timestamp %d, got %d", ts, metrics[0].Timestamp)
+	}
+	if metrics[0].BytesIn != 400 {
+		t.Fatalf("expected bytesIn 400, got %d", metrics[0].BytesIn)
+	}
+	if metrics[0].BytesOut != 600 {
+		t.Fatalf("expected bytesOut 600, got %d", metrics[0].BytesOut)
+	}
+}
+
+func TestUpsertTunnelMetricBucketsAggregatesDuplicateKeysInBatch(t *testing.T) {
+	r, err := Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	ts := time.Now().UnixMilli()
+
+	items := []*model.TunnelMetric{
+		{TunnelID: 1, NodeID: 1, Timestamp: ts, BytesIn: 10, BytesOut: 20},
+		{TunnelID: 1, NodeID: 1, Timestamp: ts, BytesIn: 30, BytesOut: 40},
+	}
+	if err := r.UpsertTunnelMetricBuckets(items); err != nil {
+		t.Fatalf("upsert buckets: %v", err)
+	}
+
+	rows, err := r.GetTunnelMetrics(1, ts-1000, ts+1000)
+	if err != nil {
+		t.Fatalf("get tunnel metrics: %v", err)
+	}
+	if len(rows) != 1 {
+		t.Fatalf("expected 1 stored row, got %d", len(rows))
+	}
+	if rows[0].BytesIn != 40 {
+		t.Fatalf("expected bytesIn 40, got %d", rows[0].BytesIn)
+	}
+	if rows[0].BytesOut != 60 {
+		t.Fatalf("expected bytesOut 60, got %d", rows[0].BytesOut)
+	}
+}
+
+func TestUpsertTunnelMetricBucketsIsSafeUnderConcurrency(t *testing.T) {
+	r, err := Open(":memory:")
+	if err != nil {
+		t.Fatalf("open repo: %v", err)
+	}
+	defer r.Close()
+
+	ts := time.Now().UnixMilli()
+
+	const workers = 20
+	const perWorkerIn = int64(5)
+	const perWorkerOut = int64(7)
+
+	var wg sync.WaitGroup
+	wg.Add(workers)
+	for i := 0; i < workers; i++ {
+		go func() {
+			defer wg.Done()
+			_ = r.UpsertTunnelMetricBuckets([]*model.TunnelMetric{{
+				TunnelID:  1,
+				NodeID:    1,
+				Timestamp: ts,
+				BytesIn:   perWorkerIn,
+				BytesOut:  perWorkerOut,
+			}})
+		}()
+	}
+	wg.Wait()
+
+	rows, err := r.GetTunnelMetrics(1, ts-1000, ts+1000)
+	if err != nil {
+		t.Fatalf("get tunnel metrics: %v", err)
+	}
+	if len(rows) != 1 {
+		t.Fatalf("expected 1 stored row, got %d", len(rows))
+	}
+
+	wantIn := int64(workers) * perWorkerIn
+	wantOut := int64(workers) * perWorkerOut
+	if rows[0].BytesIn != wantIn {
+		t.Fatalf("expected bytesIn %d, got %d", wantIn, rows[0].BytesIn)
+	}
+	if rows[0].BytesOut != wantOut {
+		t.Fatalf("expected bytesOut %d, got %d", wantOut, rows[0].BytesOut)
+	}
+}

go-backend/internal/store/repo/repository_mutations.go

@@ -997,9 +997,16 @@ func (r *Repository) DeleteGroupPermissionByIDTx(tx *gorm.DB, id int64) error {
 	return tx.Where("id = ?", id).Delete(&model.GroupPermission{}).Error
 }
 
-func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID int64, previousUserIDs, currentUserIDs []int64) error {
+// RevokedUserTunnelPair holds the (userID, tunnelID) of a deleted user_tunnel row,
+// so the handler layer can clean up associated forwarding rules.
+type RevokedUserTunnelPair struct {
+	UserID   int64
+	TunnelID int64
+}
+
+func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID int64, previousUserIDs, currentUserIDs []int64) ([]RevokedUserTunnelPair, error) {
 	if tx == nil {
-		return errors.New("database unavailable")
+		return nil, errors.New("database unavailable")
 	}
 	currentSet := make(map[int64]struct{}, len(currentUserIDs))
 	for _, uid := range currentUserIDs {
@@ -1018,7 +1025,7 @@ func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID
 		}
 	}
 	if len(removedUserIDs) == 0 {
-		return nil
+		return nil, nil
 	}
 
 	type grantRow struct {
@@ -1026,6 +1033,8 @@ func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID
 		CreatedByGroup int
 	}
 
+	var revoked []RevokedUserTunnelPair
+
 	for _, userID := range removedUserIDs {
 		var rows []grantRow
 		if err := tx.Model(&model.GroupPermissionGrant{}).
@@ -1033,7 +1042,7 @@ func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID
 			Joins("JOIN user_tunnel ON user_tunnel.id = group_permission_grant.user_tunnel_id").
 			Where("group_permission_grant.user_group_id = ? AND user_tunnel.user_id = ?", userGroupID, userID).
 			Find(&rows).Error; err != nil {
-			return err
+			return revoked, err
 		}
 
 		groupCreatedTunnelIDs := make(map[int64]struct{})
@@ -1046,28 +1055,32 @@ func (r *Repository) RevokeGroupGrantsForRemovedUsersTx(tx *gorm.DB, userGroupID
 		userTunnelIDs := tx.Model(&model.UserTunnel{}).Select("id").Where("user_id = ?", userID)
 		if err := tx.Where("user_group_id = ? AND user_tunnel_id IN (?)", userGroupID, userTunnelIDs).
 			Delete(&model.GroupPermissionGrant{}).Error; err != nil {
-			return err
+			return revoked, err
 		}
 
 		for userTunnelID := range groupCreatedTunnelIDs {
 			var remaining int64
 			if err := tx.Model(&model.GroupPermissionGrant{}).Where("user_tunnel_id = ?", userTunnelID).Count(&remaining).Error; err != nil {
-				return err
+				return revoked, err
 			}
 			if remaining == 0 {
+				var ut model.UserTunnel
+				if lookupErr := tx.Select("user_id", "tunnel_id").Where("id = ?", userTunnelID).First(&ut).Error; lookupErr == nil {
+					revoked = append(revoked, RevokedUserTunnelPair{UserID: ut.UserID, TunnelID: ut.TunnelID})
+				}
 				if err := tx.Where("id = ?", userTunnelID).Delete(&model.UserTunnel{}).Error; err != nil {
-					return err
+					return revoked, err
 				}
 			}
 		}
 	}
 
-	return nil
+	return revoked, nil
 }
 
-func (r *Repository) RevokeGroupPermissionPairTx(tx *gorm.DB, userGroupID, tunnelGroupID int64) error {
+func (r *Repository) RevokeGroupPermissionPairTx(tx *gorm.DB, userGroupID, tunnelGroupID int64) ([]RevokedUserTunnelPair, error) {
 	if tx == nil {
-		return errors.New("database unavailable")
+		return nil, errors.New("database unavailable")
 	}
 
 	type grantRow struct {
@@ -1080,7 +1093,7 @@ func (r *Repository) RevokeGroupPermissionPairTx(tx *gorm.DB, userGroupID, tunne
 		Select("user_tunnel_id, created_by_group").
 		Where("user_group_id = ? AND tunnel_group_id = ?", userGroupID, tunnelGroupID).
 		Find(&rows).Error; err != nil {
-		return err
+		return nil, err
 	}
 
 	groupCreatedTunnelIDs := make(map[int64]struct{})
@@ -1092,22 +1105,27 @@ func (r *Repository) RevokeGroupPermissionPairTx(tx *gorm.DB, userGroupID, tunne
 
 	if err := tx.Where("user_group_id = ? AND tunnel_group_id = ?", userGroupID, tunnelGroupID).
 		Delete(&model.GroupPermissionGrant{}).Error; err != nil {
-		return err
+		return nil, err
 	}
 
+	var revoked []RevokedUserTunnelPair
 	for userTunnelID := range groupCreatedTunnelIDs {
 		var remaining int64
 		if err := tx.Model(&model.GroupPermissionGrant{}).Where("user_tunnel_id = ?", userTunnelID).Count(&remaining).Error; err != nil {
-			return err
+			return revoked, err
 		}
 		if remaining == 0 {
+			var ut model.UserTunnel
+			if lookupErr := tx.Select("user_id", "tunnel_id").Where("id = ?", userTunnelID).First(&ut).Error; lookupErr == nil {
+				revoked = append(revoked, RevokedUserTunnelPair{UserID: ut.UserID, TunnelID: ut.TunnelID})
+			}
 			if err := tx.Where("id = ?", userTunnelID).Delete(&model.UserTunnel{}).Error; err != nil {
-				return err
+				return revoked, err
 			}
 		}
 	}
 
-	return nil
+	return revoked, nil
 }
 
 func (r *Repository) ReplaceFederationTunnelBindingsTx(tx *gorm.DB, tunnelID int64, bindings []FederationTunnelBinding) error {

go-backend/internal/store/repo/repository_tunnel_quality.go

@@ -0,0 +1,98 @@
+package repo
+
+import (
+	"errors"
+
+	"go-backend/internal/store/model"
+)
+
+// InsertTunnelQuality appends a tunnel quality probe result.
+// (Follows the same pattern as InsertServiceMonitorResult.)
+func (r *Repository) InsertTunnelQuality(q *model.TunnelQuality) error {
+	if r == nil || r.db == nil {
+		return errors.New("repository not initialized")
+	}
+	if q == nil || q.TunnelID <= 0 {
+		return nil
+	}
+	return r.db.Create(q).Error
+}
+
+// GetTunnelQualityHistory returns quality probe results for a tunnel
+// within a time range, ordered by timestamp ascending.
+// (Mirrors GetServiceMonitorResults pattern.)
+func (r *Repository) GetTunnelQualityHistory(tunnelID int64, startMs, endMs int64) ([]model.TunnelQuality, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var results []model.TunnelQuality
+	err := r.db.Where("tunnel_id = ? AND timestamp >= ? AND timestamp <= ?", tunnelID, startMs, endMs).
+		Order("timestamp ASC").
+		Find(&results).Error
+	return results, err
+}
+
+// GetLatestTunnelQualities returns the newest quality result per tunnel_id.
+// (Mirrors GetLatestServiceMonitorResults pattern.)
+func (r *Repository) GetLatestTunnelQualities() ([]model.TunnelQuality, error) {
+	if r == nil || r.db == nil {
+		return nil, nil
+	}
+
+	var results []model.TunnelQuality
+
+	// Use window function (works on modern SQLite 3.25+ and PostgreSQL).
+	q := `
+		SELECT id, tunnel_id, entry_to_exit_latency, exit_to_bing_latency,
+		       entry_to_exit_loss, exit_to_bing_loss, success, error_message, timestamp
+		FROM (
+			SELECT *, ROW_NUMBER() OVER (PARTITION BY tunnel_id ORDER BY timestamp DESC, id DESC) AS rn
+			FROM tunnel_quality
+		) t
+		WHERE rn = 1
+		ORDER BY tunnel_id ASC
+	`
+	if err := r.db.Raw(q).Scan(&results).Error; err == nil {
+		return results, nil
+	}
+
+	// Fallback for older SQLite
+	results = nil
+	err := r.db.Order("timestamp DESC, id DESC").Limit(5000).Find(&results).Error
+	if err != nil {
+		return nil, err
+	}
+
+	seen := make(map[int64]struct{}, len(results))
+	out := make([]model.TunnelQuality, 0, len(results))
+	for _, row := range results {
+		if row.TunnelID <= 0 {
+			continue
+		}
+		if _, ok := seen[row.TunnelID]; ok {
+			continue
+		}
+		seen[row.TunnelID] = struct{}{}
+		out = append(out, row)
+	}
+	return out, nil
+}
+
+// PruneTunnelQualityResults deletes quality results older than the given timestamp.
+// (Mirrors PruneServiceMonitorResults pattern.)
+func (r *Repository) PruneTunnelQualityResults(olderThanMs int64) error {
+	if r == nil || r.db == nil {
+		return nil
+	}
+	return r.db.Where("timestamp < ?", olderThanMs).Delete(&model.TunnelQuality{}).Error
+}
+
+// ListEnabledTunnelIDs returns IDs of all tunnels with status=1.
+func (r *Repository) ListEnabledTunnelIDs() ([]int64, error) {
+	if r == nil || r.db == nil {
+		return nil, errors.New("repository not initialized")
+	}
+	var ids []int64
+	err := r.db.Model(&model.Tunnel{}).Where("status = ?", 1).Pluck("id", &ids).Error
+	return ids, err
+}

go-backend/internal/ws/server.go

@@ -39,6 +39,7 @@ type nodeSession struct {
 	nodeID int64
 	secret string
 	conn   *connWrap
+	crypto *security.AESCrypto // 缓存的 AES 加密器，避免每条消息重建
 }
 
 type commandResponse struct {
@@ -72,6 +73,7 @@ type Server struct {
 	jwtSecret    string
 	upgrader     websocket.Upgrader
 	onNodeOnline func(nodeID int64)
+	onNodeMetric func(nodeID int64, info SystemInfo)
 
 	mu      sync.RWMutex
 	admins  map[*connWrap]struct{}
@@ -80,6 +82,22 @@ type Server struct {
 	pending map[string]pendingRequest
 }
 
+type SystemInfo struct {
+	Uptime           uint64  `json:"uptime"`
+	BytesReceived    uint64  `json:"bytes_received"`
+	BytesTransmitted uint64  `json:"bytes_transmitted"`
+	CPUUsage         float64 `json:"cpu_usage"`
+	MemoryUsage      float64 `json:"memory_usage"`
+	DiskUsage        float64 `json:"disk_usage"`
+	Load1            float64 `json:"load1"`
+	Load5            float64 `json:"load5"`
+	Load15           float64 `json:"load15"`
+	TCPConns         int64   `json:"tcp_conns"`
+	UDPConns         int64   `json:"udp_conns"`
+	NetInSpeed       int64   `json:"net_in_speed"`
+	NetOutSpeed      int64   `json:"net_out_speed"`
+}
+
 func (s *Server) SetNodeOnlineHook(fn func(nodeID int64)) {
 	if s == nil {
 		return
@@ -89,6 +107,15 @@ func (s *Server) SetNodeOnlineHook(fn func(nodeID int64)) {
 	s.mu.Unlock()
 }
 
+func (s *Server) SetNodeMetricHook(fn func(nodeID int64, info SystemInfo)) {
+	if s == nil {
+		return
+	}
+	s.mu.Lock()
+	s.onNodeMetric = fn
+	s.mu.Unlock()
+}
+
 func NewServer(repo *repo.Repository, jwtSecret string) *Server {
 	return &Server{
 		repo:      repo,
@@ -185,7 +212,12 @@ func (s *Server) handleNode(w http.ResponseWriter, r *http.Request, nodeID int64
 		_ = old.conn.conn.Close()
 		delete(s.byConn, old.conn.conn)
 	}
-	ns := &nodeSession{nodeID: nodeID, secret: secret, conn: cw}
+	// 初始化 AES 加密器并缓存（仅创建一次）
+	var nodeCrypto *security.AESCrypto
+	if strings.TrimSpace(secret) != "" {
+		nodeCrypto, _ = security.NewAESCrypto(secret)
+	}
+	ns := &nodeSession{nodeID: nodeID, secret: secret, conn: cw, crypto: nodeCrypto}
 	s.nodes[nodeID] = ns
 	s.byConn[conn] = ns
 	s.mu.Unlock()
@@ -225,18 +257,100 @@ func (s *Server) handleNode(w http.ResponseWriter, r *http.Request, nodeID int64
 			return
 		}
 
-		msg := decryptIfNeeded(payload, secret)
+		msg := decryptIfNeeded(payload, ns.crypto, secret)
 		s.tryResolvePending(nodeID, msg)
 
 		var parsed struct {
 			Type string `json:"type"`
 		}
-		if json.Unmarshal([]byte(msg), &parsed) == nil && parsed.Type == "UpgradeProgress" {
-			s.broadcastTyped(nodeID, "upgrade_progress", msg)
-		} else {
-			s.broadcastInfo(nodeID, msg)
+		if json.Unmarshal([]byte(msg), &parsed) == nil && parsed.Type != "" {
+			switch parsed.Type {
+			case "metric":
+				// Agent 新版指标消息：{type:"metric", data:{...}}
+				var envelope struct {
+					Data json.RawMessage `json:"data"`
+				}
+				if err := json.Unmarshal([]byte(msg), &envelope); err == nil && len(envelope.Data) > 0 {
+					// 解析 SystemInfo 并调用 hook
+					var sysInfo SystemInfo
+					if json.Unmarshal(envelope.Data, &sysInfo) == nil {
+						s.mu.RLock()
+						onMetric := s.onNodeMetric
+						s.mu.RUnlock()
+						if onMetric != nil {
+							go onMetric(nodeID, sysInfo)
+						}
+					}
+					// 广播内层 data 给前端（保持平坦结构兼容性）
+					s.broadcastTyped(nodeID, "metric", string(envelope.Data))
+				}
+				continue
+			case "UpgradeProgress":
+				s.broadcastTyped(nodeID, "upgrade_progress", msg)
+				continue
+			default:
+				// Unknown typed messages still get broadcast so future
+				// agent message types are not silently lost.
+				s.broadcastInfo(nodeID, msg)
+				continue
+			}
+		}
+
+		// 兼容旧版 Agent：无 type 字段的系统信息消息
+		if looksLikeSystemInfoMessage(msg) {
+			var sysInfo SystemInfo
+			if err := json.Unmarshal([]byte(msg), &sysInfo); err == nil {
+				s.mu.RLock()
+				onMetric := s.onNodeMetric
+				s.mu.RUnlock()
+				if onMetric != nil {
+					go onMetric(nodeID, sysInfo)
+				}
+				s.broadcastTyped(nodeID, "metric", msg)
+				continue
+			}
+		}
+
+		s.broadcastInfo(nodeID, msg)
+	}
+}
+
+func looksLikeSystemInfoMessage(msg string) bool {
+	// Keep this as a cheap heuristic so that arbitrary JSON objects don't get
+	// misclassified as metrics (SystemInfo unmarshal would otherwise succeed with
+	// all-zero values).
+	if strings.TrimSpace(msg) == "" {
+		return false
+	}
+	if !strings.Contains(msg, "{") {
+		return false
+	}
+
+	keys := []string{
+		"\"uptime\"",
+		"\"cpu_usage\"",
+		"\"memory_usage\"",
+		"\"disk_usage\"",
+		"\"bytes_received\"",
+		"\"bytes_transmitted\"",
+		"\"net_in_speed\"",
+		"\"net_out_speed\"",
+		"\"tcp_conns\"",
+		"\"udp_conns\"",
+		"\"load1\"",
+		"\"load5\"",
+		"\"load15\"",
+	}
+	matched := 0
+	for _, k := range keys {
+		if strings.Contains(msg, k) {
+			matched++
+			if matched >= 3 {
+				return true
+			}
 		}
 	}
+	return false
 }
 
 func (s *Server) SendCommand(nodeID int64, cmdType string, data interface{}, timeout time.Duration) (CommandResult, error) {
@@ -285,13 +399,8 @@ func (s *Server) SendCommand(nodeID int64, cmdType string, data interface{}, tim
 	}
 
 	messageData := rawCmd
-	if strings.TrimSpace(ns.secret) != "" {
-		crypto, err := security.NewAESCrypto(ns.secret)
-		if err != nil {
-			cleanup()
-			return CommandResult{}, err
-		}
-		encrypted, err := crypto.Encrypt(rawCmd)
+	if ns.crypto != nil {
+		encrypted, err := ns.crypto.Encrypt(rawCmd)
 		if err != nil {
 			cleanup()
 			return CommandResult{}, err
@@ -341,6 +450,11 @@ func (s *Server) tryResolvePending(nodeID int64, message string) {
 		return
 	}
 
