## 操作场景 NVIDIA\-Fabric Manager服务可以使多张显卡间通过NVSwitch互联。有关NVSwitch的更多信息，请参见[NVIDIA官网](https://www.nvidia.com/en-us/data-center/nvlink/)。 :::warning * 支持NVSwitch的实例请参见各[实例规格](https://www.volcengine.com/docs/6419/1088202)的特点介绍。如果未安装与GPU驱动版本对应的NVIDIA\-Fabric Manager服务，您将无法正常使用该类GPU实例。 * pni3l规格族下，配备2张及以上GPU卡的实例默认搭载NVLink设备。如遇`nvidia-fabricmanager.service`状态为`failed`属正常情况，不影响业务正常运行。 * 支持NVSwitch的实例升级GPU驱动的同时，还需同步升级Fabric Manager，否则将无法正常使用。[如何升级NVIDIA Tesla驱动？](https://www.volcengine.com/docs/6419/102549#.5aaC5L2V5Y2H57qnTlZJRElBIFRlc2xh6amx5Yqo77yf) * 云平台提供的公共镜像默认已安装NVIDIA\-Fabric Manager及devel软件包，您只需[启动NVIDIA-Fabric Manager](https://www.volcengine.com/docs/6419/73634)即可实现NVSwitch互联。 * 如果您使用未安装NVIDIA\-Fabric Manager的自定义镜像，购买了搭载支持NVSwitch的GPU实例后，则必须安装与GPU驱动版本对应的NVIDIA\-Fabric Manager软件包。 ::: ## 步骤一：安装NVIDIA\-Fabric Manager 您可以通过安装包或者源码两种方式安装NVIDIA\-Fabric Manager服务，下文以GPU驱动为470.57.02版本为例，为您介绍如何安装并启动NVIDIA\-Fabric Manager服务。如需下载其它版本，请将命令中的版本号替换为相应的GPU驱动版本号。您可以执行`nvidia-smi`命令，查看GPU驱动版本。 ### 方式一：通过安装包安装 * CentOS 8.x ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/rhel8/x86_64/nvidia-fabric-manager-470.57.02-1.x86_64.rpm rpm -ivh nvidia-fabric-manager-470.57.02-1.x86_64.rpm ``` * CentOS 7.x ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/rhel7/x86_64/nvidia-fabric-manager-470.57.02-1.x86_64.rpm rpm -ivh nvidia-fabric-manager-470.57.02-1.x86_64.rpm ``` * Ubuntu 20.04 ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/x86_64/nvidia-fabricmanager-470_470.57.02-1_amd64.deb dpkg -i nvidia-fabricmanager-470_470.57.02-1_amd64.deb ``` * Ubuntu 18.04 ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64/nvidia-fabricmanager-470_470.57.02-1_amd64.deb dpkg -i nvidia-fabricmanager-470_470.57.02-1_amd64.deb ``` * Debain 10、veLinux 1.0 ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/debian10/x86_64/nvidia-fabricmanager-470_470.57.02-1_amd64.deb dpkg -i nvidia-fabricmanager-470_470.57.02-1_amd64.deb ``` ### 方式二：通过源安装 * CentOS 8.x ```Plain Text dnf config-manager --add-repo http://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo dnf module enable -y nvidia-driver:470 dnf install -y nvidia-fabric-manager-0:470.57.02-1 ``` * CentOS 7.x ```Plain Text yum -y install yum-utils yum-config-manager --add-repo https://developer.download.nvidia.cn/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo yum install -y nvidia-fabric-manager-470.57.02-1 ``` * Ubuntu 20.04 ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600 wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/x86_64/7fa2af80.pub apt-key add 7fa2af80.pub rm 7fa2af80.pub echo "deb http://developer.download.nvidia.cn/compute/cuda/repos/ubuntu2004/x86_64 /" | tee /etc/apt/sources.list.d/cuda.list apt-get update apt-get -y install nvidia-fabricmanager-470=470.57.02-1 ``` * Ubuntu 18.04 ```Plain Text wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600 wget https://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub apt-key add 7fa2af80.pub rm 7fa2af80.pub echo "deb http://developer.download.nvidia.cn/compute/cuda/repos/ubuntu1804/x86_64 /" | tee /etc/apt/sources.list.d/cuda.list apt-get update apt-get -y install nvidia-fabricmanager-470=470.57.02-1 ``` ## 步骤二：安装Nvidia\-Fabric\-Manager\-devel * CentOS 7.x/8.x `yum install nvidia-fabric-manager-devel-470.57.02-1 -y` * Ubuntu 20.04/18.04、Debain 10、veLinux 1.0 `dpkg -i nvidia-fabric-manager-devel-470.57.02-1_amd64.deb` ## 步骤三：启动NVIDIA\-Fabric Manager 1. 执行如下命令启动Fabric Manager服务。 `sudo systemctl start nvidia-fabricmanager` 2. 执行如下命令查看Fabric Manager服务是否正常启动，回显`active（running）`表示启动成功。 `sudo systemctl status nvidia-fabricmanager` 3. 执行如下命令配置Fabric Manager服务随实例开机自启动。 `sudo systemctl enable nvidia-fabricmanager`

GPU云服务器