2020-03-09 06:29:10 +01:00
|
|
|
{
|
|
|
|
|
"$schema": "https://schema.management.azure.com/schemas/2019-04-01/deploymentTemplate.json#",
|
|
|
|
|
"contentVersion": "1.0.0.0",
|
|
|
|
|
"parameters": {
|
|
|
|
|
"adminUsername": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "ubuntu",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Username for the Virtual Machine."
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"publicKey": {
|
|
|
|
|
"type": "securestring",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "SSH Key for the Virtual Machine"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"adminPassword": {
|
|
|
|
|
"type": "securestring",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Password for the Virtual Machine and JupyterLab"
|
|
|
|
|
}
|
|
|
|
|
},
|
2020-03-12 02:46:11 +01:00
|
|
|
"headNodeSize": {
|
2020-03-09 06:29:10 +01:00
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "Standard_D2s_v3",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "The size of the head-node Virtual Machine"
|
|
|
|
|
}
|
|
|
|
|
},
|
2020-03-12 02:46:11 +01:00
|
|
|
"headNodePriority": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "Regular",
|
|
|
|
|
"allowedValues": ["Regular", "Low", "Spot"],
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Use Azure Spot instance for worker nodes"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"workerNodeSize": {
|
2020-03-09 06:29:10 +01:00
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "Standard_D2s_v3",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "The size of the worker node Virtual Machine"
|
|
|
|
|
}
|
|
|
|
|
},
|
2020-03-12 02:46:11 +01:00
|
|
|
"workerNodePriority": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "Spot",
|
|
|
|
|
"allowedValues": ["Regular", "Low", "Spot"],
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Use Azure Spot instance for worker nodes"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"workerInitial": {
|
2020-03-09 06:29:10 +01:00
|
|
|
"type": "int",
|
|
|
|
|
"defaultValue": 1,
|
|
|
|
|
"minValue": 0,
|
2020-10-27 18:27:18 -04:00
|
|
|
"maxValue": 1000,
|
2020-03-09 06:29:10 +01:00
|
|
|
"metadata": {
|
|
|
|
|
"description": "Initial number of worker nodes"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"workerMin": {
|
|
|
|
|
"type": "int",
|
|
|
|
|
"defaultValue": 1,
|
|
|
|
|
"minValue": 0,
|
2020-10-27 18:27:18 -04:00
|
|
|
"maxValue": 1000,
|
2020-03-09 06:29:10 +01:00
|
|
|
"metadata": {
|
|
|
|
|
"description": "Minimum number of worker nodes"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"workerMax": {
|
|
|
|
|
"type": "int",
|
|
|
|
|
"defaultValue": 1,
|
|
|
|
|
"minValue": 0,
|
2020-10-27 18:27:18 -04:00
|
|
|
"maxValue": 1000,
|
2020-03-09 06:29:10 +01:00
|
|
|
"metadata": {
|
|
|
|
|
"description": "Maximum number of worker nodes"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"condaEnv": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "py37_tensorflow",
|
|
|
|
|
"allowedValues": [
|
|
|
|
|
"azureml_py36_automl",
|
|
|
|
|
"azureml_py36_pytorch",
|
|
|
|
|
"azureml_py36_tensorflow",
|
|
|
|
|
"py37_default",
|
|
|
|
|
"py37_pytorch",
|
|
|
|
|
"py37_tensorflow"
|
|
|
|
|
],