+	// 快速短路：指标消息永远不含 requestId，跳过完整 JSON 解析
+	if !strings.Contains(message, "\"requestId\"") {
+		return
+	}
+
 	var resp commandResponse
 	if err := json.Unmarshal([]byte(message), &resp); err != nil {
 		return
@@ -458,18 +572,22 @@ func (s *Server) broadcastToAdmins(message string) {
 	}
 }
 
-func decryptIfNeeded(payload []byte, secret string) string {
+func decryptIfNeeded(payload []byte, crypto *security.AESCrypto, secret string) string {
 	text := string(payload)
 	var wrap encryptedMessage
 	if err := json.Unmarshal(payload, &wrap); err != nil || !wrap.Encrypted || strings.TrimSpace(wrap.Data) == "" {
 		return text
 	}
 
-	crypto, err := security.NewAESCrypto(secret)
-	if err != nil {
+	// 优先使用缓存的 crypto 实例
+	c := crypto
+	if c == nil && strings.TrimSpace(secret) != "" {
+		c, _ = security.NewAESCrypto(secret)
+	}
+	if c == nil {
 		return text
 	}
-	plain, err := crypto.Decrypt(wrap.Data)
+	plain, err := c.Decrypt(wrap.Data)
 	if err != nil {
 		return text
 	}

go-backend/tests/contract/forward_count_limit_contract_test.go

@@ -0,0 +1,346 @@
+package contract_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"go-backend/internal/auth"
+	"go-backend/internal/http/response"
+)
+
+func TestForwardCreateBlockedWhenUserNumLimitExceeded(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	now := time.Now().UnixMilli()
+
+	userID := int64(100)
+	tunnelID := int64(1)
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user(id, user, pwd, role_id, exp_time, flow, in_flow, out_flow, flow_reset_time, num, created_time, updated_time, status)
+		VALUES(?, 'num_limit_user', 'pwd', 1, 2727251700000, 99999, 0, 0, 1, 2, ?, ?, 1)
+	`, userID, now, now).Error; err != nil {
+		t.Fatalf("insert user: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(id, name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, 'num_limit_tunnel', 1.0, 1, 'tls', 99999, ?, ?, 1, NULL, 0)
+	`, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
+		VALUES(10, ?, ?, NULL, 99999, 99999, 0, 0, 1, 2727251700000, 1)
+	`, userID, tunnelID).Error; err != nil {
+		t.Fatalf("insert user_tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(1, ?, 'num_limit_user', 'existing_forward_1', ?, '8.8.8.8:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 1: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(2, ?, 'num_limit_user', 'existing_forward_2', ?, '8.8.4.4:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 2: %v", err)
+	}
+
+	token, err := auth.GenerateToken(userID, "num_limit_user", 1, secret)
+	if err != nil {
+		t.Fatalf("generate token: %v", err)
+	}
+
+	payload := `{"tunnelId":1,"name":"new_forward","remoteAddr":"1.1.1.1:53"}`
+	req := httptest.NewRequest(http.MethodPost, "/api/v1/forward/create", bytes.NewBufferString(payload))
+	req.Header.Set("Authorization", token)
+	req.Header.Set("Content-Type", "application/json")
+	res := httptest.NewRecorder()
+
+	router.ServeHTTP(res, req)
+
+	var out response.R
+	if err := json.NewDecoder(res.Body).Decode(&out); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if out.Code == 0 {
+		t.Fatalf("expected non-zero code when num limit exceeded, got code=%d msg=%q", out.Code, out.Msg)
+	}
+	if !strings.Contains(out.Msg, "转发数量已达上限") {
+		t.Fatalf("expected forward count limit message, got %q", out.Msg)
+	}
+}
+
+func TestForwardResumeBlockedWhenUserNumLimitExceeded(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	now := time.Now().UnixMilli()
+
+	userID := int64(101)
+	tunnelID := int64(1)
+	pausedForwardID := int64(3)
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user(id, user, pwd, role_id, exp_time, flow, in_flow, out_flow, flow_reset_time, num, created_time, updated_time, status)
+		VALUES(?, 'num_resume_user', 'pwd', 1, 2727251700000, 99999, 0, 0, 1, 2, ?, ?, 1)
+	`, userID, now, now).Error; err != nil {
+		t.Fatalf("insert user: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(id, name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, 'num_resume_tunnel', 1.0, 1, 'tls', 99999, ?, ?, 1, NULL, 0)
+	`, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
+		VALUES(10, ?, ?, NULL, 99999, 99999, 0, 0, 1, 2727251700000, 1)
+	`, userID, tunnelID).Error; err != nil {
+		t.Fatalf("insert user_tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(1, ?, 'num_resume_user', 'existing_forward_1', ?, '8.8.8.8:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 1: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(2, ?, 'num_resume_user', 'existing_forward_2', ?, '8.8.4.4:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 2: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(?, ?, 'num_resume_user', 'paused_forward', ?, '1.1.1.1:53', 'fifo', 0, 0, ?, ?, 0, 0)
+	`, pausedForwardID, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert paused forward: %v", err)
+	}
+
+	token, err := auth.GenerateToken(userID, "num_resume_user", 1, secret)
+	if err != nil {
+		t.Fatalf("generate token: %v", err)
+	}
+
+	req := httptest.NewRequest(http.MethodPost, "/api/v1/forward/resume", bytes.NewBufferString(`{"id":3}`))
+	req.Header.Set("Authorization", token)
+	req.Header.Set("Content-Type", "application/json")
+	res := httptest.NewRecorder()
+
+	router.ServeHTTP(res, req)
+
+	var out response.R
+	if err := json.NewDecoder(res.Body).Decode(&out); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if out.Code == 0 {
+		t.Fatalf("expected non-zero code when num limit exceeded, got code=%d msg=%q", out.Code, out.Msg)
+	}
+	if !strings.Contains(out.Msg, "转发数量已达上限") {
+		t.Fatalf("expected forward count limit message, got %q", out.Msg)
+	}
+
+	status := mustQueryInt(t, repo, `SELECT status FROM forward WHERE id = ?`, pausedForwardID)
+	if status != 0 {
+		t.Fatalf("expected forward status to remain 0, got %d", status)
+	}
+}
+
+func TestForwardCreateBlockedWhenUserTunnelNumLimitExceeded(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	now := time.Now().UnixMilli()
+
+	userID := int64(102)
+	tunnelID := int64(1)
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user(id, user, pwd, role_id, exp_time, flow, in_flow, out_flow, flow_reset_time, num, created_time, updated_time, status)
+		VALUES(?, 'ut_num_limit_user', 'pwd', 1, 2727251700000, 99999, 0, 0, 1, 99999, ?, ?, 1)
+	`, userID, now, now).Error; err != nil {
+		t.Fatalf("insert user: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(id, name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, 'ut_num_limit_tunnel', 1.0, 1, 'tls', 99999, ?, ?, 1, NULL, 0)
+	`, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
+		VALUES(10, ?, ?, NULL, 1, 99999, 0, 0, 1, 2727251700000, 1)
+	`, userID, tunnelID).Error; err != nil {
+		t.Fatalf("insert user_tunnel with num=1: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(1, ?, 'ut_num_limit_user', 'existing_tunnel_forward', ?, '8.8.8.8:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward: %v", err)
+	}
+
+	token, err := auth.GenerateToken(userID, "ut_num_limit_user", 1, secret)
+	if err != nil {
+		t.Fatalf("generate token: %v", err)
+	}
+
+	payload := `{"tunnelId":1,"name":"new_tunnel_forward","remoteAddr":"1.1.1.1:53"}`
+	req := httptest.NewRequest(http.MethodPost, "/api/v1/forward/create", bytes.NewBufferString(payload))
+	req.Header.Set("Authorization", token)
+	req.Header.Set("Content-Type", "application/json")
+	res := httptest.NewRecorder()
+
+	router.ServeHTTP(res, req)
+
+	var out response.R
+	if err := json.NewDecoder(res.Body).Decode(&out); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if out.Code == 0 {
+		t.Fatalf("expected non-zero code when user_tunnel num limit exceeded, got code=%d msg=%q", out.Code, out.Msg)
+	}
+	if !strings.Contains(out.Msg, "隧道转发数量已达上限") {
+		t.Fatalf("expected tunnel forward count limit message, got %q", out.Msg)
+	}
+}
+
+func TestForwardCreateAllowedWhenBelowUserNumLimit(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	now := time.Now().UnixMilli()
+
+	userID := int64(103)
+	tunnelID := int64(1)
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user(id, user, pwd, role_id, exp_time, flow, in_flow, out_flow, flow_reset_time, num, created_time, updated_time, status)
+		VALUES(?, 'num_ok_user', 'pwd', 1, 2727251700000, 99999, 0, 0, 1, 3, ?, ?, 1)
+	`, userID, now, now).Error; err != nil {
+		t.Fatalf("insert user: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(id, name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, 'num_ok_tunnel', 1.0, 1, 'tls', 99999, ?, ?, 1, NULL, 0)
+	`, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
+		VALUES(10, ?, ?, NULL, 99999, 99999, 0, 0, 1, 2727251700000, 1)
+	`, userID, tunnelID).Error; err != nil {
+		t.Fatalf("insert user_tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(1, ?, 'num_ok_user', 'existing_forward_1', ?, '8.8.8.8:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 1: %v", err)
+	}
+
+	token, err := auth.GenerateToken(userID, "num_ok_user", 1, secret)
+	if err != nil {
+		t.Fatalf("generate token: %v", err)
+	}
+
+	payload := `{"tunnelId":1,"name":"new_forward_ok","remoteAddr":"1.1.1.1:53"}`
+	req := httptest.NewRequest(http.MethodPost, "/api/v1/forward/create", bytes.NewBufferString(payload))
+	req.Header.Set("Authorization", token)
+	req.Header.Set("Content-Type", "application/json")
+	res := httptest.NewRecorder()
+
+	router.ServeHTTP(res, req)
+
+	var out response.R
+	if err := json.NewDecoder(res.Body).Decode(&out); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if out.Code != 0 {
+		t.Fatalf("expected success (code=0) when below num limit, got code=%d msg=%q", out.Code, out.Msg)
+	}
+}
+
+func TestForwardCreateAllowedWhenNumZero(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	now := time.Now().UnixMilli()
+
+	userID := int64(104)
+	tunnelID := int64(1)
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user(id, user, pwd, role_id, exp_time, flow, in_flow, out_flow, flow_reset_time, num, created_time, updated_time, status)
+		VALUES(?, 'num_zero_user', 'pwd', 1, 2727251700000, 99999, 0, 0, 1, 0, ?, ?, 1)
+	`, userID, now, now).Error; err != nil {
+		t.Fatalf("insert user: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(id, name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, 'num_zero_tunnel', 1.0, 1, 'tls', 99999, ?, ?, 1, NULL, 0)
+	`, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
+		VALUES(10, ?, ?, NULL, 0, 99999, 0, 0, 1, 2727251700000, 1)
+	`, userID, tunnelID).Error; err != nil {
+		t.Fatalf("insert user_tunnel with num=0: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(1, ?, 'num_zero_user', 'existing_forward_1', ?, '8.8.8.8:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 1: %v", err)
+	}
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(id, user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(2, ?, 'num_zero_user', 'existing_forward_2', ?, '8.8.4.4:53', 'fifo', 0, 0, ?, ?, 1, 0)
+	`, userID, tunnelID, now, now).Error; err != nil {
+		t.Fatalf("insert existing forward 2: %v", err)
+	}
+
+	token, err := auth.GenerateToken(userID, "num_zero_user", 1, secret)
+	if err != nil {
+		t.Fatalf("generate token: %v", err)
+	}
+
+	payload := `{"tunnelId":1,"name":"new_forward_zero","remoteAddr":"1.1.1.1:53"}`
+	req := httptest.NewRequest(http.MethodPost, "/api/v1/forward/create", bytes.NewBufferString(payload))
+	req.Header.Set("Authorization", token)
+	req.Header.Set("Content-Type", "application/json")
+	res := httptest.NewRecorder()
+
+	router.ServeHTTP(res, req)
+
+	var out response.R
+	if err := json.NewDecoder(res.Body).Decode(&out); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if out.Code != 0 {
+		t.Fatalf("expected success (code=0) when num=0 (unlimited), got code=%d msg=%q", out.Code, out.Msg)
+	}
+}

go-backend/tests/contract/issue313_entry_port_conflict_contract_test.go

@@ -112,6 +112,12 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
 		t.Fatalf("insert forward_port a: %v", err)
 	}
 
+	// Simulate legacy dirty data: tunnel A already occupies port 2000 on entryB2.
+	// When tunnel B adds entryB2, the inherited forward port should conflict cross-tunnel.
+	if err := repo.DB().Exec(`INSERT INTO forward_port(forward_id, node_id, port) VALUES(?, ?, ?)`, forwardAID, entryB2, 2000).Error; err != nil {
+		t.Fatalf("insert forward_port a on entryB2: %v", err)
+	}
+
 	if err := repo.DB().Exec(`
 		INSERT INTO user_tunnel(id, user_id, tunnel_id, speed_id, num, flow, in_flow, out_flow, flow_reset_time, exp_time, status)
 		VALUES(3132, 1, ?, NULL, 999, 99999, 0, 0, 1, 2727251700000, 1)
@@ -170,7 +176,8 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
 		t.Fatalf("expected update failure due to cross-tunnel port conflict, got success with code 0")
 	}
 
-	if !bytes.Contains([]byte(out.Msg), []byte("端口")) && !bytes.Contains([]byte(out.Msg), []byte("占用")) {
+	msgBytes := []byte(out.Msg)
+	if !bytes.Contains(msgBytes, []byte("端口")) && !bytes.Contains(msgBytes, []byte("占用")) {
 		t.Fatalf("expected port conflict error message, got %q", out.Msg)
 	}
 

go-backend/tests/contract/issue349_ipv6_entry_format_contract_test.go

@@ -0,0 +1,80 @@
+package contract_test
+
+import (
+	"testing"
+	"time"
+)
+
+func TestIssue349_ForwardListFormatsIPv6EntryAddressesContract(t *testing.T) {
+	secret := "contract-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+	adminToken := mustAdminToken(t, secret)
+	now := time.Now().UnixMilli()
+
+	if err := repo.DB().Exec(`
+		INSERT INTO tunnel(name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, in_ip, inx)
+		VALUES(?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+	`, "issue349-tunnel", 1.0, 1, "tcp", 99999, now, now, 1, nil, 0).Error; err != nil {
+		t.Fatalf("insert tunnel: %v", err)
+	}
+	tunnelID := mustLastInsertID(t, repo, "issue349-tunnel")
+
+	if err := repo.DB().Exec(`
+		INSERT INTO node(name, secret, server_ip, server_ip_v4, server_ip_v6, port, interface_name, version, http, tls, socks, created_time, updated_time, status, tcp_listen_addr, udp_listen_addr, inx)
+		VALUES(?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+	`, "issue349-entry-node-a", "entry-secret-a", "2001:db8::10", "", "2001:db8::10", "32000-32010", "", "v1", 1, 1, 1, now, now, 1, "[::]", "[::]", 0).Error; err != nil {
+		t.Fatalf("insert node a: %v", err)
+	}
+	nodeAID := mustLastInsertID(t, repo, "issue349-entry-node-a")
+
+	if err := repo.DB().Exec(`
+		INSERT INTO node(name, secret, server_ip, server_ip_v4, server_ip_v6, port, interface_name, version, http, tls, socks, created_time, updated_time, status, tcp_listen_addr, udp_listen_addr, inx)
+		VALUES(?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+	`, "issue349-entry-node-b", "entry-secret-b", "2001:db8::30", "", "2001:db8::30", "32000-32010", "", "v1", 1, 1, 1, now, now, 1, "[::]", "[::]", 1).Error; err != nil {
+		t.Fatalf("insert node b: %v", err)
+	}
+	nodeBID := mustLastInsertID(t, repo, "issue349-entry-node-b")
+
+	if err := repo.DB().Exec(`
+		INSERT INTO forward(user_id, user_name, name, tunnel_id, remote_addr, strategy, in_flow, out_flow, created_time, updated_time, status, inx)
+		VALUES(?, ?, ?, ?, ?, ?, 0, 0, ?, ?, 1, ?)
+	`, 1, "admin_user", "issue349-forward", tunnelID, "1.1.1.1:443", "fifo", now, now, 0).Error; err != nil {
+		t.Fatalf("insert forward: %v", err)
+	}
+	forwardID := mustLastInsertID(t, repo, "issue349-forward")
+
+	if err := repo.DB().Exec(`INSERT INTO forward_port(forward_id, node_id, port) VALUES(?, ?, ?)`, forwardID, nodeAID, 32001).Error; err != nil {
+		t.Fatalf("insert forward_port a: %v", err)
+	}
+	if err := repo.DB().Exec(`INSERT INTO forward_port(forward_id, node_id, port, in_ip) VALUES(?, ?, ?, ?)`, forwardID, nodeBID, 32002, "2001:db8::20").Error; err != nil {
+		t.Fatalf("insert forward_port b: %v", err)
+	}
+
+	out := requestContractEnvelope(t, router, adminToken, "/api/v1/forward/list", nil)
+	if out.Code != 0 {
+		t.Fatalf("forward list failed: code=%d msg=%q", out.Code, out.Msg)
+	}
+
+	rows := mustContractSlice(t, out.Data, "forward list data")
+	var target map[string]interface{}
+	for _, row := range rows {
+		item, ok := row.(map[string]interface{})
+		if !ok {
+			continue
+		}
+		if contractValueAsInt64(item["id"]) == forwardID {
+			target = item
+			break
+		}
+	}
+	if target == nil {
+		t.Fatalf("target forward %d not found in /forward/list response", forwardID)
+	}
+
+	if got := contractValueAsString(target["inIp"]); got != "[2001:db8::10]:32001,[2001:db8::20]:32002" {
+		t.Fatalf("expected bracketed IPv6 entry list, got %q", got)
+	}
+	if got := contractValueAsInt64(target["inPort"]); got != 32001 {
+		t.Fatalf("expected first entry port 32001, got %d", got)
+	}
+}

