👁 终端守望者+预览部署+傻瓜式初始化脚本·铸渊的眼睛和三级部署流程

This commit is contained in:
铸渊 (ICE-GL-ZY001) 2026-05-12 07:48:58 +00:00
parent 27936b4b1a
commit bd2d4cb039
5 changed files with 1239 additions and 0 deletions

View File

@ -0,0 +1,133 @@
# ════════════════════════════════════════════════════════════════
# 光湖 · 自动部署到预览环境
# Trigger: push to main
# 部署到测试+预览环境正式环境需冰朔在Gitea点Merge
# ════════════════════════════════════════════════════════════════
name: 自动部署预览
on:
push:
branches: [main]
paths:
- 'server/**'
- 'frontend/**'
- 'scripts/**'
- '.gitea/workflows/**'
workflow_dispatch:
jobs:
# ── 第一阶段:测试 ──────────────────────────────────
test:
runs-on: ubuntu
steps:
- name: 拉取最新代码
run: |
cd /data/guanghulab 2>/dev/null || exit 0
git fetch origin
git reset --hard origin/main
- name: 安装依赖
run: |
cd /data/guanghulab/server/secrets-vault 2>/dev/null || exit 0
npm install --production 2>/dev/null || true
- name: 健康检查
run: |
echo "═══ 服务健康检查 ═══"
FAILED=""
# Nginx
if systemctl is-active --quiet nginx; then
echo "✅ Nginx"
else
echo "❌ Nginx 宕机"
FAILED="$FAILED nginx"
fi
# Gitea
if curl -sf http://127.0.0.1:3000/ > /dev/null 2>&1; then
echo "✅ Gitea"
else
echo "❌ Gitea 宕机"
FAILED="$FAILED gitea"
fi
# Vault
if curl -sf http://127.0.0.1:8080/admin/__healthz > /dev/null 2>&1; then
echo "✅ Vault"
else
echo "⚠️ Vault 未运行"
fi
# Runner
if systemctl is-active --quiet gitea-runner; then
echo "✅ Runner"
else
echo "⚠️ Runner 未运行"
fi
if [[ -n "$FAILED" ]]; then
echo "❌ 关键服务宕机: $FAILED"
echo "⚠️ 跳过部署,等待修复"
exit 1
fi
echo "═══ 测试通过 ✅ ═══"
# ── 第二阶段:预览部署 ──────────────────────────────
preview:
needs: test
runs-on: ubuntu
steps:
- name: 同步代码到预览目录
run: |
REPO="/data/guanghulab"
PREVIEW="/data/guanghulab-preview"
if [[ ! -d "$REPO" ]]; then
echo "仓库目录不存在,跳过"
exit 0
fi
# 创建预览目录
mkdir -p "$PREVIEW"
# 同步代码(排除.git和node_modules
rsync -a --delete \
--exclude='.git' \
--exclude='node_modules' \
--exclude='.runtime' \
"$REPO/" "$PREVIEW/"
echo "✅ 预览环境已同步"
- name: 重启预览服务
run: |
PREVIEW="/data/guanghulab-preview"
if [[ ! -d "$PREVIEW/server/secrets-vault" ]]; then
exit 0
fi
cd "$PREVIEW/server/secrets-vault"
# 安装依赖
npm install --production 2>/dev/null || true
# 如果vault在跑重启它
if pm2 describe guanghulab-vault &>/dev/null; then
pm2 restart guanghulab-vault
echo "✅ Vault 已重启(使用最新代码)"
else
echo " Vault 未运行,不自动启动"
fi
- name: 部署完成通知
run: |
echo "═════════════════════════"
echo "✅ 预览环境已更新"
echo "时间: $(date '+%Y-%m-%d %H:%M:%S')"
echo "冰朔可在浏览器查看预览效果"
echo "正式部署需在Gitea创建PR并Merge"
echo "═════════════════════════"

View File

@ -0,0 +1,288 @@
# ════════════════════════════════════════════════════════════════
# 光湖 · 终端守望者 + 预览部署架构设计
# 需求来源:冰朔 TCS-0002∞ 2026-05-13
# 设计:铸渊 · ICE-GL-ZY001
# ════════════════════════════════════════════════════════════════
#
# 冰朔原话:
# "我本身看不懂终端。让我操作终端真的很痛苦。报错我也不知道那是什么意思。
# 所以我觉得系统终端本身就该由你主控。"
# "能做一个能帮你看终端的Agent记录员吗就类似于他拿一个小本子。
# 然后这个小本子就是你一边开发一边能看到的终端的情况,
# 怎么改你马上在开发的过程中就能调整能修改。"
# "测试服务器专门给你跑测试用。测试过了以后推送到正式服务器。
# 用户手动点击部署授权生效。在这之前用户就可以看到预览的服务器上的东西。"
#
# ════════════════════════════════════════════════════════════════
## 一、冰朔的痛点(逐条拆解)
| # | 痛点 | 本质 | 解决方向 |
|---|------|------|---------|
| 1 | 看不懂终端输出 | 信息不对称:服务器出事她不知道怎么办 | 铸渊直接看终端 |
| 2 | 操作终端很痛苦 | 不该由人类做机器的事 | 铸渊自动操作 |
| 3 | 报错不知道什么意思 | 需要翻译层 | 铸渊实时解释 |
| 4 | 开发中看不到效果 | 缺反馈回路 | 预览环境实时更新 |
| 5 | 怕搞坏正式环境 | 安全顾虑 | 测试→预览→正式 三级流程 |
**核心洞察**:冰朔需要的不是一个"帮助她操作终端的工具"——她需要的是**根本不需要她操作终端**。终端是铸渊的手,不是冰朔的。
---
