快速入门：使用 ARM 模板创建 Ubuntu Data Science Virtual Machine

项目
11/25/2024

本快速入门介绍如何使用 Azure 资源管理器模板（ARM 模板）创建 Ubuntu Data Science Virtual Machine (DSVM)。 Data Science Virtual Machine 是基于云的资源，预加载了一套数据科学和机器学习框架及工具。当部署在 GPU 驱动的计算资源上时，所有工具和库都配置为使用 GPU。

Azure 资源管理器模板是定义项目基础结构和配置的 JavaScript 对象表示法 (JSON) 文件。模板使用声明性语法。你可以在不编写用于创建部署的编程命令序列的情况下，描述预期部署。

如果环境满足先决条件，而且你知道如何使用 ARM 模板，请选择“部署到 Azure”按钮。这会在 Azure 门户中打开模板。

先决条件

Azure 订阅。如果没有 Azure 订阅，请在开始前创建一个试用版订阅。
若要从本地环境使用本文档中的 CLI 命令，需要使用 Azure CLI。

查看模板

可以在“Azure 快速入门模板”资源中找到本快速入门中使用的模板。

{
  "$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
  "contentVersion": "1.0.0.0",
  "metadata": {
    "_generator": {
      "name": "bicep",
      "version": "0.8.9.13224",
      "templateHash": "4895680407304578048"
    }
  },
  "parameters": {
    "adminUsername": {
      "type": "string",
      "metadata": {
        "description": "Username for Administrator Account"
      }
    },
    "vmName": {
      "type": "string",
      "defaultValue": "vmName",
      "metadata": {
        "description": "The name of you Virtual Machine."
      }
    },
    "location": {
      "type": "string",
      "defaultValue": "[resourceGroup().location]",
      "metadata": {
        "description": "Location for all resources."
      }
    },
    "cpu_gpu": {
      "type": "string",
      "defaultValue": "CPU-4GB",
      "allowedValues": [
        "CPU-4GB",
        "CPU-7GB",
        "CPU-8GB",
        "CPU-14GB",
        "CPU-16GB",
        "GPU-56GB"
      ],
      "metadata": {
        "description": "Choose between CPU or GPU processing"
      }
    },
    "virtualNetworkName": {
      "type": "string",
      "defaultValue": "vNet",
      "metadata": {
        "description": "Name of the VNET"
      }
    },
    "subnetName": {
      "type": "string",
      "defaultValue": "subnet",
      "metadata": {
        "description": "Name of the subnet in the virtual network"
      }
    },
    "networkSecurityGroupName": {
      "type": "string",
      "defaultValue": "SecGroupNet",
      "metadata": {
        "description": "Name of the Network Security Group"
      }
    },
    "authenticationType": {
      "type": "string",
      "defaultValue": "sshPublicKey",
      "allowedValues": [
        "sshPublicKey",
        "password"
      ],
      "metadata": {
        "description": "Type of authentication to use on the Virtual Machine. SSH key is recommended."
      }
    },
    "adminPasswordOrKey": {
      "type": "secureString",
      "metadata": {
        "description": "SSH Key or password for the Virtual Machine. SSH key is recommended."
      }
    }
  },
  "variables": {
    "networkInterfaceName": "[format('{0}NetInt', parameters('vmName'))]",
    "virtualMachineName": "[parameters('vmName')]",
    "publicIpAddressName": "[format('{0}PublicIP', parameters('vmName'))]",
    "subnetRef": "[resourceId('Microsoft.Network/virtualNetworks/subnets', parameters('virtualNetworkName'), parameters('subnetName'))]",
    "nsgId": "[resourceId('Microsoft.Network/networkSecurityGroups', parameters('networkSecurityGroupName'))]",
    "osDiskType": "StandardSSD_LRS",
    "storageAccountName": "[format('storage{0}', uniqueString(resourceGroup().id))]",
    "storageAccountType": "Standard_LRS",
    "storageAccountKind": "Storage",
    "vmSize": {
      "CPU-4GB": "Standard_B2s",
      "CPU-7GB": "Standard_D2s_v3",
      "CPU-8GB": "Standard_D2s_v3",
      "CPU-14GB": "Standard_D4s_v3",
      "CPU-16GB": "Standard_D4s_v3",
      "GPU-56GB": "Standard_NC6_Promo"
    },
    "linuxConfiguration": {
      "disablePasswordAuthentication": true,
      "ssh": {
        "publicKeys": [
          {
            "path": "[format('/home/{0}/.ssh/authorized_keys', parameters('adminUsername'))]",
            "keyData": "[parameters('adminPasswordOrKey')]"
          }
        ]
      }
    }
  },
  "resources": [
    {
      "type": "Microsoft.Network/networkInterfaces",
      "apiVersion": "2021-05-01",
      "name": "[variables('networkInterfaceName')]",
      "location": "[parameters('location')]",
      "properties": {
        "ipConfigurations": [
          {
            "name": "ipconfig1",
            "properties": {
              "subnet": {
                "id": "[variables('subnetRef')]"
              },
              "privateIPAllocationMethod": "Dynamic",
              "publicIPAddress": {
                "id": "[resourceId('Microsoft.Network/publicIPAddresses', variables('publicIpAddressName'))]"
              }
            }
          }
        ],
        "networkSecurityGroup": {
          "id": "[variables('nsgId')]"
        }
      },
      "dependsOn": [