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Conda environment to select (installed on DSVM)"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"PythonPackages": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"defaultValue": "ray[rllib] gym[atari]",
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Python packages to install (space separated)"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"PublicWebUI": {
|
|
|
|
|
"type": "bool",
|
|
|
|
|
"defaultValue": true,
|
|
|
|
|
"metadata": {
|
|
|
|
|
"description": "Open port for web UI"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"variables": {
|
2020-03-09 17:49:07 +01:00
|
|
|
"azureScriptInitUrl": "https://raw.githubusercontent.com/ray-project/ray/master/doc/azure/azure-init.sh",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[resourceGroup().location]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"vmName": "ray-node",
|
|
|
|
|
"subnetWorkers": "10.32.0.0/16",
|
|
|
|
|
"subnetHead": "10.33.0.0/16",
|
|
|
|
|
"publicIpAddressName": "[concat(variables('vmName'), '-ip' )]",
|
|
|
|
|
"networkIpConfig": "[guid(resourceGroup().id, variables('vmName'))]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"subnetName": "ray-subnet",
|
|
|
|
|
"subnetHeadName": "ray-subnet-head",
|
|
|
|
|
"subnetRef": "[resourceId('Microsoft.Network/virtualNetworks/subnets', variables('vNetName'), variables('subnetName'))]",
|
|
|
|
|
"subnetHeadRef": "[resourceId('Microsoft.Network/virtualNetworks/subnets', variables('vNetName'), variables('subnetHeadName'))]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"osDiskType": "Standard_LRS",
|
|
|
|
|
"vmNameHead": "[concat(variables('vmName'), '-head')]",
|
|
|
|
|
"vmNameWorker": "[concat(variables('vmName'), '-workers')]",
|
|
|
|
|
"networkInterfaceName": "[concat(variables('vmName'), '-nic')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"networkSecurityGroupName": "ray-nsg",
|
|
|
|
|
"vNetName": "ray-vnet",
|
2020-03-09 06:29:10 +01:00
|
|
|
"subnetNetwork": "[split(variables('subnetHead'), '/')[0]]",
|
|
|
|
|
"headInternalIP": "[concat(substring(variables('subnetNetwork'), 0, lastIndexOf(variables('subnetNetwork'), '.')), '.5')]",
|
|
|
|
|
"imagePublisher": "microsoft-dsvm",
|
|
|
|
|
"imageOffer": "ubuntu-1804",
|
|
|
|
|
"imageSku": "1804",
|
2020-10-27 18:27:18 -04:00
|
|
|
"imageVersion": "20.07.06"
|
2020-03-09 06:29:10 +01:00
|
|
|
},
|
|
|
|
|
"resources": [
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Network/networkSecurityGroups",
|
|
|
|
|
"apiVersion": "2019-02-01",
|
2020-10-27 18:27:18 -04:00
|
|
|
"name": "[variables('networkSecurityGroupName')]",
|
|
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"properties": {
|
|
|
|
|
"securityRules": [
|
|
|
|
|
{
|
|
|
|
|
"name": "SSH",
|
|
|
|
|
"properties": {
|
|
|
|
|
"priority": 1000,
|
|
|
|
|
"protocol": "TCP",
|
|
|
|
|
"access": "Allow",
|
|
|
|
|
"direction": "Inbound",
|
|
|
|
|
"sourceAddressPrefix": "*",
|
|
|
|
|
"sourcePortRange": "*",
|
|
|
|
|
"destinationAddressPrefix": "*",
|
|
|
|
|
"destinationPortRange": "22"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "JupyterLab",
|
|
|
|
|
"properties": {
|
|
|
|
|
"priority": 1001,
|
|
|
|
|
"protocol": "TCP",
|
|
|
|
|
"access": "Allow",
|
|
|
|
|
"direction": "Inbound",
