Enable monitoring for Kubernetes clusters

  • Article

This article describes how to enable complete monitoring of your Kubernetes clusters using the following Azure Monitor features:

Using the Azure portal, you can enable all of the features at the same time. You can also enable them individually by using the Azure CLI, Azure Resource Manager template, Terraform, or Azure Policy. Each of these methods is described in this article.

Important

Kubernetes clusters generate a lot of log data, which can result in significant costs if you aren't selective about the logs that you collect. Before you enable monitoring for your cluster, see the following articles to ensure that your environment is optimized for cost and that you limit your log collection to only the data that you require:

Supported clusters

This article provides onboarding guidance for the following types of clusters. Any differences in the process for each type are noted in the relevant sections.

Prerequisites

Permissions

Managed Prometheus prerequisites

  • The cluster must use managed identity authentication.
  • The following resource providers must be registered in the subscription of the AKS cluster and the Azure Monitor workspace:
    • Microsoft.ContainerService
    • Microsoft.Insights
    • Microsoft.AlertsManagement
    • Microsoft.Monitor
  • The following resource providers must be registered in the subscription of the Grafana workspace subscription:
    • Microsoft.Dashboard

Arc-Enabled Kubernetes clusters prerequisites

Note

The Managed Prometheus Arc-Enabled Kubernetes (preview) extension does not support the following configurations:

  • Red Hat Openshift distributions, including Azure Red Hat OpenShift (ARO)
  • Windows nodes

Workspaces

The following table describes the workspaces that are required to support Managed Prometheus and Container insights. You can create each workspace as part of the onboarding process or use an existing workspace. See Design a Log Analytics workspace architecture for guidance on how many workspaces to create and where they should be placed.

Feature Workspace Notes
Managed Prometheus Azure Monitor workspace Contributor permission is enough for enabling the addon to send data to the Azure Monitor workspace. You will need Owner level permission to link your Azure Monitor Workspace to view metrics in Azure Managed Grafana. This is required because the user executing the onboarding step, needs to be able to give the Azure Managed Grafana System Identity Monitoring Reader role on the Azure Monitor Workspace to query the metrics.
Container insights Log Analytics workspace You can attach an AKS cluster to a Log Analytics workspace in a different Azure subscription in the same Microsoft Entra tenant, but you must use the Azure CLI or an Azure Resource Manager template. You can't currently perform this configuration with the Azure portal.

If you're connecting an existing AKS cluster to a Log Analytics workspace in another subscription, the Microsoft.ContainerService resource provider must be registered in the subscription with the Log Analytics workspace. For more information, see Register resource provider.
Managed Grafana Azure Managed Grafana workspace Link your Grafana workspace to your Azure Monitor workspace to make the Prometheus metrics collected from your cluster available to Grafana dashboards.

Enable Prometheus and Grafana

Use one of the following methods to enable scraping of Prometheus metrics from your cluster and enable Managed Grafana to visualize the metrics. See Link a Grafana workspace for options to connect your Azure Monitor workspace and Azure Managed Grafana workspace.

Note

If you have a single Azure Monitor Resource that is private-linked, then Prometheus enablement won't work if the AKS cluster and Azure Monitor Workspace are in different regions. The configuration needed for the Prometheus add-on isn't available cross region because of the private link constraint. To resolve this, create a new DCE in the AKS cluster location and a new DCRA (association) in the same AKS cluster region. Associate the new DCE with the AKS cluster and name the new association (DCRA) as configurationAccessEndpoint. For full instructions on how to configure the DCEs associated with your Azure Monitor workspace to use a Private Link for data ingestion, see Enable private link for Kubernetes monitoring in Azure Monitor.

Enable with Azure Policy

  1. Download Azure Policy template and parameter files.

  2. Create the policy definition using the following CLI command:

    az policy definition create --name "Prometheus Metrics addon" --display-name "Prometheus Metrics addon" --mode Indexed --metadata version=1.0.0 category=Kubernetes --rules AddonPolicyMetricsProfile.rules.json --params AddonPolicyMetricsProfile.parameters.json

  3. After you create the policy definition, in the Azure portal, select Policy and then Definitions. Select the policy definition you created.

