master
78427b2e33
- Prerequisites (hardware, software, versions) - Step-by-step deployment instructions - Configuration examples - Troubleshooting table - OTA update workflow
200 lines
4.0 KiB
Markdown
200 lines
4.0 KiB
Markdown
# AI-tianyan 边缘部署操作指南
|
||
|
||
## 前置条件
|
||
|
||
| 项目 | 要求 |
|
||
|---|---|
|
||
| 硬件 | Atlas 200I DK2 (Ascend 310B4) 或 Ascend 310B4 设备 |
|
||
| OS | Ubuntu 22.04 aarch64 |
|
||
| Go | 1.18+ |
|
||
| Python | 3.9+ |
|
||
| CANN | 6.2.RC2 (Ascend Toolkit) |
|
||
| 云端 | 101.36.73.102 (API/MQTT/InfluxDB/OTA/ZLMediaKit) |
|
||
|
||
---
|
||
|
||
## 步骤 1: 拉取代码
|
||
|
||
```bash
|
||
git clone http://101.36.73.102:3112/fyah/AI-tianyan.git
|
||
cd AI-tianyan
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 2: 转换 YOLO 模型 (.onnx → .om)
|
||
|
||
```bash
|
||
# 在 CANN 环境设备上执行
|
||
source /usr/local/Ascend/ascend-toolkit/set_env.sh
|
||
|
||
# 执行 ATC 转换
|
||
bash scripts/atc_convert.sh
|
||
|
||
# 或手动转换
|
||
atc --model=yolov8n.onnx \
|
||
--framework=5 \
|
||
--output=model/model.om \
|
||
--soc_version=Ascend310B4 \
|
||
--input_format=NCHW \
|
||
--input_shape="images:1,3,640,640"
|
||
|
||
# 复制模型到指定位置
|
||
cp yolov8n.om model/model.om
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 3: 编译 Go 边缘代理
|
||
|
||
```bash
|
||
# 设置 Go 代理（国内环境）
|
||
export GOPROXY=https://goproxy.cn,direct
|
||
|
||
# 构建
|
||
bash scripts/build.sh
|
||
# 或使用 make build
|
||
# 输出: build/edge-agent
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 4: 安装部署
|
||
|
||
```bash
|
||
# 一键安装
|
||
bash scripts/install.sh
|
||
# 或 make install
|
||
|
||
# 安装脚本会:
|
||
# 1. 创建 /opt/tianyan-edge/{bin,python,config,systemd,staging,model}
|
||
# 2. 复制 edge-agent 到 /opt/tianyan-edge/bin/
|
||
# 3. 复制 infer_server.py 到 /opt/tianyan-edge/python/
|
||
# 4. 安装 Python 依赖 (opencv-python-headless, numpy)
|
||
# 5. 安装 systemd 服务文件
|
||
# 6. 启用并启动服务
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 5: 配置 edge.yaml
|
||
|
||
```bash
|
||
sudo vi /opt/tianyan-edge/config/edge.yaml
|
||
```
|
||
|
||
关键配置项:
|
||
|
||
```yaml
|
||
device_uuid: 8541db9f77826e39605ef2c032f8fb93 # 设备唯一标识
|
||
edge_id: edge-demo-001
|
||
cloud_url: http://101.36.73.102:8004 # 云端 API
|
||
mqtt_broker: tcp://101.36.73.102:1883 # MQTT 消息
|
||
rtsp_urls:
|
||
- http://101.36.73.102:8080/rtp/34020000002000000003_34020000001310000001.live.flv
|
||
infer_socket: /tmp/edge-infer.sock
|
||
infer_fps: 2 # 推理帧率
|
||
conf_threshold: 0.2 # 检测阈值
|
||
ota_url: http://101.36.73.102:8087 # OTA 更新地址
|
||
version: 1.0.0
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 6: 启动服务
|
||
|
||
### 方式 A: systemd (推荐)
|
||
|
||
```bash
|
||
# 启动推理服务
|
||
sudo systemctl start edge-infer
|
||
|
||
# 启动边缘代理
|
||
sudo systemctl start edge-agent
|
||
|
||
# 设置开机自启
|
||
sudo systemctl enable edge-infer edge-agent
|
||
|
||
# 查看状态
|
||
sudo systemctl status edge-agent edge-infer
|
||
sudo journalctl -u edge-agent -f
|
||
```
|
||
|
||
### 方式 B: 手动启动
|
||
|
||
```bash
|
||
# 启动推理服务
|
||
bash -lc 'source /usr/local/Ascend/ascend-toolkit/set_env.sh && \
|
||
NAMES_FILE=/root/AI-tianyan/model/names.txt \
|
||
CONF_THRESHOLD=0.15 \
|
||
python3 python/infer_server.py' &
|
||
|
||
# 启动边缘代理
|
||
./build/edge-agent -config config/edge.yaml
|
||
```
|
||
|
||
### 方式 C: 一键脚本
|
||
|
||
```bash
|
||
bash start.sh
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 7: 验证运行
|
||
|
||
```bash
|
||
# 检查进程
|
||
ps aux | grep -E 'edge-agent|infer_server|ffmpeg'
|
||
|
||
# 检查网络连接
|
||
ss -tunp | grep 101.36.73.102
|
||
|
||
# 检查 NPU 状态
|
||
npu-smi info
|
||
|
||
# 检查日志
|
||
sudo journalctl -u edge-agent -n 50
|
||
sudo journalctl -u edge-infer -n 50
|
||
tail -f logs/agent.log logs/infer.log
|
||
```
|
||
|
||
---
|
||
|
||
## 步骤 8: 配置 Telegraf 指标上报 (可选)
|
||
|
||
```bash
|
||
# Telegraf 已安装 (v1.21.4)
|
||
# 配置文件: /etc/telegraf/telegraf.conf
|
||
# 已配置上报到 InfluxDB (101.36.73.102:18086)
|
||
|
||
# 重启 Telegraf 使配置生效
|
||
sudo systemctl restart telegraf
|
||
sudo systemctl status telegraf
|
||
```
|
||
|
||
---
|
||
|
||
## 常见问题排查
|
||
|
||
| 问题 | 解决方法 |
|
||
|---|---|
|
||
| `acl.init failed` | 确认 `set_env.sh` 已执行，检查 `LD_LIBRARY_PATH` |
|
||
| `model.om not found` | 重新执行 ATC 转换，确认路径正确 |
|
||
| 无法连接 MQTT | 检查 `mqtt_broker` 地址和 `edge_token` |
|
||
| 拉流失败 | 确认 ZLMediaKit 端口 8080 可达 |
|
||
| 编译失败 | `GOPROXY=https://goproxy.cn,direct go mod download` |
|
||
|
||
---
|
||
|
||
## 更新部署 (OTA)
|
||
|
||
```bash
|
||
# 云端推送新版本后，edge-agent 自动检测并下载
|
||
# 也可手动更新:
|
||
git pull
|
||
make build
|
||
make install
|
||
sudo systemctl restart edge-agent edge-infer
|
||
```
|