go-backend/tests/contract/tunnel_metrics_ingestion_contract_test.go

@@ -0,0 +1,97 @@
+package contract_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+
+	"go-backend/internal/store/model"
+)
+
+func TestFlowUploadInsertsTunnelMetrics(t *testing.T) {
+	secret := "monitoring-jwt-secret"
+	router, repo := setupContractRouter(t, secret)
+
+	now := time.Now().UnixMilli()
+
+	node := &model.Node{
+		Name:          "node-1",
+		Secret:        "node-secret",
+		ServerIP:      "127.0.0.1",
+		Port:          "10000-10010",
+		TCPListenAddr: "[::]",
+		UDPListenAddr: "[::]",
+		CreatedTime:   now,
+		Status:        1,
+	}
+	if err := repo.DB().Create(node).Error; err != nil {
+		t.Fatalf("seed node: %v", err)
+	}
+
+	tunnel := &model.Tunnel{
+		Name:         "tunnel-1",
+		TrafficRatio: 1.0,
+		Type:         1,
+		Protocol:     "tls",
+		Flow:         1,
+		CreatedTime:  now,
+		UpdatedTime:  now,
+		Status:       1,
+	}
+	if err := repo.DB().Create(tunnel).Error; err != nil {
+		t.Fatalf("seed tunnel: %v", err)
+	}
+
+	forward := &model.Forward{
+		UserID:      123,
+		UserName:    "user-123",
+		Name:        "forward-1",
+		TunnelID:    tunnel.ID,
+		RemoteAddr:  "1.1.1.1:80",
+		CreatedTime: now,
+		UpdatedTime: now,
+		Status:      1,
+	}
+	if err := repo.DB().Create(forward).Error; err != nil {
+		t.Fatalf("seed forward: %v", err)
+	}
+
+	serviceName := jsonNumber(forward.ID) + "_123_0"
+	body, _ := json.Marshal([]map[string]interface{}{{
+		"n": serviceName,
+		"u": 200,
+		"d": 100,
+	}})
+
+	req := httptest.NewRequest(http.MethodPost, "/flow/upload?secret="+node.Secret, bytes.NewReader(body))
+	res := httptest.NewRecorder()
+	router.ServeHTTP(res, req)
+
+	if res.Code != http.StatusOK {
+		t.Fatalf("expected status 200, got %d", res.Code)
+	}
+
+	metrics, err := repo.GetTunnelMetrics(tunnel.ID, 0, now+60_000)
+	if err != nil {
+		t.Fatalf("get tunnel metrics: %v", err)
+	}
+	if len(metrics) != 1 {
+		t.Fatalf("expected 1 tunnel metric row, got %d", len(metrics))
+	}
+
+	if metrics[0].TunnelID != tunnel.ID {
+		t.Fatalf("expected tunnelId %d, got %d", tunnel.ID, metrics[0].TunnelID)
+	}
+	if metrics[0].NodeID != node.ID {
+		t.Fatalf("expected nodeId %d, got %d", node.ID, metrics[0].NodeID)
+	}
+	if metrics[0].BytesIn != 100 {
+		t.Fatalf("expected bytesIn 100, got %d", metrics[0].BytesIn)
+	}
+	if metrics[0].BytesOut != 200 {
+		t.Fatalf("expected bytesOut 200, got %d", metrics[0].BytesOut)
+	}
+}

go-gost/AGENTS.md

@@ -1,6 +1,9 @@
 # GO-GOST SERVICE KNOWLEDGE BASE
 
-**Generated:** Thu Feb 26 2026
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
 
 ## OVERVIEW
 Forwarding agent built on GOST v3 with a local fork of `github.com/go-gost/x` under `x/`.

go-gost/main.go

@@ -119,7 +119,9 @@ func main() {
 	log := xlogger.NewLogger()
 	logger.SetDefault(log)
 
-	wsReporter := socket.StartWebSocketReporterWithConfig(config.Addr, config.Secret, config.Http, config.Tls, config.Socks, version)
+	distro := socket.DetectDistro()
+	fullVersion := fmt.Sprintf("%s (%s/%s)", version, distro, runtime.GOARCH)
+	wsReporter := socket.StartWebSocketReporterWithConfig(config.Addr, config.Secret, config.Http, config.Tls, config.Socks, fullVersion)
 	defer wsReporter.Stop()
 	service.SetHTTPReportURL(config.Addr, config.Secret)
 

go-gost/x/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Local fork of `github.com/go-gost/x` used by `go-gost/` via `replace github.com/go-gost/x => ./x`. Most protocol/runtime behavior changes happen here. 30+ top-level packages - framework-style layout.
 

go-gost/x/api/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X API KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Gin-based management API for reading/writing config and controlling services at runtime.
 

go-gost/x/api/config_service.go

@@ -624,6 +624,11 @@ func resumeService(ctx *gin.Context) {
 	existingSvc.Close()
 	registry.ServiceRegistry().Unregister(name)
 
+	// 强制断开端口的所有连接
+	if serviceConfig.Addr != "" {
+		_ = kill.ForceClosePortConnections(serviceConfig.Addr)
+	}
+
 	// 等待端口释放
 	time.Sleep(500 * time.Millisecond)
 
@@ -1039,8 +1044,13 @@ func resumeServices(ctx *gin.Context) {
 		str.service.Close()
 		registry.ServiceRegistry().Unregister(str.name)
 
+		// 强制断开端口的所有连接
+		if str.serviceConfig.Addr != "" {
+			_ = kill.ForceClosePortConnections(str.serviceConfig.Addr)
+		}
+
 		// 等待端口释放
-		time.Sleep(100 * time.Millisecond)
+		time.Sleep(500 * time.Millisecond)
 
 		// 重新解析并启动服务
 		svc, err := parser.ParseService(str.serviceConfig)

go-gost/x/config/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X CONFIG KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Config model + parsing/loading pipeline for the `go-gost/x` runtime. This is the bridge between `gost.json`/`gost.yaml` and in-memory registries/services.
 

go-gost/x/connector/AGENTS.md

@@ -1,6 +1,9 @@
 # GOST CONNECTOR KNOWLEDGE BASE
 
-**Generated:** Fri Feb 13 2026
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
 
 ## OVERVIEW
 Connection initiators (clients) for various protocols in GOST forwarding.

go-gost/x/dialer/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X DIALERS KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Outbound dialers (client-side connection establishment) used by connectors/handlers.
 

go-gost/x/handler/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X HANDLERS KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Protocol handlers (server-side request handling) used by services defined in the GOST config.
 

go-gost/x/listener/AGENTS.md

@@ -1,5 +1,10 @@
 # GO-GOST/X LISTENERS KNOWLEDGE BASE
 
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
+
 ## OVERVIEW
 Inbound listeners (transport-level accept loops) used by services defined in the GOST config.
 

go-gost/x/registry/AGENTS.md

@@ -1,6 +1,9 @@
 # GO-GOST REGISTRY KNOWLEDGE BASE
 
-**Generated:** Wed Feb 04 2026
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
 
 ## OVERVIEW
 Central registration point for all pluggable GOST components (handlers, listeners, dialers, etc.).

go-gost/x/service/service.go

@@ -7,7 +7,6 @@ import (
 	"errors"
 	"fmt"
 	"io"
-	"log"
 	"net"
 	"os"
 	"os/exec"
@@ -63,11 +62,9 @@ func SetProtocolBlock(httpOn int, tlsOn int, socksOn int) {
 type Option func(opts *options)
 
 func init() {
-	_, err := LoadConfig("config.json")
-	fmt.Println("config.json loaded")
-	if err != nil {
-		log.Fatal(err)
-	}
+	// NOTE: This package can be imported by tests/tools that don't have a local
+	// config.json. Missing config should not crash the process.
+	_, _ = LoadConfig("config.json")
 	needWrap = isTls+isSocks+isHttp > 0
 }
 

go-gost/x/socket/AGENTS.md

@@ -1,6 +1,9 @@
 # GOST SOCKET KNOWLEDGE BASE
 
-**Generated:** Sun Feb 15 2026
+**Generated:** Fri Mar 20 2026
+**Commit:** f45f960
+**Branch:** main
+**Tag:** 2.1.9-beta6
 
 ## OVERVIEW
 WebSocket reporter and socket utilities for panel integration.

go-gost/x/socket/distro.go

@@ -0,0 +1,21 @@
+package socket
+
+import (
+	"strings"
+
+	"github.com/shirou/gopsutil/v3/host"
+)
+
+// DetectDistro returns the Linux distribution name (e.g. "ubuntu", "centos",
+// "debian"). Falls back to "linux" when detection fails.
+func DetectDistro() string {
+	info, err := host.Info()
+	if err != nil || info == nil {
+		return "linux"
+	}
+	platform := strings.ToLower(strings.TrimSpace(info.Platform))
+	if platform == "" {
+		return "linux"
+	}
+	return platform
+}

go-gost/x/socket/service.go

@@ -397,8 +397,13 @@ func resumeServices(req resumeServicesRequest) error {
 		str.service.Close()
 		registry.ServiceRegistry().Unregister(str.name)
 
+		// 强制断开端口的所有连接
+		if str.serviceConfig.Addr != "" {
+			_ = kill.ForceClosePortConnections(str.serviceConfig.Addr)
+		}
+
 		// 等待端口释放
-		time.Sleep(100 * time.Millisecond)
+		time.Sleep(500 * time.Millisecond)
 
 		// 重新解析并启动服务
 		svc, err := parser.ParseService(str.serviceConfig)

go-gost/x/socket/websocket_reporter.go

@@ -9,6 +9,7 @@ import (
 	"encoding/json"
 	"fmt"
 	"io"
+	"math/rand"
 	"net"
 	"net/http"
 	"net/url"
@@ -17,7 +18,7 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
-	"sync" // 新增：用于管理连接状态的互斥锁
+	"sync"
 	"time"
 
 	"github.com/go-gost/x/config"
@@ -25,34 +26,67 @@ import (
 	"github.com/go-gost/x/service"
 	"github.com/gorilla/websocket"
 	"github.com/shirou/gopsutil/v3/cpu"
+	"github.com/shirou/gopsutil/v3/disk"
 	"github.com/shirou/gopsutil/v3/host"
+	"github.com/shirou/gopsutil/v3/load"
 	"github.com/shirou/gopsutil/v3/mem"
 	psnet "github.com/shirou/gopsutil/v3/net"
+	"golang.org/x/net/icmp"
+	"golang.org/x/net/ipv4"
+	"golang.org/x/net/ipv6"
 )
 
 // SystemInfo 系统信息结构体
 type SystemInfo struct {
-	Uptime           uint64  `json:"uptime"`            // 开机时间	（秒）
-	BytesReceived    uint64  `json:"bytes_received"`    // 接收字节数
-	BytesTransmitted uint64  `json:"bytes_transmitted"` // 发送字节数
-	CPUUsage         float64 `json:"cpu_usage"`         // CPU使用率（百分比）
-	MemoryUsage      float64 `json:"memory_usage"`      // 内存使用率（百分比）
+	Uptime           uint64  `json:"uptime"`
+	BytesReceived    uint64  `json:"bytes_received"`
+	BytesTransmitted uint64  `json:"bytes_transmitted"`
+	CPUUsage         float64 `json:"cpu_usage"`
+	MemoryUsage      float64 `json:"memory_usage"`
+	DiskUsage        float64 `json:"disk_usage"`
+	Load1            float64 `json:"load1"`
+	Load5            float64 `json:"load5"`
+	Load15           float64 `json:"load15"`
+	TCPConns         int64   `json:"tcp_conns"`
+	UDPConns         int64   `json:"udp_conns"`
+	NetInSpeed       int64   `json:"net_in_speed"`
+	NetOutSpeed      int64   `json:"net_out_speed"`
 }
 
 // NetworkStats 网络统计信息
 type NetworkStats struct {
-	BytesReceived    uint64 `json:"bytes_received"`    // 接收字节数
-	BytesTransmitted uint64 `json:"bytes_transmitted"` // 发送字节数
+	BytesReceived    uint64 `json:"bytes_received"`
+	BytesTransmitted uint64 `json:"bytes_transmitted"`
+	BytesRecvDelta   uint64 `json:"bytes_recv_delta"`
+	BytesSentDelta   uint64 `json:"bytes_sent_delta"`
 }
 
 // CPUInfo CPU信息
 type CPUInfo struct {
-	Usage float64 `json:"usage"` // CPU使用率（百分比）
+	Usage float64 `json:"usage"`
 }
 
 // MemoryInfo 内存信息
 type MemoryInfo struct {
-	Usage float64 `json:"usage"` // 内存使用率（百分比）
+	Usage float64 `json:"usage"`
+}
+
+// DiskInfo 磁盘信息
+type DiskInfo struct {
+	Usage float64 `json:"usage"`
+}
+
+// LoadInfo 负载信息
+type LoadInfo struct {
+	Load1  float64 `json:"load1"`
+	Load5  float64 `json:"load5"`
+	Load15 float64 `json:"load15"`
+}
+
+// ConnectionInfo 连接信息
+type ConnectionInfo struct {
+	TCPConns int64 `json:"tcp_conns"`
+	UDPConns int64 `json:"udp_conns"`
 }
 
 // CommandMessage 命令消息结构体
@@ -91,9 +125,31 @@ type TcpPingResponse struct {
 	RequestId    string  `json:"requestId,omitempty"`
 }
 
+// ServiceMonitorCheckRequest service monitor check request.
+type ServiceMonitorCheckRequest struct {
+	MonitorID  int64  `json:"monitorId"`
+	Type       string `json:"type"` // tcp|icmp
+	Target     string `json:"target"`
+	TimeoutSec int    `json:"timeoutSec"`
+}
+
+// ServiceMonitorCheckResult node-executed check output.
+// CommandResponse.Success indicates command execution status.
+// Actual check success is represented by this struct.
+type ServiceMonitorCheckResult struct {
+	MonitorID    int64   `json:"monitorId"`
+	Success      bool    `json:"success"`
+	LatencyMs    float64 `json:"latencyMs"`
+	StatusCode   int     `json:"statusCode,omitempty"`
+	ErrorMessage string  `json:"errorMessage,omitempty"`
+}
+
 const (
 	reporterReadWait  = 60 * time.Second
 	reporterWriteWait = 5 * time.Second
+	wsPingInterval    = 20 * time.Second // 独立 WebSocket ping 间隔
+	initialBackoff    = 2 * time.Second  // 重连初始退避
+	maxBackoff        = 2 * time.Minute  // 重连最大退避
 )
 
 type WebSocketReporter struct {
@@ -103,15 +159,15 @@ type WebSocketReporter struct {
 	version           string // 保存版本号
 	preferredWSScheme string
 	conn              *websocket.Conn
-	reconnectTime     time.Duration
+	curBackoff        time.Duration // 当前重连退避间隔
 	pingInterval      time.Duration
 	configInterval    time.Duration
 	ctx               context.Context
 	cancel            context.CancelFunc
 	connected         bool
-	connecting        bool              // 新增：正在连接状态
-	connMutex         sync.Mutex        // 新增：连接状态锁
-	aesCrypto         *crypto.AESCrypto // 新增：AES加密器
+	connecting        bool              // 正在连接状态
+	connMutex         sync.Mutex        // 连接状态锁
+	aesCrypto         *crypto.AESCrypto // AES加密器
 }
 
 var wsDial = func(dialer *websocket.Dialer, rawURL string) (*websocket.Conn, *http.Response, error) {
@@ -133,8 +189,8 @@ func NewWebSocketReporter(serverURL string, secret string) *WebSocketReporter {
 
 	return &WebSocketReporter{
 		url:            serverURL,
-		reconnectTime:  5 * time.Second,  // 重连间隔
-		pingInterval:   2 * time.Second,  // 发送间隔改为2秒
+		curBackoff:     initialBackoff,   // 当前退避间隔
+		pingInterval:   1 * time.Second,  // 指标上报间隔（每秒采集）
 		configInterval: 10 * time.Minute, // 配置上报间隔
 		ctx:            ctx,
 		cancel:         cancel,
@@ -152,10 +208,17 @@ func (w *WebSocketReporter) Start() {
 // Stop 停止WebSocket报告器
 func (w *WebSocketReporter) Stop() {
 	w.cancel()
+	w.connMutex.Lock()
 	if w.conn != nil {
 		w.conn.Close()
 	}
+	w.connMutex.Unlock()
+}
 
+// backoffWithJitter 返回带随机抖动的退避时间（±25%）
+func backoffWithJitter(base time.Duration) time.Duration {
+	jitter := time.Duration(float64(base) * (0.75 + rand.Float64()*0.5))
+	return jitter
 }
 
 // run 主运行循环
@@ -172,23 +235,32 @@ func (w *WebSocketReporter) run() {
 
 			if needConnect {
 				if err := w.connect(); err != nil {
-					fmt.Printf("❌ WebSocket连接失败: %v，%v后重试\n", err, w.reconnectTime)
+					wait := backoffWithJitter(w.curBackoff)
+					fmt.Printf("❌ WebSocket连接失败: %v，%v后重试\n", err, wait)
+					// 指数退避：翻倍当前退避间隔，上限 maxBackoff
+					w.curBackoff *= 2
+					if w.curBackoff > maxBackoff {
+						w.curBackoff = maxBackoff
+					}
 					select {
-					case <-time.After(w.reconnectTime):
+					case <-time.After(wait):
 						continue
 					case <-w.ctx.Done():
 						return
 					}
 				}
+				// 连接成功：重置退避
+				w.curBackoff = initialBackoff
 			}
 
 			// 连接成功，开始发送消息
 			if w.connected {
 				w.handleConnection()
 			} else {
+				wait := backoffWithJitter(w.curBackoff)
 				// 如果连接失败，等待重试
 				select {
-				case <-time.After(w.reconnectTime):
+				case <-time.After(wait):
 					continue
 				case <-w.ctx.Done():
 					return
@@ -276,7 +348,7 @@ func buildWebSocketCandidates(addr string, secret string, version string, http i
 		normalizedAddr = strings.TrimSpace(addr)
 	}
 
-	query := "/system-info?type=1&secret=" + secret + "&version=" + version +
+	query := "/system-info?type=1&secret=" + url.QueryEscape(secret) + "&version=" + url.QueryEscape(version) +
 		"&http=" + strconv.Itoa(http) + "&tls=" + strconv.Itoa(tls) + "&socks=" + strconv.Itoa(socks)
 
 	schemes := []string{"wss", "ws"}
@@ -421,15 +493,34 @@ func (w *WebSocketReporter) handleConnection() {
 	// 启动消息接收goroutine
 	go w.receiveMessages()
 
-	// 主发送循环
-	ticker := time.NewTicker(w.pingInterval)
-	defer ticker.Stop()
+	// 指标上报 ticker
+	metricTicker := time.NewTicker(w.pingInterval)
+	defer metricTicker.Stop()
+
+	// 独立 WebSocket keepalive ping ticker
+	pingTicker := time.NewTicker(wsPingInterval)
+	defer pingTicker.Stop()
 
 	for {
 		select {
 		case <-w.ctx.Done():
 			return
-		case <-ticker.C:
+
+		case <-pingTicker.C:
+			// 发送 WebSocket ping 保活，独立于指标上报
+			w.connMutex.Lock()
+			conn := w.conn
+			isConnected := w.connected
+			w.connMutex.Unlock()
+			if !isConnected || conn == nil {
+				return
+			}
+			if err := conn.WriteControl(websocket.PingMessage, nil, time.Now().Add(reporterWriteWait)); err != nil {
+				fmt.Printf("❌ 发送WebSocket ping失败: %v，准备重连\n", err)
+				return
+			}
+
+		case <-metricTicker.C:
 			// 检查连接状态
 			w.connMutex.Lock()
 			isConnected := w.connected
@@ -449,11 +540,35 @@ func (w *WebSocketReporter) handleConnection() {
 	}
 }
 