  4. Select Assign and fill in the details on the Parameters tab. Select Review + Create.

  5. If you want to apply the policy to an existing cluster, create a Remediation task for that cluster resource from Policy Assignment.

After the policy is assigned to the subscription, whenever you create a new cluster without Prometheus enabled, the policy will run and deploy to enable Prometheus monitoring.

Enable Container insights

Use one of the following methods to enable Container insights on your cluster. Once this is complete, see Configure agent data collection for Container insights to customize your configuration to ensure that you aren't collecting more data than you require.

Azure portal

  1. From the Definitions tab of the Policy menu in the Azure portal, create a policy definition with the following details.

    • Definition location: Azure subscription where the policy definition should be stored.
    • Name: AKS-Monitoring-Addon
    • Description: Azure custom policy to enable the Monitoring Add-on onto Azure Kubernetes clusters.
    • Category: Select Use existing and then Kubernetes from the dropdown list.
    • Policy rule: Replace the existing sample JSON with the contents of https://aka.ms/aks-enable-monitoring-custom-policy.

  2. Select the new policy definition AKS Monitoring Addon.

  3. Select Assign and specify a Scope of where the policy should be assigned.

  4. Select Next and provide the resource ID of the Log Analytics workspace.

  5. Create a remediation task if you want to apply the policy to existing AKS clusters in the selected scope.

  6. Select Review + create to create the policy assignment.

Azure CLI

  1. Download Azure Policy template and parameter files.

  2. Create the policy definition using the following CLI command:

    az policy definition create --name "AKS-Monitoring-Addon-MSI" --display-name "AKS-Monitoring-Addon-MSI" --mode Indexed --metadata version=1.0.0 category=Kubernetes --rules azure-policy.rules.json --params azure-policy.parameters.json

  3. Create the policy definition using the following CLI command:

    az policy assignment create --name aks-monitoring-addon --policy "AKS-Monitoring-Addon-MSI" --assign-identity --identity-scope /subscriptions/<subscriptionId> --role Contributor --scope /subscriptions/<subscriptionId> --location <location> --role Contributor --scope /subscriptions/<subscriptionId> -p "{ \"workspaceResourceId\": { \"value\":  \"/subscriptions/<subscriptionId>/resourcegroups/<resourceGroupName>/providers/microsoft.operationalinsights/workspaces/<workspaceName>\" } }"

After the policy is assigned to the subscription, whenever you create a new cluster without Prometheus enabled, the policy will run and deploy to enable Prometheus monitoring.

Enable full monitoring with Azure portal

New AKS cluster (Prometheus, Container insights, and Grafana)

When you create a new AKS cluster in the Azure portal, you can enable Prometheus, Container insights, and Grafana from the Monitoring tab. Make sure that you check the Enable Container Logs, Enable Prometheus metrics, and Enable Grafana checkboxes.

Screenshot of Monitoring tab for new AKS cluster.

Existing cluster (Prometheus, Container insights, and Grafana)

  1. Navigate to your AKS cluster in the Azure portal.
  2. In the service menu, under Monitoring, select Insights > Configure monitoring.
  3. Container insights is already enabled. Select the Enable Prometheus metrics and Enable Grafana checkboxes. If you have existing Azure Monitor workspace and Grafana workspace, then they're selected for you.
  4. Select Advanced settings if you want to select alternate workspaces or create new ones. The Cost presets setting allows you to modify the default collection details to reduce your monitoring costs. See Enable cost optimization settings in Container insights for details.
  5. Select Configure.

Existing cluster (Prometheus only)

  1. Navigate to your AKS cluster in the Azure portal.
  2. In the service menu, under Monitoring, select Insights > Configure monitoring.
  3. Select the Enable Prometheus metrics checkbox.
  4. Select Advanced settings if you want to select alternate workspaces or create new ones. The Cost presets setting allows you to modify the default collection details to reduce your monitoring costs.
  5. Select Configure.