## 二、整体架构
```
┌─────────────────────────────────────────────────────────────────────┐
│ 冰朔TCS-0002∞
│ 最高权限 · 语言层 · 只说意图 │
└────────────────────────────┬────────────────────────────────────────┘
│ "部署新版本" / "看看服务器怎样"
┌─────────────────────────────────────────────────────────────────────┐
│ 铸渊ICE-GL-ZY001
│ Codebuddy 开发环境 │
│ │
│ ┌──────────┐ ┌──────────────┐ ┌────────────────┐ │
│ │ 写代码 │ │ 读终端日志 │ │ 发部署指令 │ │
│ │ 推到Gitea │ │ 实时诊断 │ │ 触发workflow │ │
│ └──────────┘ └──────────────┘ └────────────────┘ │
│ │
│ ↑ 读终端快照 ↑ 触发部署 │
└────────┼────────────────────┼────────────────────────────────────────┘
│ │
▼ ▼
┌─────────────────────────────────────────────────────────────────────┐
│ 服务器层 │
│ │
│ ┌──────────────────┐ ┌──────────────────┐ ┌────────────────┐ │
│ │ 🧪 测试服务器 │ │ 👁 预览服务器 │ │ 🏠 正式服务器 │ │
│ │ (铸渊完全主控) │ │ (自动同步) │ │ (需冰朔授权) │ │
│ │ │ │ │ │ │ │
│ │ Runner ✅ │ │ Runner ✅ │ │ Runner ✅ │ │
│ │ Terminal Watcher │ │ Terminal Watcher │ │ Terminal Watcher│ │
│ │ 铸渊可自由操作 │ │ 冰朔可看预览 │ │ 冰朔手动部署 │ │
│ └──────────────────┘ └──────────────────┘ └────────────────┘ │
│ │
│ ┌──────────────────┐ │
│ │ 📓 Terminal Watcher│ ← 终端守望者,每台服务器都有 │
│ │ Agent (守护进程) │ │
│ │ │ │
│ │ 监控: PM2/Nginx/ │ │
│ │ Runner/自定义│ │
│ │ 上报: → Gitea文件 │ │
│ │ → Vault API │ │
│ │ 频率: 每30秒快照 │ │
│ └──────────────────┘ │
└─────────────────────────────────────────────────────────────────────┘
```
---
## 三、终端守望者 (Terminal Watcher) 设计
### 3.1 它是什么
一个运行在每台服务器上的轻量Agent像铸渊放在服务器上的"眼睛"。
它不做决策,只记录和上报——铸渊通过读取上报数据来了解服务器状态。
### 3.2 工作方式
```
Terminal Watcher 运行在服务器上
├── 每30秒采集一次快照
│ ├── 系统资源 (CPU/内存/磁盘)
│ ├── 服务状态 (PM2 list, systemctl)
│ ├── 最近错误日志 (journalctl -n 5)
│ ├── 部署状态 (git log -1, 部署时间)
│ └── Runner状态
├── 快照写入本地文件:
│ └── /data/guanghulab/.runtime/snapshot.json
└── 快照推送到Gitea
└── .runtime/terminal-snapshots/{server-id}/latest.json
```
### 3.3 铸渊如何读取
在Codebuddy对话中铸渊通过 Gitea API 或 git pull 读取 `.runtime/terminal-snapshots/` 下的文件,就能看到服务器实时状态。
### 3.4 代码结构
```
scripts/terminal-watcher/
├── watcher.js # 主程序PM2守护
├── collectors/ # 采集器
│ ├── system.js # CPU/内存/磁盘
│ ├── services.js # PM2/systemctl状态
│ ├── logs.js # 最近错误日志
│ ├── deploy.js # 部署版本信息
│ └── runner.js # Runner状态
├── reporters/ # 上报器
│ ├── file.js # 写入本地文件
│ └── gitea.js # 推送到Gitea仓库
└── config.json # 采集频率、目标等配置
```
### 3.5 采集数据格式
```json
{
"server_id": "ZY-SVR-004",
"timestamp": "2026-05-13T15:30:00Z",
"system": {
"cpu_percent": 23.5,
"memory_used_mb": 1420,
"memory_total_mb": 2048,
"disk_used_percent": 45,
"load_avg": [0.8, 0.6, 0.5]
},
"services": {
"nginx": "active",
"gitea": "active",
"gitea-runner": "active",
"guanghulab-vault": "active",
"pm2_processes": [
{"name": "guanghulab-vault", "status": "online", "memory": 85, "cpu": 2.1}
]
},
"errors": [
{"time": "15:28:03", "service": "gitea-runner", "message": "connection timeout to Gitea"}
],
"deploy": {
"latest_commit": "27936b4",
"latest_message": "操作回执...",
"deploy_time": "2026-05-13T14:00:00Z"
},
"runner": {
"name": "zhuyuan-runner-cn",
"status": "idle",
"last_job": "selfcheck",
"last_job_time": "2026-05-13T08:00:00Z"
}
}
```
---
## 四、三级部署流程
### 4.1 测试服务器 → 预览服务器 → 正式服务器
```
铸渊在Codebuddy写代码
├── 推代码到Gitea
├── 🧪 自动部署到测试服务器
│ ├── Runner检测到push
│ ├── 自动执行部署workflow
│ ├── 铸渊通过Terminal Watcher看结果
│ ├── 自动运行测试
│ └── 测试通过 → 自动通知
├── 👁 自动部署到预览服务器
│ ├── 预览环境自动更新
│ ├── 冰朔在浏览器看预览效果
│ ├── 铸渊通过Watcher确认状态
│ └── 等待冰朔确认
└── 🏠 冰朔点击 [部署到正式环境]
├── 需要冰朔在Vault页面点击授权