+var lastNetBytesReceived uint64
+var lastNetBytesTransmitted uint64
+var lastNetTime int64
+
+var connInfoCached ConnectionInfo
+var connInfoCachedAt int64
+var connInfoCachedMu sync.Mutex
+
 // collectSystemInfo 收集系统信息
 func (w *WebSocketReporter) collectSystemInfo() SystemInfo {
 	networkStats := getNetworkStats()
 	cpuInfo := getCPUInfo()
 	memoryInfo := getMemoryInfo()
+	diskInfo := getDiskInfo()
+	loadInfo := getLoadInfo()
+	connInfo := getConnectionInfo()
+
+	now := time.Now().UnixMilli()
+	var netInSpeed, netOutSpeed int64
+	if lastNetTime > 0 {
+		deltaMs := now - lastNetTime
+		if deltaMs > 0 {
+			netInSpeed = int64(float64(networkStats.BytesRecvDelta) * 1000 / float64(deltaMs))
+			netOutSpeed = int64(float64(networkStats.BytesSentDelta) * 1000 / float64(deltaMs))
+		}
+	}
+	lastNetBytesReceived = networkStats.BytesReceived
+	lastNetBytesTransmitted = networkStats.BytesTransmitted
+	lastNetTime = now
 
 	return SystemInfo{
 		Uptime:           getUptime(),
@@ -461,9 +576,48 @@ func (w *WebSocketReporter) collectSystemInfo() SystemInfo {
 		BytesTransmitted: networkStats.BytesTransmitted,
 		CPUUsage:         cpuInfo.Usage,
 		MemoryUsage:      memoryInfo.Usage,
+		DiskUsage:        diskInfo.Usage,
+		Load1:            loadInfo.Load1,
+		Load5:            loadInfo.Load5,
+		Load15:           loadInfo.Load15,
+		TCPConns:         connInfo.TCPConns,
+		UDPConns:         connInfo.UDPConns,
+		NetInSpeed:       netInSpeed,
+		NetOutSpeed:      netOutSpeed,
 	}
 }
 
+// encryptPayload 加密 JSON 数据，返回加密后的消息字节（若加密失败则回退到原始数据）
+func (w *WebSocketReporter) encryptPayload(jsonData []byte) []byte {
+	if w.aesCrypto == nil {
+		return jsonData
+	}
+
+	encryptedData, err := w.aesCrypto.Encrypt(jsonData)
+	if err != nil {
+		fmt.Printf("⚠️ 加密失败，发送原始数据: %v\n", err)
+		return jsonData
+	}
+
+	encryptedMessage := map[string]interface{}{
+		"encrypted": true,
+		"data":      encryptedData,
+		"timestamp": time.Now().Unix(),
+	}
+	messageData, err := json.Marshal(encryptedMessage)
+	if err != nil {
+		fmt.Printf("⚠️ 序列化加密消息失败，发送原始数据: %v\n", err)
+		return jsonData
+	}
+	return messageData
+}
+
+// metricEnvelope wraps SystemInfo with a type field for fast identification on the panel side.
+type metricEnvelope struct {
+	Type string     `json:"type"`
+	Data SystemInfo `json:"data"`
+}
+
 // sendSystemInfo 发送系统信息
 func (w *WebSocketReporter) sendSystemInfo(sysInfo SystemInfo) error {
 	w.connMutex.Lock()
@@ -473,42 +627,19 @@ func (w *WebSocketReporter) sendSystemInfo(sysInfo SystemInfo) error {
 		return fmt.Errorf("连接未建立")
 	}
 
-	// 转换为JSON
-	jsonData, err := json.Marshal(sysInfo)
+	// 使用 type:"metric" 信封包装，Panel 可通过 type 字段直接识别指标消息
+	envelope := metricEnvelope{Type: "metric", Data: sysInfo}
+	jsonData, err := json.Marshal(envelope)
 	if err != nil {
 		return fmt.Errorf("序列化系统信息失败: %v", err)
 	}
 
-	var messageData []byte
+	messageData := w.encryptPayload(jsonData)
 
-	// 如果有加密器，则加密数据
-	if w.aesCrypto != nil {
-		encryptedData, err := w.aesCrypto.Encrypt(jsonData)
-		if err != nil {
-			fmt.Printf("⚠️ 加密失败，发送原始数据: %v\n", err)
-			messageData = jsonData
-		} else {
-			// 创建加密消息包装器
-			encryptedMessage := map[string]interface{}{
-				"encrypted": true,
-				"data":      encryptedData,
-				"timestamp": time.Now().Unix(),
-			}
-			messageData, err = json.Marshal(encryptedMessage)
-			if err != nil {
-				fmt.Printf("⚠️ 序列化加密消息失败，发送原始数据: %v\n", err)
-				messageData = jsonData
-			}
-		}
-	} else {
-		messageData = jsonData
-	}
-
-	// 设置写入超时
 	w.conn.SetWriteDeadline(time.Now().Add(5 * time.Second))
 
 	if err := w.conn.WriteMessage(websocket.TextMessage, messageData); err != nil {
-		w.connected = false // 标记连接已断开
+		w.connected = false
 		return fmt.Errorf("写入消息失败: %v", err)
 	}
 
@@ -517,23 +648,19 @@ func (w *WebSocketReporter) sendSystemInfo(sysInfo SystemInfo) error {
 
 // receiveMessages 接收服务端发送的消息
 func (w *WebSocketReporter) receiveMessages() {
+	// 获取连接引用一次即可，连接生命周期由 handleConnection 管理
+	w.connMutex.Lock()
+	conn := w.conn
+	w.connMutex.Unlock()
+	if conn == nil {
+		return
+	}
+
 	for {
 		select {
 		case <-w.ctx.Done():
 			return
 		default:
-			w.connMutex.Lock()
-			conn := w.conn
-			connected := w.connected
-			w.connMutex.Unlock()
-
-			if conn == nil || !connected {
-				return
-			}
-
-			// 设置读取超时
-			conn.SetReadDeadline(time.Now().Add(reporterReadWait))
-
 			messageType, message, err := conn.ReadMessage()
 			if err != nil {
 				if websocket.IsUnexpectedCloseError(err, websocket.CloseGoingAway, websocket.CloseAbnormalClosure) {
@@ -621,12 +748,8 @@ func (w *WebSocketReporter) handleReceivedMessage(messageType int, message []byt
 			}
 
 			if cmdMsg.Type != "call" {
-				// 其他状态变更命令保持同步，确保顺序执行
-				if cmdMsg.Type == "TcpPing" || cmdMsg.Type == "UpgradeAgent" || cmdMsg.Type == "RollbackAgent" {
-					go w.routeCommand(cmdMsg)
-				} else {
-					w.routeCommand(cmdMsg)
-				}
+				// 所有命令统一异步执行，避免阻塞消息接收循环
+				go w.routeCommand(cmdMsg)
 			}
 		} else {
 			// 处理普通消息
@@ -637,12 +760,8 @@ func (w *WebSocketReporter) handleReceivedMessage(messageType int, message []byt
 				return
 			}
 			if cmdMsg.Type != "call" {
-				// 其他状态变更命令保持同步，确保顺序执行
-				if cmdMsg.Type == "TcpPing" || cmdMsg.Type == "UpgradeAgent" || cmdMsg.Type == "RollbackAgent" {
-					go w.routeCommand(cmdMsg)
-				} else {
-					w.routeCommand(cmdMsg)
-				}
+				// 所有命令统一异步执行，避免阻塞消息接收循环
+				go w.routeCommand(cmdMsg)
 			}
 		}
 
@@ -726,6 +845,13 @@ func (w *WebSocketReporter) routeCommand(cmd CommandMessage) {
 		response.Data = tcpPingResult
 		// needSaveConfig = false (默认值)
 
+	// Service monitor check (read-only)
+	case "ServiceMonitorCheck":
+		var checkResult ServiceMonitorCheckResult
+		checkResult, err = w.handleServiceMonitorCheck(cmd.Data)
+		response.Type = "ServiceMonitorCheckResponse"
+		response.Data = checkResult
+
 	// Protocol blocking switches
 	case "SetProtocol":
 		err = w.handleSetProtocol(cmd.Data)
@@ -1309,30 +1435,7 @@ func (w *WebSocketReporter) sendResponse(response CommandResponse) {
 		return
 	}
 
-	var messageData []byte
-
-	// 如果有加密器，则加密数据
-	if w.aesCrypto != nil {
-		encryptedData, err := w.aesCrypto.Encrypt(jsonData)
-		if err != nil {
-			fmt.Printf("⚠️ 加密响应失败，发送原始数据: %v\n", err)
-			messageData = jsonData
-		} else {
-			// 创建加密消息包装器
-			encryptedMessage := map[string]interface{}{
-				"encrypted": true,
-				"data":      encryptedData,
-				"timestamp": time.Now().Unix(),
-			}
-			messageData, err = json.Marshal(encryptedMessage)
-			if err != nil {
-				fmt.Printf("⚠️ 序列化加密响应失败，发送原始数据: %v\n", err)
-				messageData = jsonData
-			}
-		}
-	} else {
-		messageData = jsonData
-	}
+	messageData := w.encryptPayload(jsonData)
 
 	// 检查消息大小，如果超过10MB则记录警告
 	if len(messageData) > 10*1024*1024 {
@@ -1381,17 +1484,21 @@ func getNetworkStats() NetworkStats {
 		return stats
 	}
 
-	// 汇总所有非回环接口的流量
 	for _, io := range ioCounters {
-		// 跳过回环接口
 		if io.Name == "lo" || strings.HasPrefix(io.Name, "lo") {
 			continue
 		}
-
 		stats.BytesReceived += io.BytesRecv
 		stats.BytesTransmitted += io.BytesSent
 	}
 
+	if lastNetBytesReceived > 0 && stats.BytesReceived >= lastNetBytesReceived {
+		stats.BytesRecvDelta = stats.BytesReceived - lastNetBytesReceived
+	}
+	if lastNetBytesTransmitted > 0 && stats.BytesTransmitted >= lastNetBytesTransmitted {
+		stats.BytesSentDelta = stats.BytesTransmitted - lastNetBytesTransmitted
+	}
+
 	return stats
 }
 
@@ -1399,8 +1506,8 @@ func getNetworkStats() NetworkStats {
 func getCPUInfo() CPUInfo {
 	var cpuInfo CPUInfo
 
-	// 获取CPU使用率
-	percentages, err := cpu.Percent(time.Second, false)
+	// 获取CPU使用率 (non-blocking)
+	percentages, err := cpu.Percent(0, false)
 	if err == nil && len(percentages) > 0 {
 		cpuInfo.Usage = percentages[0]
 	}
@@ -1422,6 +1529,69 @@ func getMemoryInfo() MemoryInfo {
 	return memInfo
 }
 
+// getDiskInfo 获取磁盘信息
+func getDiskInfo() DiskInfo {
+	var diskInfo DiskInfo
+
+	usage, err := disk.Usage("/")
+	if err != nil {
+		return diskInfo
+	}
+
+	diskInfo.Usage = usage.UsedPercent
+
+	return diskInfo
+}
+
+// getLoadInfo 获取负载信息
+func getLoadInfo() LoadInfo {
+	var loadInfo LoadInfo
+
+	avg, err := load.Avg()
+	if err != nil {
+		return loadInfo
+	}
+
+	loadInfo.Load1 = avg.Load1
+	loadInfo.Load5 = avg.Load5
+	loadInfo.Load15 = avg.Load15
+
+	return loadInfo
+}
+
+// getConnectionInfo 获取连接信息
+func getConnectionInfo() ConnectionInfo {
+	now := time.Now().UnixMilli()
+	const refreshEveryMs = int64((15 * time.Second) / time.Millisecond)
+
+	connInfoCachedMu.Lock()
+	if connInfoCachedAt > 0 && now-connInfoCachedAt < refreshEveryMs {
+		v := connInfoCached
+		connInfoCachedMu.Unlock()
+		return v
+	}
+	connInfoCachedMu.Unlock()
+
+	var connInfo ConnectionInfo
+
+	connStats, err := psnet.Connections("tcp")
+	if err == nil {
+		connInfo.TCPConns = int64(len(connStats))
+	}
+
+	udpStats, err := psnet.Connections("udp")
+	if err == nil {
+		connInfo.UDPConns = int64(len(udpStats))
+	}
+
+	connInfoCachedMu.Lock()
+	connInfoCached = connInfo
+	connInfoCachedAt = now
+	connInfoCachedMu.Unlock()
+
+	return connInfo
+}
+
 // StartWebSocketReporterWithConfig 使用配置字段启动WebSocket报告器
 func StartWebSocketReporterWithConfig(addr string, secret string, http int, tls int, socks int, version string) *WebSocketReporter {
 
@@ -1503,6 +1673,210 @@ func (w *WebSocketReporter) handleTcpPing(data interface{}) (TcpPingResponse, er
 	return response, nil
 }
 
+// handleServiceMonitorCheck executes a service monitor check on this node.
+// It always returns a result (command execution is considered successful even if the check fails).
+func (w *WebSocketReporter) handleServiceMonitorCheck(data interface{}) (ServiceMonitorCheckResult, error) {
+	jsonData, err := json.Marshal(data)
+	if err != nil {
+		return ServiceMonitorCheckResult{}, fmt.Errorf("序列化检查数据失败: %v", err)
+	}
+
+	var req ServiceMonitorCheckRequest
+	if err := json.Unmarshal(jsonData, &req); err != nil {
+		return ServiceMonitorCheckResult{}, fmt.Errorf("解析检查请求失败: %v", err)
+	}
+
+	checkType := strings.ToLower(strings.TrimSpace(req.Type))
+	target := strings.TrimSpace(req.Target)
+	res := ServiceMonitorCheckResult{MonitorID: req.MonitorID}
+
+	if checkType != "tcp" && checkType != "icmp" {
+		res.Success = false
+		res.ErrorMessage = "不支持的检查类型"
+		return res, nil
+	}
+	if target == "" {
+		res.Success = false
+		res.ErrorMessage = "检查目标为空"
+		return res, nil
+	}
+
+	timeoutSec := req.TimeoutSec
+	if timeoutSec <= 0 {
+		timeoutSec = 5
+	}
+	timeout := time.Duration(timeoutSec) * time.Second
+
+	start := time.Now()
+
+	switch checkType {
+	case "tcp":
+		// Validate and normalize host:port.
+		_, _, splitErr := net.SplitHostPort(target)
+		if splitErr != nil {
+			res.Success = false
+			res.ErrorMessage = "无效的TCP目标"
+			res.LatencyMs = float64(time.Since(start).Milliseconds())
+			return res, nil
+		}
+		conn, dialErr := net.DialTimeout("tcp", target, timeout)
+		res.LatencyMs = float64(time.Since(start).Milliseconds())
+		if dialErr != nil {
+			res.Success = false
+			res.ErrorMessage = dialErr.Error()
+			return res, nil
+		}
+		_ = conn.Close()
+		res.Success = true
+		return res, nil
+
+	case "icmp":
+		rtt, pingErr := icmpPing(target, timeout)
+		res.LatencyMs = float64(rtt.Milliseconds())
+		if pingErr != nil {
+			res.Success = false
+			res.ErrorMessage = pingErr.Error()
+			return res, nil
+		}
+		res.Success = true
+		return res, nil
+	}
+
+	res.Success = false
+	res.ErrorMessage = "未知错误"
+	res.LatencyMs = float64(time.Since(start).Milliseconds())
+	return res, nil
+}
+
+func icmpPing(target string, timeout time.Duration) (time.Duration, error) {
+	start := time.Now()
+
+	target = strings.TrimSpace(target)
+	if target == "" {
+		return time.Since(start), fmt.Errorf("无效的ICMP目标")
+	}
+	// Avoid accepting URL-like targets.
+	if strings.Contains(target, "://") {
+		return time.Since(start), fmt.Errorf("无效的ICMP目标")
+	}
+	if strings.HasPrefix(target, "[") && strings.HasSuffix(target, "]") {
+		target = strings.TrimSuffix(strings.TrimPrefix(target, "["), "]")
+	}
+
+	ipAddr, err := net.ResolveIPAddr("ip", target)
+	if err != nil || ipAddr == nil || ipAddr.IP == nil {
+		if err == nil {
+			err = fmt.Errorf("unknown address")
+		}
+		return time.Since(start), fmt.Errorf("解析目标失败: %v", err)
+	}
+
+	isV4 := ipAddr.IP.To4() != nil
+	listenAddr := "0.0.0.0"
+	proto := 1
+	var echoType icmp.Type = ipv4.ICMPTypeEcho
+	var echoReplyType icmp.Type = ipv4.ICMPTypeEchoReply
+	networks := []string{"udp4", "ip4:icmp"}
+	if !isV4 {
+		listenAddr = "::"
+		proto = 58
+		echoType = ipv6.ICMPTypeEchoRequest
+		echoReplyType = ipv6.ICMPTypeEchoReply
+		networks = []string{"udp6", "ip6:ipv6-icmp"}
+	}
+
+	var conn *icmp.PacketConn
+	selectedNetwork := ""
+	var lastErr error
+	for _, nw := range networks {
+		c, err := icmp.ListenPacket(nw, listenAddr)
+		if err == nil {
+			conn = c
+			selectedNetwork = nw
+			break
+		}
+		lastErr = err
+	}
+	if conn == nil {
+		if lastErr != nil {
+			return time.Since(start), fmt.Errorf("创建ICMP连接失败: %v", lastErr)
+		}
+		return time.Since(start), fmt.Errorf("创建ICMP连接失败")
+	}
+	defer conn.Close()
+
+	id := os.Getpid() & 0xffff
+	seq := 1
+
+	wm := icmp.Message{
+		Type: echoType,
+		Code: 0,
+		Body: &icmp.Echo{
+			ID:   id,
+			Seq:  seq,
+			Data: []byte("FLVX-PING"),
+		},
+	}
+	wb, err := wm.Marshal(nil)
+	if err != nil {
+		return time.Since(start), err
+	}
+
+	_ = conn.SetDeadline(time.Now().Add(timeout))
+
+	var dst net.Addr
+	if strings.HasPrefix(selectedNetwork, "udp") {
+		dst = &net.UDPAddr{IP: ipAddr.IP, Zone: ipAddr.Zone}
+	} else {
+		dst = &net.IPAddr{IP: ipAddr.IP, Zone: ipAddr.Zone}
+	}
+
+	if _, err := conn.WriteTo(wb, dst); err != nil {
+		return time.Since(start), err
+	}
+
+	addrIP := func(a net.Addr) net.IP {
+		switch v := a.(type) {
+		case *net.IPAddr:
+			return v.IP
+		case *net.UDPAddr:
+			return v.IP
+		default:
+			return nil
+		}
+	}
+
+	rb := make([]byte, 1500)
+	for {
+		n, peer, err := conn.ReadFrom(rb)
+		if err != nil {
+			return time.Since(start), err
+		}
+		if p := addrIP(peer); p != nil && !p.Equal(ipAddr.IP) {
+			continue
+		}
+		rm, err := icmp.ParseMessage(proto, rb[:n])
+		if err != nil {
+			continue
+		}
+		if rm.Type != echoReplyType {
+			continue
+		}
+		echo, ok := rm.Body.(*icmp.Echo)
+		if !ok {
+			continue
+		}
+		if echo.Seq != seq {
+			continue
+		}
+		// For non-privileged endpoints, the kernel may choose the ID.
+		if !strings.HasPrefix(selectedNetwork, "udp") && echo.ID != id {
+			continue
+		}
+		return time.Since(start), nil
+	}
+}
+
 // tcpPingHost 执行TCP连接测试，返回平均连接时间和失败率
 func tcpPingHost(ip string, port int, count int, timeoutMs int) (float64, float64, error) {
 	var totalTime float64