|
|
|
|
|
"sourceAddressPrefix": "*",
|
|
|
|
|
"sourcePortRange": "*",
|
|
|
|
|
"destinationAddressPrefix": "*",
|
|
|
|
|
"destinationPortRange": "8000"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "RayWebUI",
|
|
|
|
|
"properties": {
|
|
|
|
|
"priority": 1002,
|
|
|
|
|
"protocol": "TCP",
|
|
|
|
|
"access": "[if(parameters('PublicWebUI'), 'Allow', 'Deny')]",
|
|
|
|
|
"direction": "Inbound",
|
|
|
|
|
"sourceAddressPrefix": "*",
|
|
|
|
|
"sourcePortRange": "*",
|
|
|
|
|
"destinationAddressPrefix": "*",
|
|
|
|
|
"destinationPortRange": "8265"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"name": "TensorBoard",
|
|
|
|
|
"properties": {
|
|
|
|
|
"priority": 1003,
|
|
|
|
|
"protocol": "TCP",
|
|
|
|
|
"access": "[if(parameters('PublicWebUI'), 'Allow', 'Deny')]",
|
|
|
|
|
"direction": "Inbound",
|
|
|
|
|
"sourceAddressPrefix": "*",
|
|
|
|
|
"sourcePortRange": "*",
|
|
|
|
|
"destinationAddressPrefix": "*",
|
|
|
|
|
"destinationPortRange": "6006"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Network/virtualNetworks",
|
|
|
|
|
"apiVersion": "2019-11-01",
|
2020-10-27 18:27:18 -04:00
|
|
|
"name": "[variables('vNetName')]",
|
|
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"properties": {
|
|
|
|
|
"addressSpace": {
|
|
|
|
|
"addressPrefixes": [
|
|
|
|
|
"[variables('subnetHead')]",
|
|
|
|
|
"[variables('subnetWorkers')]"
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
"subnets": [
|
|
|
|
|
{
|
2020-10-27 18:27:18 -04:00
|
|
|
"name": "[variables('subnetName')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"properties": {
|
|
|
|
|
"addressPrefix": "[variables('subnetWorkers')]"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
2020-10-27 18:27:18 -04:00
|
|
|
"name": "[variables('subnetHeadName')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"properties": {
|
|
|
|
|
"addressPrefix": "[variables('subnetHead')]"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Network/publicIpAddresses",
|
|
|
|
|
"apiVersion": "2019-02-01",
|
|
|
|
|
"name": "[variables('publicIpAddressName')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"properties": {
|
|
|
|
|
"publicIpAllocationMethod": "Static",
|
|
|
|
|
"publicIPAddressVersion": "IPv4"
|
|
|
|
|
},
|
|
|
|
|
"sku": {
|
|
|
|
|
"name": "Basic",
|
|
|
|
|
"tier": "Regional"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Network/networkInterfaces",
|
2020-10-27 18:27:18 -04:00
|
|
|
"apiVersion": "2020-06-01",
|
2020-03-09 06:29:10 +01:00
|
|
|
"name": "[variables('networkInterfaceName')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"dependsOn": [
|
2020-10-27 18:27:18 -04:00
|
|
|
"[resourceId('Microsoft.Network/publicIpAddresses', variables('publicIpAddressName'))]",
|
|
|
|
|
"[resourceId('Microsoft.Network/networkSecurityGroups', variables('networkSecurityGroupName'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
],
|
|
|
|
|
"properties": {
|
|
|
|
|
"ipConfigurations": [
|
|
|
|
|
{
|
|
|
|
|
"name": "[variables('networkIpConfig')]",
|
|
|
|
|
"properties": {
|
|
|
|
|
"subnet": {
|
|
|
|
|
"id": "[variables('subnetHeadRef')]"
|
|
|
|
|
},
|
|
|
|
|
"privateIPAllocationMethod": "Static",
|
|
|
|
|
"privateIPAddress": "[variables('headInternalIP')]",
|
|
|
|
|