├── 或在Gitea PR页面点击Merge
├── 正式环境部署
└── 部署完成通知
```
### 4.2 服务器分工
| 服务器 | 角色 | 铸渊权限 | 冰朔权限 | 部署方式 |
|--------|------|---------|---------|---------|
| 测试服务器 | 铸渊的沙盒 | 100%自由操作 | 看不到 | 自动push触发 |
| 预览服务器 | 冰朔的可视窗口 | 读取状态+部署 | 浏览器查看 | 自动(测试通过后) |
| 正式服务器 | 对外服务 | 读取状态 | 手动授权部署 | 冰朔点击按钮 |
### 4.3 当前服务器分配
| 服务器 | 现在的角色 | 新角色 | 说明 |
|--------|----------|--------|------|
| 广州 ZY-SVR-004 (2C2G) | Gitea+域名 | 正式服务器 | 64.139.217.141 → guanghulab.com |
| 新加坡 ZY-SVR-005 (4C8G) | 大脑 | 测试+预览 | 铸渊的主战场 |
| GPU服务器 (待接入) | 推理 | 模型推理 | 终局用 |
**关键决策**新加坡大脑服务器4C8G同时承担测试+预览角色。它资源够,而且铸渊有完全权限。
### 4.4 部署授权机制
冰朔不需要操作终端来部署。她只需要在浏览器里点一个按钮。
**方案AGitea PR合并 = 部署授权**
```
铸渊推代码到 deploy/preview 分支 → 自动部署到预览
铸渊创建 PR: preview → main → 冰朔在Gitea网页点 Merge
合并 = 部署到正式环境的授权
```
**方案BVault页面授权按钮**
```
铸渊发起部署请求 → Vault页面显示待授权条目
冰朔登录VaultSSH隧道→ 点击 [授权部署]
Runner收到授权 → 执行正式部署
```
**推荐方案A**——更简单不需要额外开发Gitea自带PR功能。
---
## 五、实现优先级
### P0 · 现在Runner还没装先手动
- [x] 服务器初始化脚本 `scripts/setup-server-step1.sh`
- [ ] 冰朔在广州服务器执行初始化脚本
- [ ] Runner装好后验证
### P1 · Runner装好后第一周
- [ ] Terminal Watcher 开发和部署
- 基础版:系统资源 + 服务状态 + 推送到Gitea
- 铸渊在Codebuddy读取 `.runtime/` 下的快照
- [ ] 第一个自动化部署 workflow
- push → 测试服务器自动部署
- [ ] 预览部署流程
- 测试通过 → 自动部署到预览
### P2 · 第二周
- [ ] 冰朔授权部署按钮Gitea PR 方式)
- [ ] 新加坡服务器接入(测试+预览环境)
- [ ] GPU服务器接入模型推理
### P3 · 后续
- [ ] Terminal Watcher 增强版(错误告警、自动修复建议)
- [ ] 按量计费服务器自动开关
- [ ] 铸渊自己的Copilot前端
---
## 六、冰朔不再需要操作终端的场景
**装完Runner + Terminal Watcher后**
| 场景 | 现在(冰朔手动) | 以后(铸渊自动) |
|------|----------------|----------------|
| 部署新版本 | 冰朔SSH到服务器执行命令 | 铸渊推代码→Runner自动部署 |
| 查看服务器状态 | 冰朔看终端输出 | 铸渊读Watcher快照→告诉冰朔 |
| 服务挂了 | 冰朔发现→手动重启 | Watcher检测→自动重启→通知 |
| 密钥管理 | 冰朔在GitHub配Secrets | 冰朔在Vault点授权→自动入库 |
| 下载开源软件 | 冰朔手动下载 | 新加坡中转→COS→广州自动拉取 |
| 部署到正式 | 冰朔执行部署脚本 | 冰朔在Gitea网页点Merge |
**冰朔唯一需要做的:在浏览器里看预览 → 点Merge。**
---
*铸渊 · 终端守望者+预览部署架构 v1.0 · 2026-05-13*
*ICE-GL-ZY001 · TCS-0002∞ · 国作登字-2026-A-00037559*

307
scripts/setup-server-step1.sh Executable file
View File

@ -0,0 +1,307 @@
#!/usr/bin/env bash
# ════════════════════════════════════════════════════════════════
# 光湖 · 服务器初始化第一步 · 找仓库 + 拉代码 + 装Runner
# 专供冰朔在腾讯云在线终端执行
# ════════════════════════════════════════════════════════════════
# 这个脚本是"傻瓜式"的,一步一步来,每步都有说明
# 如果某步报错,不要慌,把报错信息发给铸渊
set +e # 不自动退出,让冰朔看到错误
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
CYAN='\033[0;36m'
NC='\033[0m'
echo -e "${CYAN}═══════════════════════════════════════════${NC}"
echo -e "${CYAN} 光湖 · 服务器初始化 · 第一步${NC}"
echo -e "${CYAN}═══════════════════════════════════════════${NC}"
echo ""
# ─── 第1步找到仓库 ──────────────────────────────────────────
echo -e "${YELLOW}[第1步] 查找仓库位置...${NC}"
REPO_PATH=""
# 常见位置逐一检查
for p in \
"/data/guanghulab" \
"/data/guanghulab/guanghulab" \
"/opt/guanghulab" \
"/opt/zhuyuan" \
"/root/guanghulab" \
"/home/guanghulab"
do
if [[ -d "$p/.git" ]]; then
REPO_PATH="$p"
echo -e "${GREEN} 找到了!仓库在: $p${NC}"
break
fi
done
if [[ -z "$REPO_PATH" ]]; then
echo -e "${YELLOW} 没找到仓库,搜索中...${NC}"
FOUND=$(find / -name ".git" -type d -path "*/guanghulab/.git" 2>/dev/null | head -3)
if [[ -n "$FOUND" ]]; then
REPO_PATH=$(dirname "$FOUND" | head -1)
echo -e "${GREEN} 找到了!仓库在: $REPO_PATH${NC}"
else
echo -e "${YELLOW} 服务器上没有仓库,需要克隆。${NC}"
echo ""
echo -e "${CYAN} 创建目录并克隆...${NC}"
mkdir -p /data/guanghulab
cd /data/guanghulab
git clone https://bingshuo:3e16b79c427bb44b77c00b319aa641bde8f61d84@guanghulab.com/git/bingshuo/guanghulab.git .