openspec/changes/document-existing-specs/.openspec.yaml

@@ -1,2 +0,0 @@
-schema: spec-driven
-created: 2026-02-17

openspec/changes/document-existing-specs/design.md

@@ -1,29 +0,0 @@
-## Context
-
-FLVX is a distributed system consisting of a central management panel (Backend + Frontend) and multiple forwarding agents (Nodes). The backend manages configuration, users, and billing, while agents handle the actual traffic forwarding using a modified GOST v3 stack. Communication between the panel and agents is secured and synchronized.
-
-## Goals / Non-Goals
-
-**Goals:**
-- Document the high-level architecture of the system.
-- Describe the data model for users, tunnels, and nodes.
-- Explain the communication protocol between Panel and Agent.
-- Detail the authentication and authorization mechanisms.
-
-**Non-Goals:**
-- Refactoring the existing architecture.
-- Detailed code-level documentation of every function.
-- Changing the database schema.
-
-## Decisions
-
-- **Architecture**: The system follows a client-server model where the Panel acts as the server and Agents act as clients that pull configuration and push status.
-- **Data Model**: Core entities are Users, Nodes (Agents), Tunnels (Groups of rules), and Forwarding Rules.
-- **Communication**: Agents use a heartbeat mechanism to report status and fetch configuration updates. The protocol uses AES encryption with a pre-shared key (Node Secret).
-- **Authentication**: JWT for Frontend-Backend communication; API Key (Node Secret) for Agent-Backend communication.
-
-## Risks / Trade-offs
-
-- **Security**: The security of the agent communication relies heavily on the secrecy of the Node Secret.
-- **Scalability**: Centralized management might become a bottleneck with a very large number of agents.
-- **Complexity**: Synchronizing state across distributed agents introduces complexity in handling failures and inconsistencies.

openspec/changes/document-existing-specs/proposal.md

@@ -1,28 +0,0 @@
-## Why
-
-The current system lacks formal specification documents describing its capabilities. This makes it difficult for new developers to understand the intended behavior and for existing developers to ensure consistency when adding new features. Documenting the existing functionality will serve as a baseline for future changes and help in identifying gaps or inconsistencies.
-
-## What Changes
-
-- Create formal specification documents for core system capabilities.
-- Document user management features (roles, limits).
-- Document tunnel and forwarding management (protocols, rules).
-- Document agent interactions and management.
-- Document system-level configurations.
-
-## Capabilities
-
-### New Capabilities
-- `user-management`: Authentication, user roles, and resource limits.
-- `tunnel-management`: Creation and management of traffic tunnels (TCP/UDP).
-- `forwarding-rules`: Configuration of port forwarding and tunnel forwarding rules, including rate limiting.
-- `agent-management`: Management of forwarding agents, including installation and configuration synchronization.
-- `system-config`: Global system settings and configurations.
-
-### Modified Capabilities
-<!-- None, as this is a documentation effort for existing features. -->
-
-## Impact
-
-- **Documentation**: New spec files in `openspec/specs/`.
-- **No Code Changes**: This change is purely documentation-focused.

openspec/changes/document-existing-specs/specs/agent-management/spec.md

@@ -1,29 +0,0 @@
-## ADDED Requirements
-
-### Requirement: Agent Registration
-The system SHALL require new agents (Nodes) to register using a unique node key/secret.
-
-#### Scenario: Node Connection
-- **WHEN** a new agent starts up with a valid configuration
-- **THEN** it connects to the backend and is registered as active.
-
-### Requirement: Heartbeat Monitoring
-The system SHALL monitor the status of all registered agents using periodic heartbeats.
-
-#### Scenario: Agent Status
-- **WHEN** an agent sends periodic heartbeats
-- **THEN** the system updates its last-seen timestamp and marks it as online.
-
-### Requirement: Configuration Sync
-The system MUST synchronize configuration changes (tunnels, rules) to agents securely and reliably.
-
-#### Scenario: Push Config
-- **WHEN** a configuration change is made in the panel
-- **THEN** the agent receives the updated configuration via the next heartbeat or push mechanism.
-
-### Requirement: Version Management
-The system SHOULD track the version of the agent software running on each node.
-
-#### Scenario: Version Reporting
-- **WHEN** an agent connects
-- **THEN** it reports its version number to the backend for tracking.

openspec/changes/document-existing-specs/specs/forwarding-rules/spec.md

@@ -1,22 +0,0 @@
-## ADDED Requirements
-
-### Requirement: Port Forwarding Rules
-The system SHALL support configuring port forwarding rules, defining the listening port on the node and the destination IP/port.
-
-#### Scenario: Rule Configuration
-- **WHEN** an admin creates a port forwarding rule
-- **THEN** the rule is stored and synchronized to the assigned node.
-
-### Requirement: Rate Limiting
-The system SHALL support configuring bandwidth rate limits for tunnels and users.
-
-#### Scenario: Bandwidth Restriction
-- **WHEN** a rate limit is applied to a user
-- **THEN** their total bandwidth usage does not exceed the specified limit across all their tunnels.
-
-### Requirement: Traffic Accounting
-The system MUST track incoming and outgoing traffic volume for each tunnel and user for billing and quota enforcement.
-
-#### Scenario: Traffic Calculation
-- **WHEN** traffic flows through a tunnel
-- **THEN** the system increments the user's traffic usage counter accurately.

openspec/changes/document-existing-specs/specs/system-config/spec.md

@@ -1,22 +0,0 @@
-## ADDED Requirements
-
-### Requirement: Site Settings
-The system SHALL allow customization of the site title, logo, and other branding elements.
-
-#### Scenario: Update Branding
-- **WHEN** an administrator changes the site logo
-- **THEN** the new logo is displayed across the interface.
-
-### Requirement: Notification Settings
-The system SHALL support configuring notifications for user registration, traffic limits, and other events.
-
-#### Scenario: User Limit Alert
-- **WHEN** a user approaches their traffic quota
-- **THEN** a notification is sent to the user/admin.
-
-### Requirement: Backup & Restore
-The system SHOULD provide a mechanism to backup and restore database configurations.
-
-#### Scenario: Restore Database
-- **WHEN** initiating a restore operation
-- **THEN** the system accepts a valid backup file and overwrites the current database state.

openspec/changes/document-existing-specs/specs/tunnel-management/spec.md

@@ -1,22 +0,0 @@
-## ADDED Requirements
-
-### Requirement: Tunnel Creation
-The system SHALL allow administrators to create tunnels, specifying protocols (TCP, UDP), listening ports, and destination endpoints.
-
-#### Scenario: Create TCP Tunnel
-- **WHEN** an admin creates a new TCP tunnel configuration
-- **THEN** the backend stores the tunnel definition and assigns it to a node.
-
-### Requirement: Tunnel Forwarding Configuration
-The system SHALL support both standard port forwarding (listening on a port and forwarding to a destination) and tunnel forwarding modes.
-
-#### Scenario: Configure Port Forwarding
-- **WHEN** configuring a tunnel for port forwarding
-- **THEN** traffic arriving at the specified port is forwarded to the destination IP:port.
-
-### Requirement: Tunnel Assignment
-The system SHALL allow tunnels to be assigned to specific users, tracking their usage against the user's quota.
-
-#### Scenario: User Tunnel Usage
-- **WHEN** a user is assigned a tunnel
-- **THEN** traffic passing through that tunnel is accounted for under the user's usage.

openspec/changes/document-existing-specs/specs/user-management/spec.md

@@ -1,29 +0,0 @@
-## ADDED Requirements
-
-### Requirement: User Registration
-The system SHALL allow new users to register an account with a username and password.
-
-#### Scenario: Successful Registration
-- **WHEN** a user submits valid registration details
-- **THEN** a new user account is created and the user can log in.
-
-### Requirement: User Authentication
-The system MUST authenticate users using JWT tokens. The `Authorization` header MUST contain the raw token without a `Bearer` prefix.
-
-#### Scenario: Valid Login
-- **WHEN** a user provides correct credentials
-- **THEN** the system returns a valid JWT token.
-
-### Requirement: Role Management
-The system SHALL support different user roles, specifically Administrator and Regular User, with distinct permissions.
-
-#### Scenario: Admin Access
-- **WHEN** an administrator logs in
-- **THEN** they have access to system-wide settings and all user management functions.
-
-### Requirement: Resource Quotas
-The system SHALL allow administrators to set traffic limits and connection limits for individual users.
-
-#### Scenario: Traffic Limit Enforcement
-- **WHEN** a user exceeds their traffic quota
-- **THEN** the system prevents further traffic forwarding for that user.

openspec/changes/document-existing-specs/tasks.md

@@ -1,30 +0,0 @@
-## 1. User Management Verification
-
-- [ ] 1.1 Verify User Registration logic in backend
-- [ ] 1.2 Verify JWT Authentication implementation
-- [ ] 1.3 Verify Role Management checks
-- [ ] 1.4 Verify Quota Enforcement logic
-
-## 2. Tunnel Management Verification
-
-- [ ] 2.1 Verify Tunnel Creation API
-- [ ] 2.2 Verify Forwarding Configuration parsing
-- [ ] 2.3 Verify Tunnel Assignment logic
-
-## 3. Forwarding Rules Verification
-
-- [ ] 3.1 Verify Port Forwarding rule processing
-- [ ] 3.2 Verify Rate Limiting implementation (token bucket/leaky bucket?)
-- [ ] 3.3 Verify Traffic Accounting mechanisms
-
-## 4. Agent Management Verification
-
-- [ ] 4.1 Verify Agent Registration handshake
-- [ ] 4.2 Verify Heartbeat processing
-- [ ] 4.3 Verify Config Sync protocol
-
-## 5. System Config Verification
-
-- [ ] 5.1 Verify Site Settings API
-- [ ] 5.2 Verify Notification triggers
-- [ ] 5.3 Verify Backup/Restore functionality

openspec/config.yaml

@@ -1,20 +0,0 @@
-schema: spec-driven
-
-# Project context (optional)
-# This is shown to AI when creating artifacts.
-# Add your tech stack, conventions, style guides, domain knowledge, etc.
-# Example:
-#   context: |
-#     Tech stack: TypeScript, React, Node.js
-#     We use conventional commits
-#     Domain: e-commerce platform
-
-# Per-artifact rules (optional)
-# Add custom rules for specific artifacts.
-# Example:
-#   rules:
-#     proposal:
-#       - Keep proposals under 500 words
-#       - Always include a "Non-goals" section
-#     tasks:
-#       - Break tasks into chunks of max 2 hours

openspec/project.md

@@ -1,52 +0,0 @@
-# Project Overview
-
-**Name**: FLVX (Flux Panel)
-**Description**: Traffic forwarding management system built on a forked GOST v3 stack. It provides a web-based panel for managing traffic tunnels, users, and forwarding rules.
-**Repository**: Monorepo containing Admin API, Web UI, and Forwarding Agent.
-
-## Tech Stack
-
-### Backend (`go-backend/`)
-- **Language**: Go
-- **Database**: SQLite (default), PostgreSQL (supported)
-- **Framework**: Standard library `net/http` (no heavy framework)
-- **ORM**: None (Raw SQL via `database/sql`)
-
-### Frontend (`vite-frontend/`)
-- **Framework**: React
-- **Build Tool**: Vite (using `rolldown-vite` experimental bundler)
-- **UI Library**: HeroUI
-- **Styling**: Tailwind CSS
-- **Mode**: Hybrid (Desktop + Mobile WebView support)
-
-### Agent (`go-gost/`)
-- **Language**: Go
-- **Base**: Fork of `gost` v3
-- **Extensions**: Custom extensions in `go-gost/x/`
-
-### Infrastructure
-- **Containerization**: Docker, Docker Compose (v4/v6)
-- **CI/CD**: GitHub Actions
-- **Installers**: Shell scripts (`panel_install.sh`, `install.sh`)
-
-## Architecture
-
-- **Panel**: Central management server (Go Backend + React Frontend).
-- **Agent**: Forwarding node running on remote servers.
-- **Communication**:
-    - Frontend -> Backend: REST API (JWT Auth, raw token in header).
-    - Agent -> Backend: AES-encrypted heartbeat/config sync.
-
-## Conventions
-
-- **Authentication**: `Authorization` header expects raw JWT token (do NOT add `Bearer ` prefix).
-- **API Response**: Standard envelope `{code, msg, data, ts}` (code 0 = success).
-- **Database**: Backend uses raw SQL queries. Do not introduce an ORM.
-- **File Structure**: Flat monorepo with language-prefixed directories (`go-backend`, `go-gost`).
-- **Protobuf**: Do not edit generated `.pb.go` files manually.
-
-## Development
-
-- **Backend Build**: `cd go-backend && make build`
-- **Frontend Dev**: `cd vite-frontend && npm run dev`
-- **Agent Run**: `cd go-gost && go run .`

plans/013-tunnel-entry-port-validation.md

@@ -0,0 +1,21 @@
+# 013 - 隧道入口端口校验不严格修复
+
+## Issue
+- GitHub Issue: [#373](https://github.com/Sagit-chu/flvx/issues/373)
+
+## 修复方案
+
+在 `syncTunnelForwardsEntryPorts` 中实现逐节点端口分配：
+
+- **旧入口节点**：保留原端口不变
+- **新入口节点**：通过 `resolvePortForNewEntryNode` 决策：
+  - 参考端口在范围内且未被占用 → 跟随设置一样的端口
+  - 参考端口超出范围或被占用 → 通过 `pickRandomPortForNode` 为该节点单独随机分配
+
+## 任务清单
+
+- [x] 1. 实现 `pickRandomPortForNode` 辅助方法（单节点端口随机分配）
+- [x] 2. 实现 `resolvePortForNewEntryNode` 方法（端口决策逻辑）
+- [x] 3. 重写 `syncTunnelForwardsEntryPorts` 为逐节点分配
+- [x] 4. 移除不再需要的 `isPortValidForAllEntryNodes`
+- [x] 5. 构建通过 + 全量测试通过

plans/037-monitoring-metrics-health.md

@@ -0,0 +1,104 @@
+# 037 - Monitoring: Node Metrics + Service Health Checks
+
+## Context
+This worktree introduces a monitoring feature set:
+- Node runtime metrics streamed via WebSocket (agent -> panel -> admin clients)
+- Metrics ingestion + retention in panel DB
+- Service monitoring (TCP/ICMP checks only) + result storage
+- Frontend monitor view (charts + monitor CRUD + run + results)
+- Dedicated monitor page (`/monitor`) that works for authorized non-admin users
+
+The initial implementation landed without a plan doc and had several correctness issues (API JSON shape mismatch, wrong time units, contract test hangs under SQLite single-connection mode, etc.). This plan documents what exists, what was fixed, and what is still incomplete/needs decisions.
+
+## Goals
+- Metrics endpoints return stable JSON fields matching frontend types.
+- Contract tests cover metrics + monitor CRUD and are deterministic.
+- WebSocket metric messages update node cards correctly.
+- Monitoring view queries the correct time range and renders timestamps correctly.
+- go-gost/x unit tests do not depend on a local config.json.
+
+## Non-goals (for this plan)
+- A full monitor scheduling system (jitter/backoff/concurrency budgets/per-monitor next-run) beyond the current simple loop.
+- Building a full alerting pipeline (notifications, thresholds, paging).
+
+## Current Status (as of this worktree)
+- Backend models updated with JSON tags for monitoring structs.
+- Handler endpoints for metrics + service monitors added.
+- Metrics ingestion service implemented with buffering + retention pruning.
+- Health checker implemented (panel-side when `nodeId == 0`; node-executed via WS when `nodeId > 0`) and background jobs wired.
+- Frontend monitor view added; build passes.
+- Contract tests for monitoring added.
+- Monitoring endpoints are accessible by admin users and non-admin users explicitly authorized by admin (via `monitor_permission`).
+- Frontend exposes monitoring via a dedicated `/monitor` page; admin can grant/revoke monitoring permission from the User permissions modal.
+- Frontend includes tunnel metrics charts (backed by `/api/v1/monitor/tunnels` list + `/api/v1/monitor/tunnels/:id/metrics`).
+
+## Known Semantics Gaps (need decisions)
+- `service_monitor.intervalSec` is best-effort (checker ticks every 30s; intervals shorter than that won't run faster).
+- `service_monitor_result.success` is stored as int (0/1). Frontend currently treats it as number; decide if API should expose boolean.
+
+## Admin Authorization API
+Monitoring permission management (admin-only):
+- `GET /api/v1/monitor/permission/list`
+- `POST /api/v1/monitor/permission/assign` body: `{ "userId": 123 }`
+- `POST /api/v1/monitor/permission/remove` body: `{ "userId": 123 }`
+
+## Checklist
+
+### Phase 1: Correctness + Contracts
+- [x] Align monitoring JSON response fields with frontend/contract expectations (add json tags or DTO mapping).
+- [x] Fix frontend monitor time range query (use ms start/end; avoid `start=60`).
+- [x] Fix frontend timestamp rendering (treat timestamp as UnixMilli).
+- [x] Fix node realtime metric speed field compatibility (support snake_case speed fields).
+- [x] Fix SQLite contract hang by ensuring tunnel-entry precheck uses tx-safe DB reads (no nested connection acquisition).
+- [x] Ensure monitoring contract tests pass.
+
+### Phase 2: Semantics Alignment (Decide + Implement)
+- [x] Decide "service monitors run where":
+  - Option B: node-executed when `nodeId > 0` (chosen)
+- [ ] Define interval semantics:
+  - Per-monitor next-run scheduling vs global scan loop
+  - Backoff on failures
+  - Maximum monitors + runtime cost guardrails
+- [ ] Standardize API type for `success`:
+  - Keep int for backward compatibility, or
+  - Return boolean in API responses (DTO) while storing int in DB
+
+### Phase 2.1: Partial Implementation (No Semantics Decision Yet)
+- [x] Honor `intervalSec` best-effort in panel-side checker (min cadence still bound by global loop).
+
+### Phase 2.2: Node-Executed Checks
+- [x] Add a WebSocket command for node-executed monitor checks (`ServiceMonitorCheck`).
+- [x] Panel health checker dispatches checks to the specified node when `nodeId > 0`.
+- [x] Allow unrestricted targets by policy; restrict monitoring endpoints to admin + explicitly authorized users.
+- [x] Remove HTTP checks; service monitoring supports only `tcp` and `icmp`.
+
+### Phase 3: Hardening + Performance
+- [x] Add query limits/guards for metrics endpoints (max range, max rows) to avoid accidental full-history pulls.
+- [ ] Consider indexing review and retention configurability (env or config table).
+- [ ] Review concurrency: ingestion buffer flush goroutine spawning and DB write pressure.
+- [x] Add minimal UI affordances: time range selector, empty/error states, and service monitor run/results UI.
+- [x] Ensure monitoring UI works for authorized non-admin users (dedicated `/monitor` page; no reliance on admin-only `/node/*`).
+
+### Phase 4: Hygiene
+- [x] Add `.entire/metadata/` to `.gitignore` (should never be committed).
+- [ ] Add a short developer note in docs/README if needed (API endpoints + semantics).
+
+## Test Plan
+Backend:
+```bash
+cd go-backend && go test ./... -count=1
+cd go-backend && go test ./tests/contract -count=1 -timeout 120s
+```
+
+Agent fork:
+```bash
+cd go-gost/x && go test ./... -count=1
+```
+
+Frontend:
+```bash
+cd vite-frontend && npm run build
+```
+
+## Notes
+- Node-executed checks can be used for internal probing by design; access is restricted to administrators.