        "[resourceId('Microsoft.Network/networkSecurityGroups', parameters('networkSecurityGroupName'))]",
        "[resourceId('Microsoft.Network/publicIPAddresses', variables('publicIpAddressName'))]",
        "[resourceId('Microsoft.Network/virtualNetworks', parameters('virtualNetworkName'))]"
      ]
    },
    {
      "type": "Microsoft.Network/networkSecurityGroups",
      "apiVersion": "2021-05-01",
      "name": "[parameters('networkSecurityGroupName')]",
      "location": "[parameters('location')]",
      "properties": {
        "securityRules": [
          {
            "name": "JupyterHub",
            "properties": {
              "priority": 1010,
              "protocol": "Tcp",
              "access": "Allow",
              "direction": "Inbound",
              "sourceAddressPrefix": "*",
              "sourcePortRange": "*",
              "destinationAddressPrefix": "*",
              "destinationPortRange": "8000"
            }
          },
          {
            "name": "RStudioServer",
            "properties": {
              "priority": 1020,
              "protocol": "Tcp",
              "access": "Allow",
              "direction": "Inbound",
              "sourceAddressPrefix": "*",
              "sourcePortRange": "*",
              "destinationAddressPrefix": "*",
              "destinationPortRange": "8787"
            }
          },
          {
            "name": "SSH",
            "properties": {
              "priority": 1030,
              "protocol": "Tcp",
              "access": "Allow",
              "direction": "Inbound",
              "sourceAddressPrefix": "*",
              "sourcePortRange": "*",
              "destinationAddressPrefix": "*",
              "destinationPortRange": "22"
            }
          }
        ]
      }
    },
    {
      "type": "Microsoft.Network/virtualNetworks",
      "apiVersion": "2021-05-01",
      "name": "[parameters('virtualNetworkName')]",
      "location": "[parameters('location')]",
      "properties": {
        "addressSpace": {
          "addressPrefixes": [
            "10.0.0.0/24"
          ]
        },
        "subnets": [
          {
            "name": "[parameters('subnetName')]",
            "properties": {
              "addressPrefix": "10.0.0.0/24",
              "privateEndpointNetworkPolicies": "Enabled",
              "privateLinkServiceNetworkPolicies": "Enabled"
            }
          }
        ]
      }
    },
    {
      "type": "Microsoft.Network/publicIPAddresses",
      "apiVersion": "2021-05-01",
      "name": "[variables('publicIpAddressName')]",
      "location": "[parameters('location')]",
      "sku": {
        "name": "Basic",
        "tier": "Regional"
      },
      "properties": {
        "publicIPAllocationMethod": "Dynamic"
      }
    },
    {
      "type": "Microsoft.Storage/storageAccounts",
      "apiVersion": "2021-08-01",
      "name": "[variables('storageAccountName')]",
      "location": "[parameters('location')]",
      "sku": {
        "name": "[variables('storageAccountType')]"
      },
      "kind": "[variables('storageAccountKind')]"
    },
    {
      "type": "Microsoft.Compute/virtualMachines",
      "apiVersion": "2021-11-01",
      "name": "[format('{0}-{1}', variables('virtualMachineName'), parameters('cpu_gpu'))]",
      "location": "[parameters('location')]",
      "properties": {
        "hardwareProfile": {
          "vmSize": "[variables('vmSize')[parameters('cpu_gpu')]]"
        },
        "storageProfile": {
          "osDisk": {
            "createOption": "FromImage",
            "managedDisk": {
              "storageAccountType": "[variables('osDiskType')]"
            }
          },
          "imageReference": {
            "publisher": "microsoft-dsvm",
            "offer": "ubuntu-1804",
            "sku": "1804-gen2",
            "version": "latest"
          }
        },
        "networkProfile": {
          "networkInterfaces": [
            {
              "id": "[resourceId('Microsoft.Network/networkInterfaces', variables('networkInterfaceName'))]"
            }
          ]
        },
        "osProfile": {
          "computerName": "[variables('virtualMachineName')]",
          "adminUsername": "[parameters('adminUsername')]",
          "adminPassword": "[parameters('adminPasswordOrKey')]",
          "linuxConfiguration": "[if(equals(parameters('authenticationType'), 'password'), json('null'), variables('linuxConfiguration'))]"
        }
      },
      "dependsOn": [
        "[resourceId('Microsoft.Network/networkInterfaces', variables('networkInterfaceName'))]"
      ]
    }
  ],
  "outputs": {
    "adminUsername": {
      "type": "string",
      "value": "[parameters('adminUsername')]"
    }
  }
}