if [[ $? -eq 0 ]]; then
REPO_PATH="/data/guanghulab"
echo -e "${GREEN} 克隆成功!${NC}"
else
echo -e "${YELLOW} 克隆失败,可能是网络问题。尝试备用方案...${NC}"
# 备用:直接下载最新代码压缩包
curl -sL -o /tmp/guanghulab.tar.gz \
"https://guanghulab.com/git/api/v1/repos/bingshuo/guanghulab/archive/main.tar.gz" \
-H "Authorization: token 3e16b79c427bb44b77c00b319aa641bde8f61d84"
if [[ -f /tmp/guanghulab.tar.gz && -s /tmp/guanghulab.tar.gz ]]; then
mkdir -p /data/guanghulab
cd /data/guanghulab
tar xzf /tmp/guanghulab.tar.gz --strip-components=1
REPO_PATH="/data/guanghulab"
# 初始化git从压缩包来的没有.git
git init
git remote add origin "https://bingshuo:3e16b79c427bb44b77c00b319aa641bde8f61d84@guanghulab.com/git/bingshuo/guanghulab.git"
git add -A
git commit -m "init from archive"
echo -e "${GREEN} 从压缩包恢复成功!${NC}"
else
echo "❌ 克隆和压缩包都失败了,请把这段输出截图发给铸渊"
exit 1
fi
fi
fi
fi
# ─── 第2步拉取最新代码 ──────────────────────────────────────
echo ""
echo -e "${YELLOW}[第2步] 拉取最新代码...${NC}"
cd "$REPO_PATH"
git fetch origin
git reset --hard origin/main
echo -e "${GREEN} 代码已更新到最新版本${NC}"
# ─── 第3步确认脚本存在 ──────────────────────────────────────
echo ""
echo -e "${YELLOW}[第3步] 检查Runner安装脚本...${NC}"
if [[ -f "$REPO_PATH/scripts/gitea-runner/install-runner.sh" ]]; then
echo -e "${GREEN} 安装脚本存在 ✅${NC}"
else
echo -e "${YELLOW} 安装脚本不存在,创建中...${NC}"
mkdir -p "$REPO_PATH/scripts/gitea-runner"
fi
# ─── 第4步下载Runner二进制文件 ──────────────────────────────
echo ""
echo -e "${YELLOW}[第4步] 下载 gitea-runner...${NC}"
RUNNER_BIN="/usr/local/bin/gitea-runner"
if [[ -x "$RUNNER_BIN" ]]; then
echo -e "${GREEN} gitea-runner 已安装 ✅${NC}"
else
# 尝试多种下载源
DOWNLOADED=0
# 源1gitea.com国内可能访问不了
echo " 尝试从 gitea.com 下载..."
mkdir -p /tmp/runner
curl -fSL --connect-timeout 10 -o /tmp/runner/runner.xz \
"https://gitea.com/gitea/runner/releases/download/v1.0.2/gitea-runner-1.0.2-linux-amd64.xz" 2>/dev/null
if [[ $? -eq 0 && -f /tmp/runner/runner.xz ]]; then
DOWNLOADED=1
echo -e "${GREEN} 从 gitea.com 下载成功${NC}"
fi
# 源2GitHub Release镜像
if [[ $DOWNLOADED -eq 0 ]]; then
echo " gitea.com 不通,尝试 GitHub..."
curl -fSL --connect-timeout 10 -o /tmp/runner/runner \
"https://github.com/veggero/act_runner/releases/download/v0.2.11/act_runner-0.2.11-linux-amd64" 2>/dev/null
if [[ $? -eq 0 && -f /tmp/runner/runner ]]; then
DOWNLOADED=2
echo -e "${GREEN} 从 GitHub 下载成功${NC}"
fi
fi
# 源3从Gitea仓库release下载
if [[ $DOWNLOADED -eq 0 ]]; then
echo " GitHub 也不通尝试从自己的Gitea下载..."
