Autoscale using guest metrics in a Linux scale set template
There are two broad types of metrics in Azure that are gathered from VMs and scale sets: Host metrics and Guest metrics. At a high level, if you would like to use standard CPU, disk, and network metrics, then host metrics are a good fit. If, however, you need a larger selection of metrics, then guest metrics should be looked into.
Host metrics do not require additional setup because they are collected by the host VM, whereas guest metrics require you to install the Windows Azure Diagnostics extension or the Linux Azure Diagnostics extension in the guest VM. One common reason to use guest metrics instead of host metrics is that guest metrics provide a larger selection of metrics than host metrics. One such example is memory-consumption metrics, which are only available via guest metrics. The supported host metrics are listed here, and commonly used guest metrics are listed here. This article shows how to modify the basic viable scale set template to use autoscale rules based on guest metrics for Linux scale sets.
Change the template definition
In a previous article we had created a basic scale set template. We will now use that earlier template and modify it to create a template that deploys a Linux scale set with guest metric based autoscale.
First, add parameters for storageAccountName and storageAccountSasToken. The diagnostics agent stores metric data in a table in this storage account. As of the Linux Diagnostics Agent version 3.0, using a storage access key is no longer supported. Instead, use a SAS Token.
},
"adminPassword": {
"type": "securestring"
+ },
+ "storageAccountName": {
+ "type": "string"
+ },
+ "storageAccountSasToken": {
+ "type": "securestring"
}
},
Next, modify the scale set extensionProfile to include the diagnostics extension. In this configuration, specify the resource ID of the scale set to collect metrics from, as well as the storage account and SAS token to use to store the metrics. Specify how frequently the metrics are aggregated (in this case, every minute) and which metrics to track (in this case, percent used memory). For more detailed information on this configuration and metrics other than percent used memory, see this documentation.
}
}
]
+ },
+ "extensionProfile": {
+ "extensions": [
+ {
+ "name": "LinuxDiagnosticExtension",
+ "properties": {
+ "publisher": "Microsoft.Azure.Diagnostics",
+ "type": "LinuxDiagnostic",
+ "typeHandlerVersion": "3.0",
+ "settings": {
+ "StorageAccount": "[parameters('storageAccountName')]",
+ "ladCfg": {
+ "diagnosticMonitorConfiguration": {
+ "performanceCounters": {
+ "sinks": "WADMetricJsonBlob",
+ "performanceCounterConfiguration": [
+ {
+ "unit": "percent",
+ "type": "builtin",
+ "class": "memory",
+ "counter": "percentUsedMemory",
+ "counterSpecifier": "/builtin/memory/percentUsedMemory",
+ "condition": "IsAggregate=TRUE"
+ }
+ ]
+ },
+ "metrics": {
+ "metricAggregation": [
+ {
+ "scheduledTransferPeriod": "PT1M"
+ }
+ ],
+ "resourceId": "[resourceId('Microsoft.Compute/virtualMachineScaleSets', 'myScaleSet')]"
+ }
+ }
+ }
+ },
+ "protectedSettings": {
+ "storageAccountName": "[parameters('storageAccountName')]",
+ "storageAccountSasToken": "[parameters('storageAccountSasToken')]",
+ "sinksConfig": {
+ "sink": [
+ {
+ "name": "WADMetricJsonBlob",
+ "type": "JsonBlob"
+ }
+ ]
+ }
+ }
+ }
+ }
+ ]
}
}
}
Finally, add an autoscaleSettings resource to configure autoscale based on these metrics. This resource has a dependsOn clause that references the scale set to ensure that the scale set exists before attempting to autoscale it. If you choose a different metric to autoscale on, you would use the counterSpecifier from the diagnostics extension configuration as the metricName in the autoscale configuration. For more information on autoscale configuration, see the autoscale best practices and the Azure Monitor REST API reference documentation.
+ },
+ {
+ "type": "Microsoft.Insights/autoscaleSettings",
+ "apiVersion": "2015-04-01",
+ "name": "guestMetricsAutoscale",
+ "location": "[resourceGroup().location]",
+ "dependsOn": [
+ "Microsoft.Compute/virtualMachineScaleSets/myScaleSet"
+ ],
+ "properties": {
+ "name": "guestMetricsAutoscale",
+ "targetResourceUri": "[resourceId('Microsoft.Compute/virtualMachineScaleSets', 'myScaleSet')]",
+ "enabled": true,
+ "profiles": [
+ {
+ "name": "Profile1",
+ "capacity": {
+ "minimum": "1",
+ "maximum": "10",
+ "default": "3"
+ },
+ "rules": [
+ {
+ "metricTrigger": {
+ "metricName": "/builtin/memory/percentUsedMemory",
+ "metricNamespace": "",
+ "metricResourceUri": "[resourceId('Microsoft.Compute/virtualMachineScaleSets', 'myScaleSet')]",
+ "timeGrain": "PT1M",
+ "statistic": "Average",
+ "timeWindow": "PT5M",
+ "timeAggregation": "Average",
+ "operator": "GreaterThan",
+ "threshold": 60
+ },
+ "scaleAction": {
+ "direction": "Increase",
+ "type": "ChangeCount",
+ "value": "1",
+ "cooldown": "PT1M"
+ }
+ },
+ {
+ "metricTrigger": {
+ "metricName": "/builtin/memory/percentUsedMemory",
+ "metricNamespace": "",
+ "metricResourceUri": "[resourceId('Microsoft.Compute/virtualMachineScaleSets', 'myScaleSet')]",
+ "timeGrain": "PT1M",
+ "statistic": "Average",
+ "timeWindow": "PT5M",
+ "timeAggregation": "Average",
+ "operator": "LessThan",
+ "threshold": 30
+ },
+ "scaleAction": {
+ "direction": "Decrease",
+ "type": "ChangeCount",
+ "value": "1",
+ "cooldown": "PT1M"
+ }
+ }
+ ]
+ }
+ ]
+ }
}
]
}
Next steps
You can deploy the preceding template by following the Azure Resource Manager documentation.
You can start this tutorial series from the basic scale set template article.
You can see how to modify the basic scale set template to deploy the scale set into an existing virtual network.
You can see how to modify the basic scale set template to deploy the scale set with a custom image.
You can see how to modify the basic scale set template to deploy a Linux scale set with guest-based autoscale.
For more information about scale sets, refer to the scale set overview page.
Feedback
Coming soon: Throughout 2024 we will be phasing out GitHub Issues as the feedback mechanism for content and replacing it with a new feedback system. For more information see: https://aka.ms/ContentUserFeedback.
Submit and view feedback for