xiaoxia-saas/deploy/gpu_worker/scripts/setup-gpu-node.sh

#!/bin/bash
# GPU节点一键初始化脚本 - 在全新GPU机器上执行

set -e

echo "=== 1. 安装系统依赖 ==="
sudo apt-get update -qq
sudo apt-get install -y -qq python3 python3-pip python3-venv ffmpeg wget curl git

echo "=== 2. 创建目录 ==="
mkdir -p ~/projects/MuseTalk ~/projects/gpu-webhook /opt/xiaoxia-gpu-worker

echo "=== 3. 安装nvidia-container-toolkit（如需要Docker）==="
# 可选，当前不使用Docker，跳过
# distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
# curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
# curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
# sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
# sudo nvidia-ctk runtime configure --runtime=docker
# sudo systemctl restart docker

echo "=== 4. 拉取服务配置和脚本 ==="
REPO_URL="https://git.xiaoxiajianji.com/xiaoxia/xiaoxia-saas/raw/branch/develop/deploy/gpu_worker"
wget -q -O /tmp/musetalk-worker.service "$REPO_URL/musetalk-worker.service"
wget -q -O /tmp/gpu-poll.service "$REPO_URL/gpu-poll.service"
wget -q -O /tmp/gpu-poll.timer "$REPO_URL/gpu-poll.timer"
wget -q -O ~/projects/update-gpu-worker.sh "$REPO_URL/scripts/update-gpu-worker.sh"
wget -q -O ~/projects/gpu-webhook/poll_and_update.sh "$REPO_URL/scripts/poll_and_update.sh"
chmod +x ~/projects/update-gpu-worker.sh ~/projects/gpu-webhook/poll_and_update.sh

echo "=== 5. 安装systemd服务 ==="
sudo cp /tmp/musetalk-worker.service /etc/systemd/system/
sudo cp /tmp/gpu-poll.service /etc/systemd/system/
sudo cp /tmp/gpu-poll.timer /etc/systemd/system/

echo "=== 6. 配置sudo免密 ==="
sudo bash -c 'cat > /etc/sudoers.d/ying-gpu-update << EOF
ying ALL=(ALL) NOPASSWD: /bin/systemctl restart musetalk-worker
ying ALL=(ALL) NOPASSWD: /bin/systemctl restart xiaoxia-gpu-worker
ying ALL=(ALL) NOPASSWD: /bin/systemctl status musetalk-worker
ying ALL=(ALL) NOPASSWD: /bin/systemctl status xiaoxia-gpu-worker
ying ALL=(ALL) NOPASSWD: /bin/systemctl daemon-reload
ying ALL=(ALL) NOPASSWD: /usr/bin/journalctl
ying ALL=(ALL) NOPASSWD: /bin/cp
ying ALL=(ALL) NOPASSWD: /bin/chmod
ying ALL=(ALL) NOPASSWD: /usr/bin/tee
EOF'
sudo chmod 440 /etc/sudoers.d/ying-gpu-update

echo "=== 7. 首次拉取代码并启动服务 ==="
bash ~/projects/update-gpu-worker.sh
sudo systemctl daemon-reload
sudo systemctl enable musetalk-worker xiaoxia-gpu-worker gpu-poll.timer
sudo systemctl start musetalk-worker xiaoxia-gpu-worker gpu-poll.timer

echo "=== 完成！ ==="
echo "检查服务状态："
echo "  sudo systemctl status musetalk-worker"
echo "  sudo systemctl status xiaoxia-gpu-worker"
echo "  sudo systemctl status gpu-poll.timer"
echo "健康检查：curl http://127.0.0.1:7861/health"
echo "更新日志：tail -f ~/gpu-worker-update.log"
echo "轮询日志：tail -f ~/gpu-poll.log"