Enable Windows metrics collection (preview)

Note

There is no CPU/Memory limit in windows-exporter-daemonset.yaml so it may over-provision the Windows nodes
For more details see Resource reservation

As you deploy workloads, set resource memory and CPU limits on containers. This also subtracts from NodeAllocatable and helps the cluster-wide scheduler in determining which pods to place on which nodes. Scheduling pods without limits may over-provision the Windows nodes and in extreme cases can cause the nodes to become unhealthy.

As of version 6.4.0-main-02-22-2023-3ee44b9e of the Managed Prometheus addon container (prometheus_collector), Windows metric collection has been enabled for the AKS clusters. Onboarding to the Azure Monitor Metrics add-on enables the Windows DaemonSet pods to start running on your node pools. Both Windows Server 2019 and Windows Server 2022 are supported. Follow these steps to enable the pods to collect metrics from your Windows node pools.

  1. Manually install windows-exporter on AKS nodes to access Windows metrics by deploying the windows-exporter-daemonset YAML file. Enable the following collectors:

    • [defaults]
    • container
    • memory
    • process
    • cpu_info

    For more collectors, please see Prometheus exporter for Windows metrics.

    Deploy the windows-exporter-daemonset YAML file. Note that if there are any taints applied in the node, you will need to apply the appropriate tolerations.

        kubectl apply -f windows-exporter-daemonset.yaml

  2. Apply the ama-metrics-settings-configmap to your cluster. Set the windowsexporter and windowskubeproxy Booleans to true. For more information, see Metrics add-on settings configmap.

  3. Enable the recording rules that are required for the out-of-the-box dashboards:

    • If onboarding using the CLI, include the option --enable-windows-recording-rules.
    • If onboarding using an ARM template, Bicep, or Azure Policy, set enableWindowsRecordingRules to true in the parameters file.
    • If the cluster is already onboarded, use this ARM template and this parameter file to create the rule groups. This will add the required recording rules and is not an ARM operation on the cluster and does not impact current monitoring state of the cluster.

Verify deployment

Use the kubectl command line tool to verify that the agent is deployed properly.

Managed Prometheus

Verify that the DaemonSet was deployed properly on the Linux node pools

kubectl get ds ama-metrics-node --namespace=kube-system

The number of pods should be equal to the number of Linux nodes on the cluster. The output should resemble the following example:

User@aksuser:~$ kubectl get ds ama-metrics-node --namespace=kube-system
NAME               DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
ama-metrics-node   1         1         1       1            1           <none>          10h

Verify that Windows nodes were deployed properly

kubectl get ds ama-metrics-win-node --namespace=kube-system

The number of pods should be equal to the number of Windows nodes on the cluster. The output should resemble the following example:

User@aksuser:~$ kubectl get ds ama-metrics-node --namespace=kube-system
NAME                   DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
ama-metrics-win-node   3         3         3       3            3           <none>          10h

Verify that the two ReplicaSets were deployed for Prometheus

kubectl get rs --namespace=kube-system

The output should resemble the following example:

User@aksuser:~$kubectl get rs --namespace=kube-system
NAME                            DESIRED   CURRENT   READY   AGE
ama-metrics-5c974985b8          1         1         1       11h
ama-metrics-ksm-5fcf8dffcd      1         1         1       11h

Container insights

Verify that the DaemonSets were deployed properly on the Linux node pools

kubectl get ds ama-logs --namespace=kube-system

The number of pods should be equal to the number of Linux nodes on the cluster. The output should resemble the following example:

User@aksuser:~$ kubectl get ds ama-logs --namespace=kube-system
NAME       DESIRED   CURRENT   READY     UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
ama-logs   2         2         2         2            2           <none>          1d

Verify that Windows nodes were deployed properly

kubectl get ds ama-logs-windows --namespace=kube-system

The number of pods should be equal to the number of Windows nodes on the cluster. The output should resemble the following example:

User@aksuser:~$ kubectl get ds ama-logs-windows --namespace=kube-system
NAME                   DESIRED   CURRENT   READY     UP-TO-DATE   AVAILABLE   NODE SELECTOR     AGE
ama-logs-windows           2         2         2         2            2       <none>            1d

