Collect client performance counters by using VM insights

Host metrics describe the Azure resources allocated to your VM. To inspect performance inside the guest operating system, enable Azure Monitor Agent (AMA) and a performance data collection rule (DCR).

This exercise uses logs-based VM insights monitoring. Its predefined performance data is stored in the InsightsMetrics table in the Log Analytics workspace you prepared. You analyze it with VM insights Performance views and Kusto Query Language (KQL).

Enable logs-based performance monitoring

1. Open monitored-linux-vm in the Azure portal and select Monitor from its menu.

2. Select Configure to open Configure monitor.

3. Select the logs-based collection option and clear OpenTelemetry metrics for this exercise.

4. Select Customize infrastructure monitoring. Select the Log Analytics workspace created earlier rather than accepting an unspecified default workspace.

5. Allow the portal to create the performance DCR. If you select an existing DCR instead, confirm that it's configured for logs-based VM insights performance collection, sends data to the intended workspace, and doesn't enable process/dependency mapping.

6. Leave process/dependency mapping disabled. Don't install Dependency Agent.

7. Recommended host alerts were configured in the preceding VM exercise. Clear Enable recommended alerts in this onboarding operation to avoid creating an additional set.

8. Review the configuration. Confirm logs-based collection, the intended Log Analytics workspace, and the selected VM. Record the DCR name and its actual resource group for cleanup.

   

9. Select Review + Enable, then Enable.

Allow several minutes for installation, configuration, and initial data delivery. On the VM, open Settings > Extensions + applications and confirm that AzureMonitorLinuxAgent reports successful provisioning.

View and confirm performance data

1. In the VM menu, expand Monitoring, select Insights, and open Performance.

2. Select a time range that includes the period after onboarding. Inspect the CPU, memory, and logical-disk performance views.

   

3. Open Logs from the VM menu. Confirm the query scope is this VM.

4. If Logs opens in Simple mode, switch to KQL mode. Close the example-query dialog if necessary.

5. Run the following query:

   InsightsMetrics
   | where TimeGenerated > ago(30m)
   | where Origin == "vm.azm.ms"
   | summarize Samples = count() by Namespace, Name
   | order by Namespace asc, Name asc
   

The results should list recently collected performance measurements for this VM. The Performance charts and recent InsightsMetrics records are the success criteria; a VM Overview widget or an agent heartbeat alone isn't proof of performance collection.

For example, chart guest CPU utilization:

InsightsMetrics
| where TimeGenerated > ago(30m)
| where Origin == "vm.azm.ms"
| where Namespace == "Processor" and Name == "UtilizationPercentage"
| summarize AverageCpu = avg(Val) by bin(TimeGenerated, 5m), Computer
| render timechart

These queries assume the VM resource scope. If you open Logs from a workspace instead, constrain the query to the VM's full resource ID with an _ResourceId filter.

If no performance records appear, check the extension, managed identity, DCR association, destination workspace, and required outbound connectivity. Initial charts need time to accumulate samples. See Troubleshoot AMA on Linux.

For more examples, see InsightsMetrics queries and VM insights Performance views.

Application dependency monitoring

For a website instrumented with Application Insights, Application map can visualize application components and recorded dependencies. This requires an application, supported instrumentation, and generated request/dependency telemetry. This lab creates a VM, not an instrumented website, so an application map isn't an expected output.

Application-level mapping isn't a replacement for arbitrary host-process and port discovery. AMA guest performance collection alone doesn't create either kind of map.