"publicIpAddress": {
|
|
|
|
|
"id": "[resourceId('Microsoft.Network/publicIPAddresses', variables('publicIPAddressName'))]"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
],
|
|
|
|
|
"networkSecurityGroup": {
|
2020-10-27 18:27:18 -04:00
|
|
|
"id": "[resourceId('Microsoft.Network/networkSecurityGroups', variables('networkSecurityGroupName'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Compute/virtualMachines",
|
2020-10-27 18:27:18 -04:00
|
|
|
"apiVersion": "2020-06-01",
|
2020-03-09 06:29:10 +01:00
|
|
|
"name": "[variables('vmNameHead')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"dependsOn": [
|
2020-10-27 18:27:18 -04:00
|
|
|
"[resourceId('Microsoft.Network/networkInterfaces', variables('networkInterfaceName'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
],
|
|
|
|
|
"properties": {
|
|
|
|
|
"hardwareProfile": {
|
2020-03-12 02:46:11 +01:00
|
|
|
"vmSize": "[parameters('headNodeSize')]"
|
2020-03-09 06:29:10 +01:00
|
|
|
},
|
2020-03-12 02:46:11 +01:00
|
|
|
"priority": "[parameters('headNodePriority')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"storageProfile": {
|
|
|
|
|
"osDisk": {
|
|
|
|
|
"createOption": "fromImage",
|
|
|
|
|
"managedDisk": {
|
|
|
|
|
"storageAccountType": "[variables('osDiskType')]"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"imageReference": {
|
|
|
|
|
"publisher": "[variables('imagePublisher')]",
|
|
|
|
|
"offer": "[variables('imageOffer')]",
|
|
|
|
|
"sku": "[variables('imageSku')]",
|
|
|
|
|
"version": "[variables('imageVersion')]"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"networkProfile": {
|
|
|
|
|
"networkInterfaces": [
|
|
|
|
|
{
|
|
|
|
|
"id": "[resourceId('Microsoft.Network/networkInterfaces', variables('networkInterfaceName'))]"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
"osProfile": {
|
|
|
|
|
"computerName": "[variables('vmNameHead')]",
|
|
|
|
|
"adminUsername": "[parameters('adminUsername')]",
|
|
|
|
|
"adminPassword": "[parameters('adminPassword')]",
|
|
|
|
|
"linuxConfiguration": {
|
|
|
|
|
"disablePasswordAuthentication": false,
|
|
|
|
|
"ssh": {
|
|
|
|
|
"publicKeys": [
|
|
|
|
|
{
|
|
|
|
|
"path": "[concat('/home/', parameters('adminUsername'), '/.ssh/authorized_keys')]",
|
|
|
|
|
"keyData": "[parameters('publicKey')]"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"resources": [
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Compute/virtualMachines/extensions",
|
|
|
|
|
"name": "[concat(variables('vmNameHead'), '/HeadNodeInitScript')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"apiVersion": "2020-06-01",
|
|
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"dependsOn": [
|
2020-10-27 18:27:18 -04:00
|
|
|
"[resourceId('Microsoft.Compute/virtualMachines', variables('vmNameHead'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
],
|
|
|
|
|
"properties": {
|
|
|
|
|
"publisher": "Microsoft.Azure.Extensions",
|
|
|
|
|
"type": "CustomScript",
|
|
|
|
|
"typeHandlerVersion": "2.1",
|
|
|
|
|
"autoUpgradeMinorVersion": true,
|
|
|
|
|
"settings": {
|
|
|
|
|
"commandToExecute": "[concat('sh azure-init.sh ', parameters('adminUsername'), ' ', parameters('condaEnv'), ' \"', parameters('PythonPackages'), '\" ignore head 2>&1 >/var/log/ray-head.log')]",