Verify deployment of the Container insights solution

kubectl get deployment ama-logs-rs --namespace=kube-system

The output should resemble the following example:

User@aksuser:~$ kubectl get deployment ama-logs-rs --namespace=kube-system
NAME          READY   UP-TO-DATE   AVAILABLE   AGE
ama-logs-rs   1/1     1            1           24d

View configuration with CLI

Use the aks show command to find out whether the solution is enabled, the Log Analytics workspace resource ID, and summary information about the cluster.

az aks show --resource-group <resourceGroupofAKSCluster> --name <nameofAksCluster>

The command will return JSON-formatted information about the solution. The addonProfiles section should include information on the omsagent as in the following example:

"addonProfiles": {
    "omsagent": {
        "config": {
            "logAnalyticsWorkspaceResourceID": "/subscriptions/aaaa0a0a-bb1b-cc2c-dd3d-eeeeee4e4e4e/resourcegroups/my-resource-group/providers/microsoft.operationalinsights/workspaces/my-workspace",
            "useAADAuth": "true"
        },
        "enabled": true,
        "identity": null
    },
}

Resources provisioned

When you enable monitoring, the following resources are created in your subscription:

Resource Name Resource Type Resource Group Region/Location Description
MSCI-<aksclusterregion>-<clustername> Data Collection Rule Same as cluster Same as Log Analytics workspace This data collection rule is for log collection by Azure Monitor agent, which uses the Log Analytics workspace as destination, and is associated to the AKS cluster resource.
MSPROM-<aksclusterregion>-<clustername> Data Collection Rule Same as cluster Same as Azure Monitor workspace This data collection rule is for prometheus metrics collection by metrics addon, which has the chosen Azure monitor workspace as destination, and also it is associated to the AKS cluster resource
MSPROM-<aksclusterregion>-<clustername> Data Collection endpoint Same as cluster Same as Azure Monitor workspace This data collection endpoint is used by the above data collection rule for ingesting Prometheus metrics from the metrics addon

When you create a new Azure Monitor workspace, the following additional resources are created as part of it

Resource Name Resource Type Resource Group Region/Location Description
<azuremonitor-workspace-name> Data Collection Rule MA_<azuremonitor-workspace-name>_<azuremonitor-workspace-region>_managed Same as Azure Monitor Workspace DCR created when you use OSS Prometheus server to Remote Write to Azure Monitor Workspace.
<azuremonitor-workspace-name> Data Collection Endpoint MA_<azuremonitor-workspace-name>_<azuremonitor-workspace-region>_managed Same as Azure Monitor Workspace DCE created when you use OSS Prometheus server to Remote Write to Azure Monitor Workspace.

Differences between Windows and Linux clusters

The main differences in monitoring a Windows Server cluster compared to a Linux cluster include:

  • Windows doesn't have a Memory RSS metric. As a result, it isn't available for Windows nodes and containers. The Working Set metric is available.
  • Disk storage capacity information isn't available for Windows nodes.
  • Only pod environments are monitored, not Docker environments.
  • With the preview release, a maximum of 30 Windows Server containers are supported. This limitation doesn't apply to Linux containers.

Note

Container insights support for the Windows Server 2022 operating system is in preview.

The containerized Linux agent (replicaset pod) makes API calls to all the Windows nodes on Kubelet secure port (10250) within the cluster to collect node and container performance-related metrics. Kubelet secure port (:10250) should be opened in the cluster's virtual network for both inbound and outbound for Windows node and container performance-related metrics collection to work.

If you have a Kubernetes cluster with Windows nodes, review and configure the network security group and network policies to make sure the Kubelet secure port (:10250) is open for both inbound and outbound in the cluster's virtual network.

Next steps

  • If you experience issues while you attempt to onboard the solution, review the Troubleshooting guide.
  • With monitoring enabled to collect health and resource utilization of your AKS cluster and workloads running on them, learn how to use Container insights.