plans/038-monitoring-bugfixes-and-optimizations.md

@@ -0,0 +1,59 @@
+# 038 - Monitoring Bug Fixes + Optimizations
+
+## Context
+Monitoring in FLVX currently spans:
+- Agent -> panel WebSocket realtime system metrics (CPU/mem/disk/net/load/conns)
+- Panel-side ingestion + retention pruning (`node_metric`)
+- Service monitors (TCP/ICMP) with scheduled checks + stored results
+- Frontend monitor page (`/monitor`) with charts + monitor CRUD/run/results
+
+While the feature set works end-to-end, there are a few correctness footguns and a couple of obvious performance hot spots (agent-side sampling cost and frontend N+1 polling patterns).
+
+## Goals
+- Service monitor updates do not accidentally clear `nodeId` / `enabled` when fields are omitted.
+- Checker cadence is explicit (intervals below the scan cadence are clamped / best-effort).
+- Reduce frontend requests for service monitor status (avoid per-monitor polling).
+- Reduce agent sampling overhead and DB write volume without breaking UI expectations.
+- Avoid misclassifying arbitrary JSON as a metric message on the WS channel.
+
+## Non-goals
+- A full scheduler (per-monitor next-run queue, jitter/backoff, concurrency budgets).
+- Alerting/notifications.
+- Implementing full tunnel-metrics ingestion (connections/errors/latency) beyond current endpoints.
+
+## Checklist
+
+### Phase 1: Backend Correctness + Hardening
+- [x] Make `/api/v1/monitor/services/update` treat `nodeId` and `enabled` as optional fields (no accidental zeroing).
+- [x] Clamp `intervalSec` to a minimum that matches the checker scan cadence (and apply the same clamp in the checker).
+- [x] Add `GET /api/v1/monitor/services/latest-results` returning the latest result per monitor (for frontend list rendering).
+- [x] WS metric parsing: only treat messages as metrics when they look like a system-metric payload.
+
+### Phase 2: Frontend UX + Request Reduction
+- [x] Fix “立即检查” toast severity (failure should be an error toast).
+- [x] Use `latest-results` endpoint to render service monitor status without N+1 polling.
+- [x] Add a small hint when chart data is truncated by backend row limits.
+
+### Phase 3: Agent Sampling Optimizations
+- [x] Reduce default WS metric send interval (2s -> 5s).
+- [x] Make CPU sampling non-blocking and cache heavy metrics (e.g. connection counts) to reduce per-sample cost.
+
+## Test Plan
+Backend:
+```bash
+cd go-backend && go test ./... -count=1
+```
+
+Agent fork:
+```bash
+cd go-gost/x && go test ./... -count=1
+```
+
+Frontend (best-effort in this environment):
+```bash
+cd vite-frontend && npm run build
+```
+
+## Rollout Notes
+- Agent sampling interval change reduces metric resolution and DB growth; charts remain usable and realtime UI remains responsive.
+- Existing monitors with very small `intervalSec` are best-effort; effective cadence remains bounded by the checker scan loop.

plans/039-monitoring-tunnel-metrics-ingestion.md

@@ -0,0 +1,41 @@
+# 039 - Monitoring: Tunnel Metrics Ingestion
+
+## Context
+The `/monitor` UI includes tunnel metric charts backed by:
+- `GET /api/v1/monitor/tunnels` (list)
+- `GET /api/v1/monitor/tunnels/:id/metrics` (timeseries)
+
+The backend has the `tunnel_metric` table + query endpoints, but there is no production code path that writes tunnel metrics. As a result, tunnel charts are typically empty.
+
+## Goal
+Persist tunnel traffic timeseries based on agent flow uploads (`POST /flow/upload`).
+
+## Scope
+- Write `tunnel_metric` rows from flow uploads.
+- Keep write volume bounded (aggregate per minute).
+- Provide contract coverage that a flow upload creates tunnel metrics.
+
+## Non-goals
+- Populate connections/errors/latency for tunnel metrics (remain 0 for now).
+- A full aggregation pipeline across multiple nodes per tunnel at query time.
+
+UI note:
+- The tunnel chart only exposes the Traffic view for now; other tabs are hidden.
+
+## Design
+- Agent reports per-service traffic deltas via `/flow/upload` with items `{n,u,d}`.
+- Backend derives `forward_id` from service name (`<forwardID>_<userID>_<userTunnelID>[...suffix]`).
+- Map `forward_id -> tunnel_id` in batch.
+- Aggregate per `(node_id, tunnel_id, minute_bucket)` and upsert into `tunnel_metric` using an UPDATE-then-INSERT fallback.
+
+## Checklist
+- [x] Add repository helper: map forward IDs to tunnel IDs.
+- [x] Add repository helper: upsert per-minute tunnel metric buckets.
+- [x] Extend `/flow/upload` handler to record tunnel metrics from incoming items.
+- [x] Add contract test verifying flow upload produces tunnel metrics.
+- [x] Run backend tests.
+
+## Test Plan
+```bash
+cd go-backend && go test ./... -count=1
+```

plans/040-service-monitor-configurable-limits-and-ui-hints.md

@@ -0,0 +1,40 @@
+# 040 - Service Monitor Limits Config + UI Hints
+
+## Goal
+Make service monitor interval/timeout constraints configurable (instead of hard-coded clamps) and make the UI clearly communicate the effective limits.
+
+## Current Pain
+- Backend clamps `intervalSec` and `timeoutSec` with hard-coded constants.
+- Checker scan cadence is also hard-coded, so users can set values that will never be honored.
+- Frontend form does not explain allowed ranges or why values may change.
+
+## Approach
+- Add frontend-configurable limits stored in `vite_config` (with safe defaults matching current behavior).
+- Backend always normalizes using the configured limits.
+- Expose the current limits via a monitoring endpoint so the UI can render accurate hints.
+- Frontend shows min/max and validates before submit.
+- Admin can edit the limits on `/config`.
+
+## Config Keys (vite_config)
+- `service_monitor_checker_scan_interval_sec` (default: 30)
+- `service_monitor_min_interval_sec` (default: 30; auto-raised to at least scan interval)
+- `service_monitor_default_interval_sec` (default: 60)
+- `service_monitor_min_timeout_sec` (default: 1)
+- `service_monitor_default_timeout_sec` (default: 5)
+- `service_monitor_max_timeout_sec` (default: 60)
+
+## Checklist
+Backend:
+- [x] Introduce shared `ServiceMonitorLimits` config loader.
+- [x] Use limits for create/update normalization.
+- [x] Use limits in checker (scan interval + timeout clamp).
+- [x] Add `GET /api/v1/monitor/services/limits` to return current limits.
+
+Frontend:
+- [x] Fetch limits once and render input descriptions.
+- [x] Validate interval/timeout client-side and show inline errors.
+- [x] Add `/config` items to edit the `vite_config` keys.
+
+Verification:
+- [x] `cd go-backend && go test ./... -count=1`
+- [x] `cd vite-frontend && npm run lint && npm run build`

plans/041-monitoring-reliability-realtime-and-ux-hardening.md

@@ -0,0 +1,224 @@
+# 041 - Monitoring Reliability, Realtime, and UX Hardening
+
+## Context
+Current monitoring support in FLVX already covers three major areas:
+- Node runtime metrics from agent WebSocket telemetry, buffered into `node_metric`, exposed by `/api/v1/monitor/nodes*`, and rendered on `/monitor`.
+- Tunnel metrics derived from `/flow/upload`, stored in `tunnel_metric`, exposed by `/api/v1/monitor/tunnels*`, and rendered on `/monitor`.
+- Service monitoring for `tcp` and `icmp`, including CRUD, scheduled checks, manual run, history, and non-admin authorization via `monitor_permission`.
+
+The feature set is usable, but the audit found several correctness, reliability, and UX gaps:
+- The monitor page is not truly realtime and can lag DB ingestion by tens of seconds.
+- Tunnel metrics are only partially implemented and are not aggregated correctly for multi-node tunnels.
+- Some monitoring writes fail silently, which can hide data-loss and retention issues.
+- Service monitor scheduling is functional but too naive for larger monitor sets and restart scenarios.
+- The monitoring UI exposes incomplete semantics, weak freshness cues, and inconsistent permission/error affordances.
+- Several monitoring endpoints and edge cases still lack direct automated coverage.
+
+This plan collects all currently known monitoring follow-up work into one implementation document.
+
+## Goals
+- Make node monitoring data freshness explicit and reduce stale or misleading chart behavior.
+- Make tunnel metrics correct for multi-node tunnels and align schema/query/UI semantics.
+- Harden service monitor scheduling, persistence, and cleanup behavior.
+- Improve observability so monitoring ingestion and result writes never fail silently.
+- Upgrade the monitoring UI so operators can understand status, freshness, scope, and failures at a glance.
+- Expand automated coverage for all monitoring APIs and the highest-risk aggregation/scheduler cases.
+
+## Non-goals
+- Add a full alerting or notification pipeline.
+- Add brand-new monitor protocols beyond the current `tcp` and `icmp` scope.
+- Build a large analytics dashboard outside the existing monitoring page structure.
+- Introduce frontend test infrastructure for broad component/unit testing unless required by an implementation step.
+
+## Audit Findings To Address
+- Node metrics on `/monitor` are DB-polled rather than realtime-streamed.
+- Node metrics are buffered for 30s, so charts can lag behind observed node state.
+- Tunnel metrics are stored per `(tunnel_id, node_id, timestamp)` but queried and rendered as if they were already tunnel-level aggregates.
+- Tunnel metric minute-bucket upsert uses update-then-insert without uniqueness guarantees.
+- Tunnel metrics only populate `bytesIn` and `bytesOut`; `connections`, `errors`, and `avgLatencyMs` are placeholder values.
+- Node/tunnel/service-monitor writes can fail silently due to ignored errors.
+- Service monitor scheduler is serial and uses in-memory `lastRun`, causing restart skew and slow-monitor head-of-line blocking.
+- Deleting a service monitor does not clean up related historical results.
+- `expectedCode` exists on the model but is not implemented in behavior or UX.
+- The monitoring page/menu is exposed before permission is known, leading to avoidable denied-entry UX.
+- Service monitor UI does not clearly show latest result freshness, last check time, or whether a displayed row is stale.
+- Chart labels and units are not operator-friendly for long time windows and network-heavy views.
+- Monitoring API coverage is incomplete for list, permission, limits, latest-results, multi-node tunnel aggregation, and concurrency paths.
+
+## Design
+
+### 1. Node Monitoring Freshness and Realtime Model
+- Keep the existing WebSocket node telemetry stream as the source of live state.
+- Preserve DB-backed metrics queries for historical charts, but explicitly separate them from live cards/status.
+- On `/monitor`, add a lightweight realtime subscription path reusing the existing admin WebSocket feed already used by the node page.
+- Use realtime events for:
+  - node online/offline state,
+  - a small “latest value” strip or summary above charts,
+  - freshness timestamp display.
+- Keep charts historical and DB-backed by default, but add a visible freshness hint such as:
+  - `历史图表，最近落库延迟约 0-30s`, or
+  - `最近入库时间: ...`.
+- Do not remove buffered ingestion immediately; first make lag transparent in UI and observable in logs/metrics.
+- Optional second-step optimization: reduce flush interval or add a bounded flush-on-latest-view mode if DB pressure remains acceptable.
+
+### 2. Tunnel Metrics Data Model and Query Semantics
+- Decide and document one API contract:
+  - `GET /api/v1/monitor/tunnels/:id/metrics` must return tunnel-level aggregated series for the selected time range, not raw per-node rows.
+- Keep storage per `(tunnel_id, node_id, timestamp)` because it is useful for future drill-down.
+- Change query behavior so the tunnel metrics endpoint aggregates rows by timestamp across all nodes for the tunnel:
+  - `SUM(bytes_in)`,
+  - `SUM(bytes_out)`,
+  - `SUM(connections)`,
+  - `SUM(errors)`,
+  - `AVG` or weighted-average strategy for latency, if latency is later implemented.
+- Return a single point per timestamp to the frontend.
+- If future node drill-down is needed, add a separate endpoint rather than mixing per-node rows into the current chart API.
+
+### 3. Tunnel Metric Upsert Safety
+- Replace the current update-then-insert fallback with a uniqueness-backed upsert strategy.
+- Add a unique index on `(tunnel_id, node_id, timestamp)`.
+- Implement DB-safe upsert behavior compatible with SQLite and PostgreSQL via GORM clauses or equivalent dialect-safe SQL.
+- Preserve additive semantics for traffic counters inside the bucket.
+- Add concurrency coverage proving that parallel uploads for the same bucket do not create duplicate rows.
+
+### 4. Tunnel Metric Scope Clarification
+- Short term: make the UI and API explicitly traffic-only where the backend only has traffic truth.
+- Remove or hide unsupported tunnel metric modes from the current UX until real data exists.
+- Do not expose zero-filled placeholders as if they were valid telemetry.
+- Keep schema fields if future support is planned, but label them as unimplemented in code comments and avoid rendering them as live features.
+
+### 5. Service Monitor Scheduler Hardening
+- Replace the current fully serial best-effort loop with bounded concurrency:
+  - retain a global scan loop or next-run calculation,
+  - collect monitors due for execution,
+  - execute them with a configurable worker limit,
+  - avoid one slow node/target delaying all others.
+- Move scheduling semantics from pure in-memory `lastRun` toward persisted or history-derived next-run safety:
+  - on restart, do not fire an uncontrolled burst for all monitors if they just ran;
+  - use latest persisted result timestamp or a persisted scheduler state to calculate due-ness.
+- Keep interval clamping behavior aligned with configured limits.
+- Continue supporting local panel execution for `tcp` and node execution for `tcp`/`icmp`.
+
+### 6. Service Monitor Data Lifecycle
+- Define monitor deletion semantics explicitly:
+  - either cascade-delete historical `service_monitor_result` rows when a monitor is deleted, or
+  - retain them intentionally and exclude orphan rows from latest/list endpoints.
+- Preferred approach: delete associated results with the monitor so the UI/API model stays simple.
+- Either implement `expectedCode` fully or remove it from the model/API surface for now.
+- Because service monitoring currently supports only `tcp` and `icmp`, and no HTTP checks are implemented, `expectedCode` should likely be removed from the data model/API until a real HTTP monitor exists.
+
+### 7. Observability and Failure Handling
+- Stop swallowing monitoring persistence errors.
+- For all node/tunnel/service-monitor writes:
+  - log structured errors with entity identifiers and operation names,
+  - increment internal counters if an existing metrics/logging primitive exists,
+  - keep request/loop behavior resilient, but make failure visible.
+- Apply this to:
+  - node metric batch flush,
+  - tunnel metric bucket writes,
+  - scheduled service monitor result writes,
+  - manual service monitor result writes,
+  - pruning failures.
+- Avoid user-facing hard failures for background ingestion, but surface operational diagnostics in logs.
+
+### 8. Monitoring UI Semantics and Navigation
+- Keep `/monitor` accessible only to authenticated users, but improve pre-entry affordances:
+  - hide or disable the navigation item for users without monitor permission when role/permission data is known,
+  - or show a locked state with explanation instead of allowing a full denied page transition.
+- Preserve the backend permission check as the source of truth.
+- On the page itself, upgrade semantics:
+  - distinguish `enabled/disabled` from `healthy/unhealthy` in the service monitor table,
+  - show `last checked at`,
+  - show `latest result` separately from monitor switch state,
+  - show whether the latest displayed result is stale.
+- Add an at-a-glance monitoring summary near the top:
+  - online/offline node counts,
+  - monitors healthy/unhealthy/disabled counts,
+  - latest data freshness text.
+
+### 9. Monitoring UI Readability Improvements
+- Improve chart axis labeling for long ranges:
+  - use date + time formatting for 24h windows,
+  - keep shorter labels for short ranges.
+- Format bytes and rates into human-readable units (`KB/s`, `MB/s`, `GB`) instead of raw integers.
+- Expose clear empty states and fetch-error states instead of silent failures.
+- Make tunnel charts explicitly say `流量趋势` if only traffic is supported.
+- Show `statusCode` in the results modal only if the corresponding monitor type ever uses it; otherwise omit it.
+
+### 10. API and Test Coverage Expansion
+- Add contract coverage for:
+  - `GET /api/v1/monitor/nodes`,
+  - `GET /api/v1/monitor/tunnels`,
+  - `GET /api/v1/monitor/services/latest-results`,
+  - `GET /api/v1/monitor/services/limits`,
+  - monitor permission list/assign/remove endpoints.
+- Add backend tests for:
+  - multi-node tunnel aggregation returning one point per timestamp,
+  - tunnel upsert concurrency safety,
+  - service monitor restart/due scheduling semantics,
+  - service monitor delete cleanup behavior,
+  - background write failure logging where practical.
+- Keep existing build/test targets green for backend, agent, and frontend.
+
+## Checklist
+
+### Phase 1: Correctness Fixes
+- [x] Aggregate `GET /api/v1/monitor/tunnels/:id/metrics` by timestamp across all node rows for the selected tunnel.
+- [x] Add a unique index for tunnel metric minute buckets and replace the race-prone update-then-insert flow with safe upsert logic.
+- [x] Stop exposing unsupported tunnel metric dimensions (`connections`, `errors`, `latency`) as active UI features while the backend still stores placeholders.
+- [x] Define and implement service monitor deletion cleanup so history does not leave orphaned result rows.
+- [x] Remove or fully implement `expectedCode`; do not keep dead monitoring fields in the live API/model contract.
+
+### Phase 2: Reliability and Scheduling
+- [x] Replace serial service monitor execution with bounded-concurrency execution for due monitors.
+- [x] Persist or derive service monitor next-run behavior so process restarts do not trigger uncontrolled immediate reruns.
+- [x] Ensure monitor scheduler semantics remain aligned with configured min interval and checker scan cadence.
+- [x] Add structured logging for all monitoring persistence failures and prune failures.
+- [x] Audit all ignored monitoring write errors and convert them into visible operational diagnostics.
+
+### Phase 3: Realtime and Freshness UX
+- [x] Reuse the existing admin WebSocket stream on `/monitor` for live node status and latest-value freshness indicators.
+- [x] Add visible chart freshness metadata so users know historical charts are DB-backed and may lag ingestion.
+- [x] Decide whether to reduce node metric flush interval after instrumentation confirms acceptable DB impact (decision: keep 30s default for now; revisit after observing DB write rate and UI staleness in production).
+- [x] Add a monitoring summary strip showing online nodes, unhealthy monitors, and latest data time.
+
+### Phase 4: Monitoring Page UX Cleanup
+- [x] Separate service monitor switch state (`启用/禁用`) from probe health (`成功/失败`).
+- [x] Add `last checked at` to the service monitor list and results modal context.
+- [x] Mark stale results clearly when the latest result is older than the configured interval budget.
+- [x] Format traffic and speed values in human-readable units instead of raw bytes.
+- [x] Improve chart time labels for 24h windows to include date context.
+- [x] Replace silent frontend fetch failures with explicit inline error or toast handling.
+- [x] Rename or relabel tunnel chart UI to make its current scope unambiguous.
+
+### Phase 5: Permission and Navigation UX
+- [x] Avoid showing a fully interactive monitor nav entry to users who lack monitoring permission once permission state is known.
+- [x] Preserve backend authorization as the final gate and keep denied responses intact.
+- [x] Improve denied-state copy so users understand whether they need admin grant vs role change.
+
+### Phase 6: Automated Coverage
+- [x] Add contract tests for monitor node list, tunnel list, latest service monitor results, limits, and permission endpoints.
+- [x] Add contract or repository tests for multi-node tunnel aggregation correctness.
+- [x] Add concurrency tests for tunnel metric upsert safety.
+- [x] Add scheduler tests covering restart behavior, due monitor selection, and slow-monitor isolation.
+- [x] Keep existing monitoring contract tests passing after all changes.
+
+## Implementation Notes
+- Prefer backward-compatible API changes where possible, but favor correctness over preserving misleading tunnel metric semantics.
+- Do not introduce a fake realtime chart if the data source remains DB-backed; label it honestly.
+- If permission visibility requires an extra frontend capability call, keep it lightweight and cacheable.
+- If schema/index changes are introduced, they must remain compatible with both SQLite and PostgreSQL.
+
+## Final Verification Targets
+- [x] `GET /api/v1/monitor/tunnels/:id/metrics` returns one aggregated point per timestamp even when multiple nodes report the same tunnel bucket.
+- [x] Parallel `/flow/upload` calls for the same tunnel/node/minute do not create duplicate bucket rows.
+- [x] `/monitor` clearly distinguishes live state from historical persisted charts and surfaces data freshness to the operator.
+- [x] Service monitor list shows enabled state, latest health result, latest check time, and stale-state semantics correctly.
+- [x] Deleting a service monitor no longer leaves dangling historical data in list-facing APIs.
+- [x] Monitoring ingestion/result write failures are visible in logs and no longer fail silently.
+- [x] Non-admin users without monitoring permission do not get a confusing monitor-entry experience, while granted users continue to access monitoring successfully.
+- [x] Backend monitoring contract tests pass.
+- [x] New repository/scheduler tests pass.
+- [x] `cd go-backend && go test ./... -count=1` passes.
+- [x] `cd go-gost/x && go test ./socket/... -count=1` passes.
+- [x] `cd vite-frontend && npm run build` passes.