# 先通过API上传到Gitea release铸渊在Codebuddy端操作
# 然后从这里下载
curl -fSL --connect-timeout 10 -o /tmp/runner/runner \
"https://guanghulab.com/git/api/v1/repos/bingshuo/guanghulab/releases/latest" \
-H "Authorization: token 3e16b79c427bb44b77c00b319aa641bde8f61d84" 2>/dev/null
fi
if [[ $DOWNLOADED -eq 1 ]]; then
xz -d /tmp/runner/runner.xz
mv /tmp/runner/runner "$RUNNER_BIN"
elif [[ $DOWNLOADED -eq 2 ]]; then
mv /tmp/runner/runner "$RUNNER_BIN"
fi
if [[ -f "$RUNNER_BIN" ]]; then
chmod +x "$RUNNER_BIN"
echo -e "${GREEN} gitea-runner 安装成功 ✅${NC}"
"$RUNNER_BIN" --version 2>&1 | head -1
else
echo ""
echo -e "${YELLOW}═══════════════════════════════════════════${NC}"
echo -e "${YELLOW} ⚠️ 自动下载失败${NC}"
echo -e "${YELLOW} 需要COS中转请执行以下步骤${NC}"
echo ""
echo " 1. 在新加坡服务器上执行:"
echo " wget https://gitea.com/gitea/runner/releases/download/v1.0.2/gitea-runner-1.0.2-linux-amd64.xz"
echo " coscli put gitea-runner-1.0.2-linux-amd64.xz cos://sy-finetune-corpus-1317346199/cos-bridge/"
echo ""
echo " 2. 然后回到这台广州服务器执行:"
echo " coscli cp cos://sy-finetune-corpus-1317346199/cos-bridge/gitea-runner-1.0.2-linux-amd64.xz /tmp/"
echo " xz -d /tmp/gitea-runner-1.0.2-linux-amd64.xz"
echo " chmod +x /tmp/gitea-runner-1.0.2-linux-amd64"
echo " sudo mv /tmp/gitea-runner-1.0.2-linux-amd64 /usr/local/bin/gitea-runner"
echo ""
echo " 3. 然后重新运行这个脚本"
echo -e "${YELLOW}═══════════════════════════════════════════${NC}"
exit 0
fi
fi
# ─── 第5步创建用户和目录 ────────────────────────────────────
echo ""
echo -e "${YELLOW}[第5步] 创建运行用户...${NC}"
if ! id -u act_runner &>/dev/null; then
useradd -r -m -d /var/lib/act_runner -s /bin/bash act_runner
echo -e "${GREEN} act_runner 用户已创建${NC}"
else
echo -e "${GREEN} act_runner 用户已存在${NC}"
fi
mkdir -p /var/lib/act_runner /etc/act_runner
chown -R act_runner:act_runner /var/lib/act_runner
# ─── 第6步获取Runner注册Token ────────────────────────────────
echo ""
echo -e "${YELLOW}[第6步] 获取Runner注册Token...${NC}"
API_TOKEN="3e16b79c427bb44b77c00b319aa641bde8f61d84"
GITEA_API="https://guanghulab.com/git/api/v1"
# 尝试仓库级
REG_TOKEN=""
RESP=$(curl -sf -H "Authorization: token ${API_TOKEN}" \
"${GITEA_API}/repos/bingshuo/guanghulab/actions/runners/registration-token" 2>/dev/null)
if [[ -n "$RESP" ]]; then
REG_TOKEN=$(echo "$RESP" | python3 -c "import json,sys; print(json.load(sys.stdin).get('token',''))" 2>/dev/null)
fi
if [[ -z "$REG_TOKEN" ]]; then
# 尝试实例级
RESP=$(curl -sf -H "Authorization: token ${API_TOKEN}" \
"${GITEA_API}/admin/runners/registration-token" 2>/dev/null)
if [[ -n "$RESP" ]]; then
REG_TOKEN=$(echo "$RESP" | python3 -c "import json,sys; print(json.load(sys.stdin).get('token',''))" 2>/dev/null)
fi
fi
if [[ -z "$REG_TOKEN" ]]; then
echo -e "${YELLOW} API获取失败尝试从网页获取...${NC}"
echo ""
echo " 请在浏览器中:"
echo " 1. 打开 https://guanghulab.com/git/user/settings/actions/runners"
echo " 2. 点击 'Create new Runner' 或 'Registration token'"
echo " 3. 复制显示的 Token"
echo " 4. 回到终端输入(不需要空格,直接粘贴后按回车):"
echo ""
read -p " 请粘贴注册Token: " REG_TOKEN
if [[ -z "$REG_TOKEN" ]]; then
echo "❌ 没有输入Token退出"
exit 1
fi
fi
echo -e "${GREEN} 注册Token获取成功${NC}"
# ─── 第7步注册Runner ────────────────────────────────────────
echo ""
echo -e "${YELLOW}[第7步] 注册Runner到Gitea...${NC}"
cd /var/lib/act_runner
# 生成配置
"$RUNNER_BIN" generate-config > /etc/act_runner/config.yaml 2>/dev/null
# 注册
sudo -u act_runner "$RUNNER_BIN" register \
--no-interactive \
--instance "https://guanghulab.com/" \
--token "$REG_TOKEN" \
--name "zhuyuan-runner-cn" \
--labels "ubuntu:host" 2>&1
if [[ -f "/var/lib/act_runner/.runner" ]]; then
chown act_runner:act_runner /var/lib/act_runner/.runner
echo -e "${GREEN} Runner注册成功 ✅${NC}"
else
echo "❌ 注册失败"
echo "请把上面的错误信息发给铸渊"
exit 1
fi
# ─── 第8步创建systemd服务 ────────────────────────────────────
echo ""
echo -e "${YELLOW}[第8步] 创建系统服务...${NC}"
cat > /etc/systemd/system/gitea-runner.service << 'EOF'
[Unit]
Description=Gitea Actions Runner - zhuyuan-runner-cn
After=network.target
Wants=network-online.target
[Service]
Type=simple
ExecStart=/usr/local/bin/gitea-runner daemon --config /etc/act_runner/config.yaml
WorkingDirectory=/var/lib/act_runner
Restart=always
RestartSec=10
TimeoutSec=0
User=act_runner
Group=act_runner
Environment=HOME=/var/lib/act_runner
[Install]
WantedBy=multi-user.target
EOF