|
|
|
|
|
"fileUris": [
|
|
|
|
|
"[variables('azureScriptInitUrl')]"
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Compute/virtualMachineScaleSets",
|
|
|
|
|
"name": "[variables('vmNameWorker')]",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"apiVersion": "2019-07-01",
|
|
|
|
|
"dependsOn": [
|
2020-10-27 18:27:18 -04:00
|
|
|
"[resourceId('Microsoft.Network/virtualNetworks', variables('vNetName'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
],
|
|
|
|
|
"sku": {
|
2020-03-12 02:46:11 +01:00
|
|
|
"name": "[parameters('workerNodeSize')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"tier": "Standard",
|
|
|
|
|
"capacity": "[parameters('workerInitial')]"
|
|
|
|
|
},
|
|
|
|
|
"properties": {
|
|
|
|
|
"upgradePolicy": {
|
|
|
|
|
"mode": "Manual"
|
|
|
|
|
},
|
|
|
|
|
"virtualMachineProfile": {
|
2020-03-12 02:46:11 +01:00
|
|
|
"priority": "[parameters('workerNodePriority')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"storageProfile": {
|
|
|
|
|
"osDisk": {
|
|
|
|
|
"createOption": "fromImage",
|
|
|
|
|
"managedDisk": {
|
|
|
|
|
"storageAccountType": "[variables('osDiskType')]"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"imageReference": {
|
|
|
|
|
"publisher": "[variables('imagePublisher')]",
|
|
|
|
|
"offer": "[variables('imageOffer')]",
|
|
|
|
|
"sku": "[variables('imageSku')]",
|
|
|
|
|
"version": "[variables('imageVersion')]"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"osProfile": {
|
|
|
|
|
"computerNamePrefix": "[variables('vmNameWorker')]",
|
|
|
|
|
"adminUsername": "[parameters('adminUsername')]",
|
|
|
|
|
"adminPassword": "[parameters('adminPassword')]",
|
|
|
|
|
"linuxConfiguration": {
|
|
|
|
|
"disablePasswordAuthentication": false,
|
|
|
|
|
"ssh": {
|
|
|
|
|
"publicKeys": [
|
|
|
|
|
{
|
|
|
|
|
"path": "[concat('/home/', parameters('adminUsername'), '/.ssh/authorized_keys')]",
|
|
|
|
|
"keyData": "[parameters('publicKey')]"
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
"networkProfile": {
|
|
|
|
|
"networkInterfaceConfigurations": [
|
|
|
|
|
{
|
|
|
|
|
"name": "[concat(variables('vmNameWorker'),'-nic')]",
|
|
|
|
|
"properties": {
|
|
|
|
|
"primary": true,
|
|
|
|
|
"ipConfigurations": [
|
|
|
|
|
{
|
|
|
|
|
"name": "worker-ip-config",
|
|
|
|
|
"properties": {
|
|
|
|
|
"subnet": {
|
|
|
|
|
"id": "[variables('subnetRef')]"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
},
|
|
|
|
|
"extensionProfile": {
|
|
|
|
|
"extensions": [
|
|
|
|
|
{
|
|
|
|
|
"name": "RayWorkerInitScript",
|
|
|
|
|
"properties": {
|
|
|
|
|
"publisher": "Microsoft.Azure.Extensions",
|
|
|
|
|
"type": "CustomScript",
|
|
|
|
|
"typeHandlerVersion": "2.1",
|
|
|
|
|
"autoUpgradeMinorVersion": true,
|
|
|
|
|
"settings": {
|
|
|
|
|
"commandToExecute": "[concat('sh azure-init.sh ', parameters('adminUsername'), ' ', parameters('condaEnv'), ' \"', parameters('PythonPackages'), '\" ', variables('headInternalIP'), ' worker 2>&1 >/var/log/ray-worker.log')]",
|
|
|
|
|
"fileUris": [
|
|
|
|
|
"[variables('azureScriptInitUrl')]"
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"type": "Microsoft.Insights/autoscaleSettings",
|
|
|
|
|
"apiVersion": "2015-04-01",
|
|
|
|
|
"name": "cpuautoscale",
|
2020-10-27 18:27:18 -04:00
