fix(monitor): harden server delete and remove challenge docs

- Delete server state before monitored server to avoid FK 500s
- Keep legacy cleanup for obsolete challenge tables
- Rewrite monitor docs to API key-only flow
This commit is contained in:
zhi
2026-03-20 08:02:19 +00:00
parent 8e0f158266
commit 9b5e2dc15c
3 changed files with 136 additions and 534 deletions

View File

@@ -1,68 +1,76 @@
# OpenClaw Monitor Agent Plugin 开发计划(草案)
# HarborForge Monitor / OpenClaw Plugin Connector Plan
## 目标
让被监测服务器通过 WebSocket 主动接入 HarborForge Backend并持续上报
- OpenClaw 版本
- agent 列表
- 每 5 分钟主机指标CPU/MEM/DISK/SWAP
- agent 状态变更事件
## 握手流程
1. Admin 在 HarborForge 后台添加 server identifier
2. Admin 生成 challenge10 分钟有效)
3. 插件请求 `GET /monitor/public/server-public-key` 获取公钥
4. 插件构造 payload
- `identifier`
- `challenge_uuid`
- `nonce`(随机)
- `ts`ISO8601
5. 使用 RSA-OAEP(SHA256) 公钥加密base64 后作为 `encrypted_payload` 发给 `WS /monitor/server/ws`
6. 握手成功后进入事件上报通道
使用 **API Key + HTTP heartbeat** 连接 HarborForge Monitor 与远程 OpenClaw 节点。
## 插件事件协议
### server.hello
## 认证方式
- 管理员为服务器生成 API Key
- 插件通过 `X-API-Key` 调用 heartbeat 接口
- 不再使用 challenge / RSA 公钥 / WebSocket 握手
## 上报接口
`POST /monitor/server/heartbeat-v2`
### Headers
- `X-API-Key: <server-api-key>`
### Payload
```json
{
"event": "server.hello",
"payload": {
"openclaw_version": "x.y.z",
"agents": [{"id": "a1", "name": "agent-1", "status": "idle"}]
"identifier": "vps.t1",
"openclaw_version": "OpenClaw 2026.3.13 (61d171a)",
"plugin_version": "0.1.0",
"agents": [
{ "id": "agent-bot1", "name": "agent-bot1", "status": "configured" }
],
"cpu_pct": 12.3,
"mem_pct": 45.6,
"disk_pct": 78.9,
"swap_pct": 0,
"load_avg": [0.12, 0.08, 0.03],
"uptime_seconds": 12345
}
```
## 语义
- `openclaw_version`: 远程主机上的 OpenClaw 版本
- `plugin_version`: harborforge-monitor 插件版本
## 插件生命周期
- 插件注册到 Gateway
-`gateway_start` 启动 `server/telemetry.mjs`
-`gateway_stop` 停止 sidecar
## 配置位置
`~/.openclaw/openclaw.json`
```json
{
"plugins": {
"entries": {
"harborforge-monitor": {
"enabled": true,
"config": {
"enabled": true,
"backendUrl": "http://127.0.0.1:8000",
"identifier": "vps.t1",
"apiKey": "your-api-key"
}
}
}
}
}
```
### server.metrics每 5 分钟)
```json
{
"event": "server.metrics",
"payload": {
"cpu_pct": 21.3,
"mem_pct": 42.1,
"disk_pct": 55.9,
"swap_pct": 0.0,
"agents": [{"id": "a1", "name": "agent-1", "status": "busy"}]
}
}
```
## 已废弃
### agent.status_changed可选
```json
{
"event": "agent.status_changed",
"payload": {
"agents": [{"id": "a1", "name": "agent-1", "status": "focus"}]
}
}
```
## 实施里程碑
- M1: Node/Python CLI 插件最小握手联通
- M2: 指标采集 + 周期上报
- M3: agent 状态采集与变更事件
- M4: 守护化systemd+ 断线重连 + 本地日志
## 风险与注意事项
- 时钟漂移会导致 `ts` 校验失败(建议 NTP
- challenge 仅一次可用,重复使用会被拒绝
- nonce 重放会被拒绝
- 需要保证插件本地安全保存 identifier/challenge短期
- challenge UUID
- server public key
- WebSocket telemetry
- encrypted handshake payload