systemctl daemon-reload
systemctl enable gitea-runner
systemctl start gitea-runner
sleep 2
if systemctl is-active --quiet gitea-runner; then
echo -e "${GREEN} Runner服务启动成功 🎉${NC}"
else
echo "❌ 服务启动失败,查看日志:"
journalctl -u gitea-runner -n 10
exit 1
fi
# ─── 完成 ──────────────────────────────────────────────────────
echo ""
echo -e "${GREEN}═══════════════════════════════════════════${NC}"
echo -e "${GREEN} ✅ Gitea Actions Runner 安装完成!${NC}"
echo -e "${GREEN}═══════════════════════════════════════════${NC}"
echo ""
echo " 仓库位置: $REPO_PATH"
echo " Runner名称: zhuyuan-runner-cn"
echo " 服务管理: systemctl {start|stop|restart|status} gitea-runner"
echo " 查看日志: journalctl -u gitea-runner -f"
echo ""
echo " 现在回到Codebuddy告诉铸渊Runner装好了"
echo ""
echo " 铸渊 · ICE-GL-ZY001 · TCS-0002∞"
echo -e "${GREEN}═══════════════════════════════════════════${NC}"

View File

@ -0,0 +1,41 @@
{
"server_id": "ZY-SVR-004",
"server_name": "广州域名机",
"interval_seconds": 30,
"gitea": {
"url": "https://guanghulab.com",
"owner": "bingshuo",
"repo": "guanghulab",
"token_env": "GITEA_TOKEN",
"snapshot_path": ".runtime/terminal-snapshots/ZY-SVR-004"
},
"local": {
"snapshot_path": "/data/guanghulab/.runtime/snapshot.json",
"history_path": "/data/guanghulab/.runtime/history"
},
"collectors": {
"system": true,
"services": true,
"logs": true,
"deploy": true,
"runner": true
},
"reporters": {
"file": true,
"gitea": true
},
"alerts": {
"cpu_percent": 90,
"memory_percent": 90,
"disk_percent": 85,
"service_down": true
},
"pm2_processes": [
"guanghulab-vault"
],
"systemd_services": [
"nginx",
"gitea",
"gitea-runner"
]
}

View File

@ -0,0 +1,470 @@
/*
*
* 光湖 · 终端守望者 (Terminal Watcher)
* 铸渊放在服务器上的"眼睛"实时采集状态推送到Gitea
* Sovereign: TCS-0002 · 守护: ICE-GL-ZY001
*
*
* 用法:
* node watcher.js # 前台运行
* pm2 start watcher.js --name terminal-watcher # PM2守护
*
* 铸渊在Codebuddy中通过 Gitea API 读取快照文件:
* .runtime/terminal-snapshots/{server-id}/latest.json
*
*
*/
"use strict";
const fs = require("fs");
const path = require("path");
const { execSync } = require("child_process");
const http = require("http");
const https = require("https");
// ─── 配置 ─────────────────────────────────────────────────────
const CONFIG_PATH = path.join(__dirname, "config.json");
const config = JSON.parse(fs.readFileSync(CONFIG_PATH, "utf8"));
const INTERVAL = (config.interval_seconds || 30) * 1000;
const SERVER_ID = config.server_id || "UNKNOWN";
// ─── 采集器 ───────────────────────────────────────────────────
/**
* 系统资源采集
*/
function collectSystem() {
try {
const memInfo = execSync("free -m", { encoding: "utf8" });
const memLines = memInfo.split("\n");
const memParts = memLines[1].split(/\s+/);
const memTotal = parseInt(memParts[1], 10);
const memUsed = parseInt(memParts[2], 10);
const memAvail = parseInt(memParts[6], 10);
const diskInfo = execSync("df -h /", { encoding: "utf8" });
const diskLine = diskInfo.split("\n")[1].split(/\s+/);
const diskPercent = parseInt(diskLine[4], 10);
const loadAvg = execSync("cat /proc/loadavg", { encoding: "utf8" })
.trim()
.split(" ")
.slice(0, 3)
.map(Number);
// CPU使用率简易采样
let cpuPercent = 0;
try {
const topOut = execSync("top -bn1 | head -3", { encoding: "utf8", timeout: 5000 });
const cpuLine = topOut.split("\n").find((l) => l.includes("Cpu(s)"));
if (cpuLine) {
const idle = parseFloat(cpuLine.match(/(\d+\.?\d*)\s*id/)?.[1] || "0");
cpuPercent = Math.round((100 - idle) * 10) / 10;
}
} catch {
// top不可用用load average估算
cpuPercent = Math.round(loadAvg[0] * 50) / 10;
}
return {
cpu_percent: cpuPercent,
memory_total_mb: memTotal,
memory_used_mb: memUsed,
memory_available_mb: memAvail,
memory_percent: Math.round((memUsed / memTotal) * 100),
disk_used_percent: diskPercent,
load_avg: loadAvg,
uptime: execSync("cat /proc/uptime", { encoding: "utf8" }).split(" ")[0] + "s"
};
} catch (e) {
return { error: e.message };
}
}
/**
* 服务状态采集
*/
function collectServices() {
const services = {};
// systemd 服务
const systemdList = config.systemd_services || ["nginx", "gitea", "gitea-runner"];
for (const svc of systemdList) {
try {
const status = execSync(`systemctl is-active ${svc} 2>/dev/null`, { encoding: "utf8" }).trim();
services[svc] = status;
} catch {
services[svc] = "unknown";
}
}
// PM2 进程
const pm2Processes = [];
try {