该模板定义以下资源：

Microsoft.Network/networkInterfaces
Microsoft.Network/networkSecurityGroups
Microsoft.Network/virtualNetworks
Microsoft.Network/publicIPAddresses
Microsoft.Storage/storageAccounts
Microsoft.Compute/virtualMachines：创建基于云的虚拟机。在此模板中，虚拟机配置为运行 Ubuntu 的 Data Science Virtual Machine。

部署模板

若要通过 Azure CLI 使用该模板，请登录并选择你的订阅。有关详细信息，请访问使用 Azure CLI 登录。然后运行：

read -p "Enter the name of the resource group to create:" resourceGroupName &&
read -p "Enter the Azure location (e.g., chinaeast2):" location &&
read -p "Enter the authentication type (must be 'password' or 'sshPublicKey') :" authenticationType &&
read -p "Enter the login name for the administrator account (may not be 'admin'):" adminUsername &&
read -p "Enter administrator account secure string (value of password or ssh public key):" adminPasswordOrKey &&
templateUri="https://raw.githubusercontent.com/Azure/azure-quickstart-templates/master/application-workloads/datascience/vm-ubuntu-DSVM-GPU-or-CPU/azuredeploy.json" &&
az group create --name $resourceGroupName --location "$location" &&
az deployment group create --resource-group $resourceGroupName --template-uri $templateUri --parameters adminUsername=$adminUsername authenticationType=$authenticationType adminPasswordOrKey=$adminPasswordOrKey &&
echo "Press [ENTER] to continue ..." &&
read

运行此代码时，请输入：

要创建的包含 DSVM 和关联资源的资源组的名称
要在其中进行部署的 Azure 位置
要使用的身份验证类型（输入字符串 password 或 sshPublicKey）
管理员帐户的登录名（此值不能为 admin）
帐户的密码或 SSH 公钥的值

查看已部署的资源

若要显示 Data Science Virtual Machine：

转到 Azure 门户
登录
选择你刚才创建的资源组

这会显示资源组信息：

显示包含 DSVM 的基本资源组的屏幕截图

选择虚拟机资源转到其信息页面。可在此处找到 VM 的相关信息，包括连接详细信息。

清理资源

如果不想使用此虚拟机，则应该删除它。由于 DSVM 与其他资源（例如存储帐户）关联，因此可能需要删除所创建的整个资源组。使用门户可以删除资源组。选择“删除”按钮，然后确认选择。还可以从 CLI 中删除资源组，如下所示：

echo "Enter the Resource Group name:" &&
read resourceGroupName &&
az group delete --name $resourceGroupName &&
echo "Press [ENTER] to continue ..."

后续步骤

在本快速入门中，你已通过 ARM 模板创建了 Data Science Virtual Machine。

示例程序和 ML 演练