plans/045-release-2-1-8.md

@@ -0,0 +1,10 @@
+# PLAN: Release 2.1.8
+
+## Overview
+Synchronize with the main branch and release a new tag `2.1.8`.
+
+## Tasks
+- [x] Pull latest changes from `main`
+- [x] Update `AGENTS.md` with new tag and current commit hash (PR #337 created and set to auto-merge)
+- [x] Create git tag `2.1.8`
+- [x] Push git tag `2.1.8` to origin

plans/046-monitor-nezha-redesign.md

@@ -0,0 +1,23 @@
+# Monitor Nezha Redesign
+
+## Summary
+Redesign the monitor overview to resemble Nezha dashboard. Phase 1 introduced ServerCard grid. Phase 2 hides node metrics from the overview and adds a detail view with charts and service monitor latency sparklines.
+
+## Objectives
+- [x] Analyze `monitor-view.tsx` layout boundaries.
+- [x] Create `ServerCard` utilizing `Progress` component.
+- [x] Render metrics (CPU, RAM, Disk, System Load, Connections, Network speeds).
+- [x] Implement robust real-time updates and seamless state linkage via existing hooks.
+- [x] Adjust layout positioning for impact at top-of-page.
+- [x] Hide node metrics from overview — click node card to enter detail view.
+- [x] Detail view: back button + node header + realtime KPI cards.
+- [x] Detail view: node metrics chart (CPU/Memory/Disk/Network/Load/Connections).
+- [x] Detail view: tunnel traffic chart.
+- [x] Detail view: service monitors rendered as latency sparkline charts (Nezha-style).
+- [x] Service monitor cards include status dot, type chip, target, interval, and dropdown actions.
+
+## Technical Details
+- Phase 1: Injected a `ServerCard` inline component with progress bars and real-time metrics.
+- Phase 2: Added `detailNodeId` state for drill-down navigation. Grid view shows only summary bar + clickable server cards. Detail view shows KPI summary cards, node metrics chart (reusing existing recharts setup), tunnel traffic chart, and service monitors as a responsive grid of cards each containing a latency-over-time sparkline chart. Reduced file from 2108 to 1799 lines by consolidating the old overview card into the summary bar.
+
+## Status: Complete

plans/046-node-card-cleanup-monitor-redesign.md

@@ -0,0 +1,7 @@
+# 046 - Node Card Cleanup & Monitor Redesign
+
+## Tasks
+
+- [x] 1. Remove duplicate system metrics from node cards in `node.tsx` (CPU, memory, upload/download speed, upload/download traffic, disk, load)
+- [x] 2. Redesign monitor `ServerCard` in `monitor-view.tsx` to match the node card style from `node.tsx`
+- [x] 3. Make all charts in monitor view update incrementally (streaming) instead of full page reload

plans/047-beautify-monitor-tab.md

@@ -0,0 +1,9 @@
+# 047 - Beautify Monitor Tab
+
+## Goal
+Redesign the node monitoring cards in the Monitor tab (监控标签) to make them more visually appealing. Incorporate a dynamic, clean layout mimicking the Nezha monitoring dashboard.
+
+## Tasks
+- [x] Redesign `ServerCard` in `vite-frontend/src/pages/node/monitor-view.tsx` with gradients, neon statuses, clean layout, up/down arrows for speeds.
+- [x] Update the realtime KPI cards in the Monitor detailed view to have subtle gradients and dynamically colored text.
+- [x] Fix TypeScript linting issues with `Progress` component colors (`classNames` vs `color`).

plans/048-release-2-1-9-alpha5.md

@@ -0,0 +1,11 @@
+# 048 - Release 2.1.9-alpha5
+
+## Goal
+Release the current state of `main` (including the monitor tab beautification and user page improvements) under the tag `2.1.9-alpha5`.
+
+## Tasks
+- [x] Update `AGENTS.md` with current date, branch, and anticipated tag.
+- [x] Create a release branch `release/2.1.9-alpha5`.
+- [x] Commit and push the branch.
+- [x] Create a Pull Request and merge it into `main`.
+- [x] Publish the new tag `2.1.9-alpha5`.

plans/049-monitor-list-view.md

@@ -0,0 +1,32 @@
+# 049 - Monitor List View
+
+## Objective
+The user requested that the Monitor page should switch its card-based grid view to a list view similar to a provided screenshot, and a view mode toggle should be added to the top right.
+
+## Expected Features
+1. View Mode Toggle
+   - Add a state `viewMode` in `pages/monitor.tsx`.
+   - Add a toggle button with LayoutGrid/List icons next to the refresh button.
+   - Pass `viewMode` down to `MonitorView` component.
+2. List View Implementation
+   - Extend `MonitorViewProps` with `viewMode: "list" | "grid"`.
+   - Render the `ServerCard` grid when `viewMode === "grid"`.
+   - Render a `Table` when `viewMode === "list"`.
+   - The list view should include:
+     - 状态 (Status: Colored dot depending on `isOnline`).
+     - 名称 (Name: Node name).
+     - 速率 (Speed: Up/Down speeds styled appropriately).
+     - 流量 (Traffic: Total Up/Down bytes).
+     - 开机时长 (Uptime).
+     - 连接数 (Connections: TCP/UDP).
+     - CPU (Progress bar).
+     - RAM (Progress bar).
+     - 存储 (Storage / Disk Progress bar).
+     - 操作 (Actions: Eye view icon to open detailed monitor).
+
+## Checklist
+- [x] Create plan document.
+- [x] Add viewMode state and toggle in `monitor.tsx`.
+- [x] Receive viewMode in `MonitorView` and selectively render Grid vs List views.
+- [x] Ensure list correctly visualizes node info, speed, traffic, uptime, conns, CPU/RAM/Disk usages, and action icons.
+- [x] Fix HeroUI missing TableProps typings (remove `removeWrapper` prop).

plans/050-monitor-tunnel-top-level.md

@@ -0,0 +1,40 @@
+# 050 - 监控页面隧道独立入口
+
+## 背景
+当前监控页面的隧道流量监控嵌套在节点详情内，用户需要先点击某个节点才能看到隧道流量。
+需要将隧道监控提升为与节点列表同级的入口，展示隧道质量（延迟）和流量统计。
+
+## 设计
+
+### 监控页面结构变更
+在 `monitor.tsx` 页面增加 Tab 式结构（或并列区域），新增"隧道监控"入口与节点列表同级：
+- **节点** Tab：保持现有节点卡片/列表 + 详情视图
+- **隧道** Tab：新增隧道卡片/列表视图
+  - 每个隧道卡片显示：隧道名称、状态、质量指标（入口→出口延迟、出口→Bing TCP延迟）
+  - 点击隧道进入详情：隧道流量趋势图表（已有的 tunnel metrics chart）
+
+### 隧道质量指标
+使用已有的隧道诊断 API (`/tunnel/diagnose`) 进行 TCP 延迟测试：
+- 入口→出口延迟
+- 出口→Bing TCP 延迟（bing.com:443）
+
+新增一个轻量级的后端 API 来获取隧道质量快照（定时 TCP 测试延迟），
+或者前端在隧道 Tab 里手动触发诊断。
+
+### 实现方案（前端为主）
+由于后端已有完整的 tunnel metrics 和 diagnosis API，本次改动主要在前端：
+
+1. `monitor.tsx` 增加 Tab 切换（节点 / 隧道）
+2. 新建 `tunnel-monitor-view.tsx` 组件
+3. 隧道列表用卡片展示质量指标
+4. 点击进入隧道详情，展示流量图表
+5. 从 `monitor-view.tsx` 的节点详情中移除隧道流量图表
+
+## 任务清单
+
+- [x] 1. 修改 `monitor.tsx`，增加节点/隧道 Tab 切换
+- [x] 2. 新建 `tunnel-monitor-view.tsx` 组件，实现隧道列表视图
+- [x] 3. 隧道卡片展示名称、状态、质量指标（支持手动诊断获取延迟）
+- [x] 4. 隧道详情视图，展示流量趋势图表（复用已有 tunnel metrics API）
+- [x] 5. 从节点详情视图中移除隧道流量图表
+- [x] 6. TypeScript 编译通过验证

plans/051-sync-main-and-push-tag.md

@@ -0,0 +1,15 @@
+# Plan 051: Sync Main and Push Tag
+
+## Goal
+Submit all changes, push them to the remote repository, create a pull request to main, merge it, and push a new tag.
+
+## Tasks
+- [ ] Determine the next tag version.
+- [ ] Add all changes to the staging area.
+- [ ] Commit the changes.
+- [ ] Push changes to the current branch.
+- [ ] Create a pull request via `gh` CLI.
+- [ ] Merge the pull request via `gh` CLI.
+- [ ] Sync the `main` branch locally.
+- [ ] Create a new tag.
+- [ ] Push the new tag to the remote repository.

plans/052-node-user-list-view.md

@@ -0,0 +1,10 @@
+# 052 Node and User List View
+
+## Objective
+Add a list view and a card view toggle for Node Page and User Page, so they match exactly how the Monitor Page does it.
+
+## Tasks
+- [x] Add `viewMode` state toggle with `LayoutGrid` and `List` icons to `pages/node.tsx` header area.
+- [x] Implement Table view rendering for Node list in `pages/node.tsx`.
+- [x] Add `viewMode` state toggle with `LayoutGrid` and `List` icons to `pages/user.tsx` header area.
+- [x] Implement Table view rendering for User list in `pages/user.tsx`.

plans/053-list-view-styling.md

@@ -0,0 +1,16 @@
+# 053 List View Styling
+
+## Objective
+Update the design of Node and User list table views to match the provided screenshot:
+- Use colored dots for status instead of separate "Status" text columns.
+- The action buttons should have text labels (e.g. "安装", "编辑") with `variant="flat"` instead of icons.
+- Add version column for Node list.
+- Remove traffic columns from Node list as requested.
+- Adjust User list to match this clean style.
+
+## Tasks
+- [x] Update `pages/node.tsx` list view to use the new column layout (Node name with dot, Address, Version, Actions).
+- [x] Update action buttons in `pages/node.tsx` list view to use text instead of icons.
+- [x] Update `pages/user.tsx` list view to use the status dot pattern.
+- [x] Update action buttons in `pages/user.tsx` list view to use text instead of icons.
+- [x] Ensure `selectionMode="multiple"` (or similar) is properly reflected.

plans/054-agent-panel-comm-optimization.md

@@ -0,0 +1,147 @@
+# Agent-Panel 通信优化：提升稳定性与效率
+
+## 背景
+
+Agent（`go-gost/x/socket/websocket_reporter.go`）与 Panel（`go-backend/internal/ws/server.go`）之间通过 WebSocket 进行实时通信，包括指标上报（每 5s）、命令下发/响应、和流量上报（HTTP）。经过代码审查，以下是发现的问题和优化建议。
+
+---
+
+## 发现的问题
+
+### 1. Keepalive 时序不匹配 —— 导致误断连
+
+| 参数 | Agent 侧 | Panel 侧 |
+|------|----------|----------|
+| Read deadline | `reporterReadWait` = 60s | `wsPongWait` = 45s |
+| Ping 发送间隔 | 无主动 ping（靠指标数据 5s 续命） | `wsPingPeriod` = 15s |
+| Write timeout | `reporterWriteWait` = 5s | `wsWriteWait` = 5s |
+
+**问题**：Panel 每 15s 发 ping，Agent read deadline 60s，但 Panel pong deadline 只有 45s。如果 Agent 的指标消息被延迟（网络抖动），Panel 可能因 pong 超时而关闭连接。两侧的超时参数缺乏协调设计。
+
+### 2. 固定重连间隔 —— 无退避策略
+
+Agent 断线后以固定 5s 间隔重试（`reconnectTime = 5 * time.Second`），在 Panel 长时间不可用（升级、网络故障）的情况下，会产生大量无用连接尝试。
+
+### 3. Panel 侧每次解密都重建 AES 加密器
+
+`ws/server.go` 的 `decryptIfNeeded()` 和 `SendCommand()` 每次调用都 `security.NewAESCrypto(secret)` 重新创建 cipher（SHA256 + AES-GCM 初始化），对于高频指标消息（5s/次 × N 节点），有不必要的 CPU 开销。
+
+### 4. 指标消息使用 JSON Text 格式传输
+
+每 5s 发送一次包含 13 个字段的 SystemInfo JSON，加密后还需 base64 编码，一条消息约 300-500 bytes（加密后约 700 bytes）。对于大量节点场景，存在优化空间。
+
+### 5. `receiveMessages` 紧循环中有频繁锁竞争
+
+`receiveMessages()` 在每次 `ReadMessage()` 前都要 `Lock/Unlock connMutex` 检查连接状态，但 `ReadMessage` 本身是阻塞的，实际不需要在循环外检查。
+
+### 6. 状态变更命令阻塞读消息循环
+
+`routeCommand` 中的 Service/Chain/Limiter CRUD 命令是同步执行的，包括 `saveConfig()` 文件写入。执行期间会阻塞 `receiveMessages` 的读取循环。
+
+---
+
+## 推荐的优化方案（按优先级排列）
+
+### P0 — 高收益、低风险
+
+#### 优化 1：协调 Keepalive 参数
+
+**文件**：`websocket_reporter.go`
+
+- Agent 增加独立的 WebSocket ping 发送（每 20s），不依赖指标数据来维持连接
+- 统一 read deadline 设置，确保两侧 read timeout > 2×ping interval
+
+#### 优化 2：指数退避重连
+
+**文件**：`websocket_reporter.go`
+
+- 初始间隔 2s，按指数退避增长至最大 2 分钟
+- 连接成功后立即重置退避
+- 增加随机抖动（jitter）避免大量 Agent 同时重连
+
+#### 优化 3：Panel 侧缓存 AES 加密器
+
+**文件**：`ws/server.go`
+
+- 将 `AESCrypto` 实例缓存在 `nodeSession` 中，避免每条消息重建
+- `SendCommand` 复用缓存实例
+
+### P1 — 中等收益
+
+#### 优化 4：减少 `receiveMessages` 锁竞争
+
+**文件**：`websocket_reporter.go`
+
+- 将连接状态检查移到循环外，只在出错/关闭时通过 channel 通知退出
+- 用 `context.WithCancel` 代替锁检查 `connected` flag 来控制生命周期
+
+#### 优化 5：异步化状态变更命令处理
+
+**文件**：`websocket_reporter.go`
+
+- 所有命令统一异步执行（通过 goroutine + response channel），避免阻塞 readLoop
+- 当前只有 TcpPing/ServiceMonitorCheck/UpgradeAgent/RollbackAgent 是异步的
+
+---
+
+## 具体代码变更
+
+### Agent 侧 (`go-gost/x/socket`)
+
+---
+
+#### [MODIFY] [websocket_reporter.go](file:///Users/sagit/Documents/github/flvx/go-gost/x/socket/websocket_reporter.go)
+
+1. **指数退避重连**：将 `reconnectTime` 从固定 `5s` 改为动态退避字段，增加 `curBackoff/maxBackoff` 字段
+2. **独立 Ping 发送**：在 `handleConnection()` 中增加 WebSocket ping ticker（20s），独立于指标上报
+3. **减少锁竞争**：`receiveMessages` 中只在循环入口检查一次连接，此后靠 `ReadMessage` 的 error 退出
+4. **统一命令异步化**：所有 `routeCommand` 调用统一使用 goroutine
+
+---
+
+### Panel 侧 (`go-backend/internal/ws`)
+
+---
+
+#### [MODIFY] [server.go](file:///Users/sagit/Documents/github/flvx/go-backend/internal/ws/server.go)
+
+1. **缓存 AES 加密器**：在 `nodeSession` 中增加 `crypto *security.AESCrypto` 字段，节点连接时初始化
+2. **`decryptIfNeeded` 接收 crypto 参数**而非 secret 字符串
+3. **`SendCommand` 使用缓存 crypto** 实例
+
+---
+
+## Verification Plan
+
+### Automated Tests
+
+```bash
+# 运行现有 agent 侧单元测试（验证不回归）
+(cd go-gost/x && go test ./socket/... -v -count=1)
+
+# 运行现有流量上报测试
+(cd go-gost/x && go test ./service/... -v -count=1)
+
+# 运行 panel 侧全部测试
+(cd go-backend && go test ./... -count=1)
+```
+
+### Manual Verification
+
+> [!IMPORTANT]
+> 本次改动涉及实时通信核心路径，建议在 staging 环境部署后观察至少 30 分钟：
+> 1. 检查节点在面板中状态是否正常显示为在线
+> 2. 手动停止面板后观察 Agent 日志，确认重连间隔呈指数增长
+> 3. 恢复面板后确认 Agent 能自动恢复连接并恢复指标上报
+> 4. 通过面板下发命令（如添加/删除 service），确认命令执行成功
+
+---
+
+## 任务清单
+
+- [x] 优化 1：Agent 增加独立 WebSocket ping 发送
+- [x] 优化 2：Agent 指数退避重连
+- [x] 优化 3：Panel 缓存 AES 加密器
+- [x] 优化 4：Agent 减少 receiveMessages 锁竞争
+- [x] 优化 5：Agent 命令处理统一异步化
+- [x] 运行现有测试验证不回归