|
|
|
"location": "[variables('location')]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"dependsOn": [
|
2020-10-27 18:27:18 -04:00
|
|
|
"[resourceId('Microsoft.Compute/virtualMachineScaleSets', variables('vmNameWorker'))]"
|
2020-03-09 06:29:10 +01:00
|
|
|
],
|
|
|
|
|
"properties": {
|
|
|
|
|
"name": "cpuautoscale",
|
2020-10-27 18:27:18 -04:00
|
|
|
"targetResourceUri": "[concat(resourceGroup().id, '/providers/Microsoft.Compute/virtualMachineScaleSets/', variables('vmNameWorker'))]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"enabled": true,
|
|
|
|
|
"profiles": [
|
|
|
|
|
{
|
|
|
|
|
"name": "Profile1",
|
|
|
|
|
"capacity": {
|
|
|
|
|
"minimum": "[parameters('workerMin')]",
|
|
|
|
|
"maximum": "[parameters('workerMax')]",
|
|
|
|
|
"default": "[parameters('workerInitial')]"
|
|
|
|
|
},
|
|
|
|
|
"rules": [
|
|
|
|
|
{
|
|
|
|
|
"metricTrigger": {
|
|
|
|
|
"metricName": "Percentage CPU",
|
2020-10-27 18:27:18 -04:00
|
|
|
"metricResourceUri": "[concat(resourceGroup().id, '/providers/Microsoft.Compute/virtualMachineScaleSets/', variables('vmNameWorker'))]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"timeGrain": "PT1M",
|
|
|
|
|
"statistic": "Average",
|
|
|
|
|
"timeWindow": "PT10M",
|
|
|
|
|
"timeAggregation": "Average",
|
|
|
|
|
"operator": "GreaterThan",
|
|
|
|
|
"threshold": 80
|
|
|
|
|
},
|
|
|
|
|
"scaleAction": {
|
|
|
|
|
"direction": "Increase",
|
|
|
|
|
"type": "ChangeCount",
|
|
|
|
|
"value": "1",
|
|
|
|
|
"cooldown": "PT5M"
|
|
|
|
|
}
|
|
|
|
|
},
|
|
|
|
|
{
|
|
|
|
|
"metricTrigger": {
|
|
|
|
|
"metricName": "Percentage CPU",
|
2020-10-27 18:27:18 -04:00
|
|
|
"metricResourceUri": "[concat(resourceGroup().id, '/providers/Microsoft.Compute/virtualMachineScaleSets/', variables('vmNameWorker'))]",
|
2020-03-09 06:29:10 +01:00
|
|
|
"timeGrain": "PT1M",
|
|
|
|
|
"statistic": "Average",
|
|
|
|
|
"timeWindow": "PT30M",
|
|
|
|
|
"timeAggregation": "Average",
|
|
|
|
|
"operator": "LessThan",
|
|
|
|
|
"threshold": 20
|
|
|
|
|
},
|
|
|
|
|
"scaleAction": {
|
|
|
|
|
"direction": "Decrease",
|
|
|
|
|
"type": "ChangeCount",
|
|
|
|
|
"value": "1",
|
|
|
|
|
"cooldown": "PT5M"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
]
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
],
|
|
|
|
|
"outputs": {
|
|
|
|
|
"JupyterLabURL": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"value": "[concat('https://', reference(variables('publicIpAddressName')).ipAddress, ':8000')]"
|
|
|
|
|
},
|
|
|
|
|
"SSH": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"value": "[concat('ssh -t -L 8265:localhost:8265 -L 8888:localhost:8888 ', parameters('adminUsername'),'@', reference(variables('publicIpAddressName')).ipAddress)]"
|
|
|
|
|
},
|
|
|
|
|
"RayWebUIURL": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"value": "[concat('http://', reference(variables('publicIpAddressName')).ipAddress, ':8265')]",
|
|
|
|
|
"condition": "[parameters('PublicWebUI')]"
|
|
|
|
|
},
|
|
|
|
|
"TensorBoard": {
|
|
|
|
|
"type": "string",
|
|
|
|
|
"value": "[concat('http://', reference(variables('publicIpAddressName')).ipAddress, ':6006')]",
|
|
|
|
|
"condition": "[parameters('PublicWebUI')]"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