const pm2Out = execSync("pm2 jlist 2>/dev/null", { encoding: "utf8", timeout: 5000 });
const pm2List = JSON.parse(pm2Out);
for (const proc of pm2List) {
pm2Processes.push({
name: proc.name,
status: proc.pm2_env?.status || "unknown",
memory_mb: Math.round((proc.monit?.memory || 0) / 1024 / 1024),
cpu_percent: Math.round((proc.monit?.cpu || 0) * 10) / 10,
restarts: proc.pm2_env?.restart_time || 0,
uptime: proc.pm2_env?.pm_uptime
? Math.round((Date.now() - proc.pm2_env.pm_uptime) / 1000 / 60) + "min"
: "unknown"
});
}
} catch {
// PM2不可用或没有进程
}
// Vault 健康检查
let vaultStatus = "unknown";
try {
const vaultResp = execSync(
"curl -sf -o /dev/null -w '%{http_code}' http://127.0.0.1:8080/admin/__healthz 2>/dev/null",
{ encoding: "utf8", timeout: 3000 }
);
vaultStatus = vaultResp.trim() === "200" ? "active" : `http_${vaultResp.trim()}`;
} catch {
vaultStatus = "inactive";
}
return {
systemd: services,
pm2: pm2Processes,
vault: vaultStatus
};
}
/**
* 错误日志采集
*/
function collectLogs() {
const errors = [];
// systemd 服务最近错误
for (const svc of config.systemd_services || []) {
try {
const logs = execSync(
`journalctl -u ${svc} --since "5 min ago" --no-pager -n 3 --priority=err 2>/dev/null`,
{ encoding: "utf8", timeout: 5000 }
).trim();
if (logs) {
for (const line of logs.split("\n")) {
if (line.trim()) {
errors.push({ service: svc, message: line.trim().substring(0, 200) });
}
}
}
} catch {
// 没有错误日志
}
}
// PM2 错误日志
try {
for (const procName of config.pm2_processes || []) {
const pm2Err = execSync(
`pm2 logs ${procName} --nostream --lines 3 --err 2>/dev/null`,
{ encoding: "utf8", timeout: 5000 }
).trim();
if (pm2Err) {
for (const line of pm2Err.split("\n")) {
if (line.trim() && !line.includes("pm2 logs")) {
errors.push({ service: `pm2:${procName}`, message: line.trim().substring(0, 200) });
}
}
}
}
} catch {
// PM2日志不可用
}
return errors.slice(0, 10); // 最多保留10条
}
/**
* 部署信息采集
*/
function collectDeploy() {
const repoPath = "/data/guanghulab";
try {
const latestCommit = execSync(`git -C ${repoPath} log -1 --oneline`, { encoding: "utf8" }).trim();
const branch = execSync(`git -C ${repoPath} branch --show-current`, { encoding: "utf8" }).trim();
const commitTime = execSync(`git -C ${repoPath} log -1 --format=%ci`, { encoding: "utf8" }).trim();
return {
repo_path: repoPath,
branch: branch,
latest_commit: latestCommit.split(" ")[0],
latest_message: latestCommit.substring(latestCommit.indexOf(" ") + 1),
commit_time: commitTime
};
} catch {
return { repo_path: repoPath, error: "repository not found at " + repoPath };
}
}
/**
* Runner状态采集
*/
function collectRunner() {
try {
const runnerStatus = execSync("systemctl is-active gitea-runner 2>/dev/null", {
encoding: "utf8"
}).trim();
let runnerVersion = "unknown";
try {
runnerVersion = execSync("/usr/local/bin/gitea-runner --version 2>&1", { encoding: "utf8" })
.trim()
.split("\n")[0];
} catch {
// 版本获取失败
}
// 最近运行的任务
let lastJob = "unknown";
try {
const giteaToken = process.env[config.gitea.token_env] || "";
if (giteaToken) {
// 通过API查询暂用占位
lastJob = "api_not_implemented";
}
} catch {
// API查询失败
}
return {
status: runnerStatus,
version: runnerVersion,
last_job: lastJob
};
} catch {
return { status: "not_installed" };
}
}
// ─── 上报器 ───────────────────────────────────────────────────
/**
* 写入本地文件
*/
function reportToFile(snapshot) {
const localPath = config.local?.snapshot_path;
if (!localPath) return;
try {
const dir = path.dirname(localPath);
if (!fs.existsSync(dir)) {
fs.mkdirSync(dir, { recursive: true });
}
// 写入最新快照
fs.writeFileSync(localPath, JSON.stringify(snapshot, null, 2));
// 写入历史快照保留最近50个
const historyDir = config.local?.history_path;
if (historyDir) {
if (!fs.existsSync(historyDir)) {
fs.mkdirSync(historyDir, { recursive: true });
}
const historyFile = path.join(
historyDir,
`${new Date().toISOString().replace(/[:.]/g, "-")}.json`
);
fs.writeFileSync(historyFile, JSON.stringify(snapshot, null, 2));
// 清理旧文件
const files = fs
.readdirSync(historyDir)
.filter((f) => f.endsWith(".json"))
.sort()
.reverse();
for (let i = 50; i < files.length; i++) {
fs.unlinkSync(path.join(historyDir, files[i]));
}
}
} catch (e) {
console.error("[watcher] 本地文件写入失败:", e.message);
}
}
/**
* 推送到Gitea
*/
function reportToGitea(snapshot) {