plans/055-tunnel-quality-periodic-probing.md

@@ -0,0 +1,31 @@
+# 055 - 隧道质量定时探测 + 实时展示 + 历史图表
+
+## 背景
+当前隧道质量检测是手动触发的：用户点击"诊断"按钮 → 后端调用节点 TcpPing → 返回结果。
+需求：改为**后端定时（每10秒）自动探测**所有启用隧道的质量（入口→出口延迟、出口→Bing延迟），
+结果保留历史（24h），前端隧道 Tab 实时展示 + 图表历史趋势。
+
+## 设计原则：与服务监控复用
+
+| 复用点 | 服务监控 | 隧道质量 |
+|--------|---------|---------|
+| 调度方式 | `health.Checker.Start(ctx)` via `jobs.go` | `tunnelQualityProber.Start(ctx)` via `jobs.go` |
+| 存储模式 | `service_monitor_result` (history, insert) | `tunnel_quality` (history, insert) |
+| 清理方式 | `PruneServiceMonitorResults(olderThanMs)` | `PruneTunnelQualityResults(olderThanMs)` |
+| 最新查询 | `GetLatestServiceMonitorResults()` (window func) | `GetLatestTunnelQualities()` (window func) |
+| 历史查询 | `GetServiceMonitorResults(id, limit)` | `GetTunnelQualityHistory(id, start, end)` |
+| API 模式 | `GET /monitor/services/{id}/results` | `GET /monitor/tunnels/{id}/quality` |
+| 前端图表 | Recharts LineChart (延迟趋势) | Recharts LineChart (同样模式) |
+
+## 任务清单
+
+- [x] 1. `TunnelQuality` model 改为历史存储（composite index, 非 unique）
+- [x] 2. Repo 改为 insert（非 upsert），复用服务监控的查询模式
+- [x] 3. 添加 `PruneTunnelQualityResults` + `GetLatestTunnelQualities` + `GetTunnelQualityHistory`
+- [x] 4. Prober 生命周期集成到 `jobs.go`（与 healthCheck 同级）
+- [x] 5. Prober 添加 24h 清理周期
+- [x] 6. 添加 API `GET /monitor/tunnels/{id}/quality` 返回历史
+- [x] 7. 前端添加 `getMonitorTunnelQualityHistory()` API
+- [x] 8. 前端详情页添加质量趋势图表（复用服务监控图表组件模式）
+- [x] 9. Go 编译 + 测试通过
+- [x] 10. TypeScript 编译通过

plans/056-tunnel-uptime-kuma-bars.md

@@ -0,0 +1,16 @@
+# Tunnel Quality Uptime Kuma Display
+**Plan ID**: 056-tunnel-uptime-kuma-bars.md
+
+## Objective
+The goal is to modify the "Monitor - Tunnel" UI page to remove the isolated "Quality" column/chip, and replace the specific latency readouts for entry->exit and exit->Bing with an Uptime Kuma style row of visual history bars.
+
+## Steps
+- [x] Remove the individual "Quality" column from the tunnel monitor table view.
+- [x] Remove the individual "Quality" chip from the tunnel monitor grid view.
+- [x] Implement `<UptimeHistoryBar />` to display a historical sequence of up to 30 metrics, coloring by latency (success, warning, danger) and packet loss (danger).
+- [x] Modify the front-end to poll/load initial tunnel quality history, effectively padding the history bars instead of them starting empty.
+- [x] Update `getMonitorTunnelQuality` auto-polling locally to append strictly to the local history state, truncating appropriately.
+- [x] Retain current latency display next to or below the Uptime bars to allow numerical visibility.
+
+## Status
+Completed. The interface will now load history for all displayed tunnels upon open, and continue tracking with bars appending in real time every 10-second polling interval.

plans/057-hide-remote-nodes-monitor.md

@@ -0,0 +1,13 @@
+# 057 Hide Remote Nodes from Monitor View
+
+## Goal
+Do not display remote nodes on the Monitoring page.
+
+## Changes Made
+- Modified `ListMonitorNodes` in `go-backend/internal/store/repo/repository_monitor_nodes.go` by adding a `.Where("is_remote = ?", 0)` constraint so that remote nodes (imported via Federation feature) are entirely excluded from the returned payload for API `/api/v1/monitor/nodes`.
+- This efficiently removes remote nodes from both the grid/list displaying Node stats in the Monitoring tab and also eliminates remote nodes from the selection dropdown when creating new Service Monitors.
+
+## Checklist
+- [x] Identify how "remote node" is defined in the database structure (`IsRemote` = 1 or 0).
+- [x] Add SQL query constraint to filter out remote nodes from the `/monitor/nodes` API response.
+- [x] Verify changes compile successfully.

plans/058-sync-main-and-push-tag.md

@@ -0,0 +1,13 @@
+# Plan 058: Sync Main and Push Tag
+
+Sync current changes to `main`, merge to `master` (if needed) or just PR/merge on `main` (if that's the flow), and push a new tag `2.1.9-beta7`.
+
+## Tasks
+
+- [x] Determine the next tag version (`2.1.9-beta7`).
+- [x] Create a feature branch `release/2.1.9-beta7`.
+- [x] Commit all current changes to the feature branch.
+- [x] Push the feature branch to remote.
+- [x] Create a PR to `main`.
+- [x] Merge the PR to `main`.
+- [x] Create and push tag `2.1.9-beta7`.

plans/059-theme-system-design.md

@@ -0,0 +1,59 @@
+# 059 - 主题系统设计（v2 — 完整可扩展架构）
+
+## 概述
+
+设计一个高度可扩展的主题包架构，允许第三方作者通过代码提交的方式创建主题，覆盖前端所有元素——从 CSS 变量到组件实现、布局结构、甚至整个页面。
+
+## 架构
+
+```
+src/themes/
+├── types.ts                  # ThemePackage 接口定义
+├── registry.ts               # 主题注册表 + CSS 注入引擎
+├── context.tsx                # React Context + Provider + Hooks
+├── index.ts                  # 公共 API barrel
+├── loader.ts                 # 主题加载器（注册所有内置主题）
+├── README.md                 # 主题开发指南
+│
+├── default/                  # 默认主题（参考实现）
+│   └── index.ts
+│
+├── example-cyberpunk/        # 示例主题（赛博朋克）
+│   ├── index.ts
+│   └── components/
+│       └── button.tsx        # 组件覆盖示范
+│
+└── <your-theme>/             # 第三方主题
+    ├── index.ts
+    ├── components/
+    ├── layouts/
+    ├── pages/
+    └── assets/
+```
+
+## 覆盖层级
+
+| 层级 | 字段 | 说明 |
+|------|------|------|
+| CSS 变量 | `tokens.light` / `tokens.dark` | 80+ 个设计 token（颜色、字体、圆角） |
+| 原始 CSS | `css` | 注入自定义 CSS（动画、字体、阴影等） |
+| 组件替换 | `components` | 替换任意 UI 组件（30+ 个可替换组件键） |
+| 布局替换 | `layouts` | 替换 4 种布局（Admin / H5 / H5Simple / Default） |
+| 页面替换 | `pages` | 替换 14 个页面路由实现 |
+| 生命周期 | `onActivate` / `onDeactivate` | 主题启用/停用回调 |
+
+## 任务清单
+
+- [x] **T1**: 创建 `src/themes/types.ts` — ThemePackage 接口 + 所有可覆盖键定义
+- [x] **T2**: 创建 `src/themes/registry.ts` — 主题注册/激活/停用/CSS 注入引擎
+- [x] **T3**: 创建 `src/themes/context.tsx` — React Context + ThemeProvider + hooks
+- [x] **T4**: 创建 `src/themes/index.ts` — 公共 API barrel
+- [x] **T5**: 创建 `src/themes/loader.ts` — 自动加载所有内置主题
+- [x] **T6**: 创建 `src/themes/default/` — 默认主题参考实现
+- [x] **T7**: 创建 `src/themes/example-cyberpunk/` — 完整示例主题（含组件覆盖 + CSS + 生命周期）
+- [x] **T8**: 重构 `use-theme.tsx` — 向后兼容包装
+- [x] **T9**: 重构 `theme-provider.tsx` — 集成新主题系统
+- [x] **T10**: 编写 `README.md` — 主题开发完整指南
+- [x] **T11**: TypeScript 编译验证通过
+- [ ] **T12**: (后续) 设置页面集成主题选择器 UI
+- [ ] **T13**: (后续) 将现有组件导入逐步迁移到 `useThemedComponent` 模式

plans/060-tunnel-quality-polling.md

@@ -0,0 +1,42 @@
+# 060 Nezha-style Monitoring (1s test, 30s report)
+
+## Objective
+Update all monitoring subsystems to test every 1 second and report (write to DB) every 30 seconds, matching Nezha-style monitoring behavior.
+
+## Changes
+
+### 1. Tunnel Quality Prober (`go-backend/internal/http/handler/tunnel_quality_prober.go`)
+- [x] Change `tunnelQualityProbeInterval` from 10s to 1s
+- [x] Add `tunnelQualityReportInterval = 30s` for DB write throttling
+- [x] Update `storeResult` to cache in-memory every tick, write to DB only every 30s per tunnel
+- [x] Add atomic `probing` flag to prevent overlapping `probeAll()` goroutine pile-up
+- [x] Increase `maxWorkers` from 4 to 20
+
+### 2. Service Monitor Checker (`go-backend/internal/health/checker.go`)
+- [x] Add `serviceMonitorReportInterval = 30s` for DB write throttling
+- [x] Add `latestResults` in-memory map and `lastDBWrite` map per monitor
+- [x] Add `GetLatestCached()` method for real-time API reads
+- [x] Add atomic `checking` flag to prevent overlapping `runChecks()` goroutine pile-up
+- [x] Modify worker goroutines to always update in-memory cache, only write to DB every 30s
+
+### 3. Service Monitor Limits (`go-backend/internal/monitoring/limits.go`)
+- [x] Change `CheckerScanIntervalSec` default from 30 to 1
+- [x] Change `WorkerLimit` default from 5 to 20
+- [x] Change `MinIntervalSec` default from 30 to 1
+- [x] Change `DefaultIntervalSec` default from 60 to 1
+
+### 4. Monitoring API Handler (`go-backend/internal/http/handler/monitoring.go`)
+- [x] Update `monitorServiceLatestResultsHandler` to prefer in-memory cached results from `healthCheck.GetLatestCached()`
+
+### 5. Agent WebSocket Reporter (`go-gost/x/socket/websocket_reporter.go`)
+- [x] Change `pingInterval` (metric reporting) from 5s to 1s
+
+### 6. Frontend - Tunnel Monitor (`vite-frontend/src/pages/node/tunnel-monitor-view.tsx`)
+- [x] Change `QUALITY_POLL_INTERVAL` from 10s to 1s 
+- [x] Update detail view text: "自动探测中（每秒测试，30秒上报）"
+- [x] Update list view text: "每秒探测 · 更新于 ..."
+
+### 7. Frontend - Service Monitor (`vite-frontend/src/pages/node/monitor-view.tsx`)
+- [x] Change `DEFAULT_SERVICE_MONITOR_LIMITS` defaults to match backend (1s intervals)
+- [x] Change service monitor + latest results polling from 30s to 1s
+- [x] Update info bar text: "每秒测试，30秒上报"

plans/061-node-os-logo.md

@@ -0,0 +1,15 @@
+# Plan 061: Node Logo By OS Type (Linux Distro)
+
+## Objective
+Display different logos for nodes in the 'Monitor - Node - Card/List View' based on their Linux distribution (Ubuntu, Debian, CentOS, Alpine, etc.).
+
+## Tasks
+- [x] Agent: Use `gopsutil/v3/host.Info().Platform` to detect the Linux distro and include it in the version string (`distro.go`).
+- [x] Agent: Update `main.go` to call `socket.DetectDistro()` instead of `runtime.GOOS`.
+- [x] Agent: Ensure version parameter is URL-escaped since it now includes distro info with special chars.
+- [x] Backend: Select `version` column in `ListMonitorNodes`.
+- [x] Backend: Include `version` in `monitorNodeListItem` JSON response.
+- [x] Frontend: Add `version` to TypeScript interfaces (`MonitorNodeApiItem`, `MonitorNode`, `MonitorViewProps`).
+- [x] Frontend: Create `distro-icon.tsx` component with SVG logos for Ubuntu, Debian, CentOS/Rocky/Alma, Alpine, Fedora, Arch/Manjaro, and a default Linux (Tux) fallback.
+- [x] Frontend: Use `DistroIcon` in `ServerCard` (card view) and list view name column with branded colors per distro.
+- [x] All three projects compile cleanly (`go build`, `tsc --noEmit`).

plans/062-release-beta14.md

@@ -0,0 +1,49 @@
+# Plan 062: Commit, PR, Merge and Tag
+
+## Overview
+This plan outlines the steps to commit all changes, create a PR, merge it, and then publish a new tag for the FLVX project.
+
+## Checklist
+- [ ] Check current git status for any unexpected changes
+- [ ] Create a feature branch `feat-node-os-logo-release`
+- [ ] Stage and commit all modifications and untracked files
+- [ ] Push the feature branch to origin
+- [ ] Create a Pull Request (PR) from the feature branch to `main`
+- [ ] Merge the PR to `main`
+- [ ] Update `AGENTS.md` with the new tag and commit hash
+- [ ] Create and push new tag `2.1.9-beta14`
+
+## Detailed Steps
+
+### 1. Create Feature Branch
+```bash
+git checkout -b feat-node-os-logo-release
+```
+
+### 2. Commit all changes
+Add all modified and untracked files:
+```bash
+git add .
+git commit -m "feat: node OS logo support, UI rate overlap fix and tunnel monitoring updates"
+```
+
+### 3. Push and PR
+Push to `origin`:
+```bash
+git push origin feat-node-os-logo-release
+```
+Create PR via `gh pr create` if possible.
+
+### 4. Merge to Main
+```bash
+git checkout main
+git merge feat-node-os-logo-release
+git push origin main
+```
+
+### 5. Create Tag
+Increment the current tag `2.1.9-beta13` to `2.1.9-beta14`.
+```bash
+git tag 2.1.9-beta14
+git push origin 2.1.9-beta14
+```

plans/063-release-2-1-9-rc6.md

@@ -0,0 +1,12 @@
+# Plan 063: Release 2.1.9-rc6
+
+Sync all changes, bump version to `2.1.9-rc6`, create PR, merge, and publish tag.
+
+## Tasks
+
+- [x] Update `AGENTS.md` with new tag (`2.1.9-rc6`) and today's date (`Tue Mar 24 2026`).
+- [x] Commit all changes to branch `chore/rc6-bump`.
+- [x] Push branch to remote.
+- [x] Create Pull Request using `gh`.
+- [x] Merge Pull Request using `gh`.
+- [x] Create and push tag `2.1.9-rc6`.

plans/064-release-2-1-9-rc7.md

@@ -0,0 +1,12 @@
+# Plan 064: Release 2.1.9-rc7
+
+Sync all changes, bump version to `2.1.9-rc7`, create PR, merge, and publish tag.
+
+## Tasks
+
+- [x] Update `AGENTS.md` with new tag (`2.1.9-rc7`) and today's date (`Tue Mar 24 2026`).
+- [x] Commit all changes to branch `chore/rc7-bump`.
+- [x] Push branch to remote.
+- [x] Create Pull Request using `gh`.
+- [x] Merge Pull Request using `gh`.
+- [x] Create and push tag `2.1.9-rc7`.

plans/065-fix-node-metrics-postgres-type.md

@@ -0,0 +1,11 @@
+# Fix Node Metrics PostgreSQL Type Encoding
+
+## Objective
+Fix the PostgreSQL type encoding error (`failed to encode args[0]: unable to encode 5 into text format for text (OID 25)`) and `integer out of range` error when querying node metrics for time ranges greater than 1 hour.
+
+## Tasks
+- [x] Identify the problematic downsampled SQL aggregation in `GetNodeMetrics`.
+- [x] Fix the `? AS node_id` placeholder which confused PostgreSQL's type inference by directly embedding the `nodeID` using `fmt.Sprintf("%d AS node_id")`.
+- [x] Change all `CAST(X AS INTEGER)` to `CAST(X AS BIGINT)` to prevent 32-bit integer overflow on Unix millisecond timestamps in PostgreSQL.
+- [x] Verify the build and tests pass.
+- [ ] Commit all changes, create a new branch, push, create a Pull Request, merge the PR into `main`, and publish a new tag `2.1.9-rc9`.

plans/066-issue-349-ipv6-entry-format.md

@@ -0,0 +1,11 @@
+# 066 - Issue 349 IPv6 Entry Format
+
+## Goal
+- 修复规则入口 IPv6 地址在列表/复制场景下缺少方括号的问题，确保 API 返回与前端展示都能直接使用 `[IPv6]:port` 格式。
+
+## Checklist
+- [x] 定位规则入口地址的生成链路，确认问题来自后端入口地址拼接格式。
+- [x] 修复 IPv6 入口地址拼接逻辑，统一输出可直接复制的标准格式。
+- [x] 增加回归验证，覆盖 `/api/v1/forward/list` 的 IPv6 入口地址格式。
+- [x] 运行相关测试并确认通过。
+- [ ] 提交修复分支、创建 PR，并合并到 `main`。