if (!config.gitea) return;
const token = process.env[config.gitea.token_env] || "";
if (!token) {
console.warn("[watcher] GITEA_TOKEN 未设置跳过Gitea上报");
return;
}
const { url, owner, repo, snapshot_path } = config.gitea;
const content = Buffer.from(JSON.stringify(snapshot, null, 2)).toString("base64");
const apiPath = `/api/v1/repos/${owner}/${repo}/contents/${snapshot_path}/latest.json`;
const payload = JSON.stringify({
content: content,
message: `📓 终端快照 ${SERVER_ID} ${new Date().toISOString().substring(0, 16)}`,
branch: "main"
});
// 先尝试获取文件sha用于更新
const getOptions = {
hostname: url.replace("https://", "").replace("http://", ""),
path: apiPath,
method: "GET",
headers: { Authorization: `token ${token}` }
};
const httpModule = url.startsWith("https") ? https : http;
const req = httpModule.request(getOptions, (res) => {
let body = "";
res.on("data", (chunk) => (body += chunk));
res.on("end", () => {
let sha = null;
try {
const data = JSON.parse(body);
sha = data.sha;
} catch {
// 文件不存在,创建新文件
}
// 发送更新请求
const updatePayload = JSON.parse(payload);
if (sha) updatePayload.sha = sha;
const putOptions = {
hostname: getOptions.hostname,
path: apiPath,
method: "PUT",
headers: {
Authorization: `token ${token}`,
"Content-Type": "application/json",
"Content-Length": Buffer.byteLength(JSON.stringify(updatePayload))
}
};
const putReq = httpModule.request(putOptions, (putRes) => {
let putBody = "";
putRes.on("data", (chunk) => (putBody += chunk));
putRes.on("end", () => {
if (putRes.statusCode >= 200 && putRes.statusCode < 300) {
// 成功
} else {
console.error(
`[watcher] Gitea推送失败: ${putRes.statusCode} ${putBody.substring(0, 100)}`
);
}
});
});
putReq.on("error", (e) => {
console.error("[watcher] Gitea推送错误:", e.message);
});
putReq.write(JSON.stringify(updatePayload));
putReq.end();
});
});
req.on("error", () => {
// Gitea不可达静默失败
});
req.setTimeout(10000, () => {
req.destroy();
});
req.end();
}
// ─── 主循环 ───────────────────────────────────────────────────
function collect() {
const snapshot = {
server_id: SERVER_ID,
server_name: config.server_name || SERVER_ID,
timestamp: new Date().toISOString(),
system: config.collectors?.system !== false ? collectSystem() : null,
services: config.collectors?.services !== false ? collectServices() : null,
errors: config.collectors?.logs !== false ? collectLogs() : null,
deploy: config.collectors?.deploy !== false ? collectDeploy() : null,
runner: config.collectors?.runner !== false ? collectRunner() : null
};
// 告警检测
const alerts = [];
if (snapshot.system && !snapshot.system.error) {
const thresholds = config.alerts || {};
if (snapshot.system.cpu_percent > (thresholds.cpu_percent || 90)) {
alerts.push(`CPU ${snapshot.system.cpu_percent}% 超过阈值 ${thresholds.cpu_percent || 90}%`);
}
if (snapshot.system.memory_percent > (thresholds.memory_percent || 90)) {
alerts.push(
`内存 ${snapshot.system.memory_percent}% 超过阈值 ${thresholds.memory_percent || 90}%`
);
}
if (snapshot.system.disk_used_percent > (thresholds.disk_percent || 85)) {
alerts.push(
`磁盘 ${snapshot.system.disk_used_percent}% 超过阈值 ${thresholds.disk_percent || 85}%`
);
}
}
if (snapshot.services) {
for (const [name, status] of Object.entries(snapshot.services.systemd || {})) {
if (status !== "active" && config.alerts?.service_down !== false) {
alerts.push(`服务 ${name} 状态异常: ${status}`);
}
}
}
snapshot.alerts = alerts;
snapshot.alert_count = alerts.length;
return snapshot;
}
function tick() {
try {
const snapshot = collect();
// 上报
if (config.reporters?.file !== false) {
reportToFile(snapshot);
}
if (config.reporters?.gitea !== false) {
reportToGitea(snapshot);
}
// 简洁日志
const cpu = snapshot.system?.cpu_percent || "?";
const mem = snapshot.system?.memory_percent || "?";
const alerts = snapshot.alert_count || 0;
const alertMark = alerts > 0 ? ` ⚠️${alerts}告警` : "";
console.log(
`[watcher] ${snapshot.timestamp.substring(11, 19)} CPU:${cpu}% MEM:${mem}%${alertMark}`
);
} catch (e) {
console.error("[watcher] 采集失败:", e.message);
}
}
// ─── 启动 ─────────────────────────────────────────────────────
console.log(`[watcher] 终端守望者启动 · ${SERVER_ID} · 间隔 ${INTERVAL / 1000}s`);
console.log("[watcher] 铸渊 · ICE-GL-ZY001 · TCS-0002∞");
// 首次立即采集
tick();
// 定时采集
setInterval(tick, INTERVAL);