Scale your Applications using Application Gateway Metrics (Beta)

As incoming traffic increases, it becomes crucial to scale up your applications based on the demand.

In the following tutorial, we explain how you can use Application Gateway's AvgRequestCountPerHealthyHost metric to scale up your application. AvgRequestCountPerHealthyHost is measure of average request that are sent to a specific backend pool and backend http setting combination.

We are going to use following two components:

  1. Azure K8S Metric Adapter - We will using the metric adapter to expose Application Gateway metrics through the metric server.
  2. Horizontal Pod Autoscaler - We will use HPA to use Application Gateway metrics and target a deployment for scaling.

Setting up Azure K8S Metric Adapter

  1. We will first create an Azure AAD service principal and assign it Monitoring Reader access over Application Gateway's resource group. Paste the following lines in your Azure Cloud Shell:

    applicationGatewayGroupId=$(az group show -g $applicationGatewayGroupName -o tsv --query "id")
    az ad sp create-for-rbac -n "azure-k8s-metric-adapter-sp" --role "Monitoring Reader" --scopes applicationGatewayGroupId

  2. Now, We will deploy the Azure K8S Metric Adapter using the AAD service principal created above.

    kubectl create namespace custom-metrics
    # use values from service principle created above to create secret
    kubectl create secret generic azure-k8s-metrics-adapter -n custom-metrics \
        --from-literal=azure-tenant-id=<tenantid> \
        --from-literal=azure-client-id=<clientid> \
    kubectl apply -f kubectl apply -f -n custom-metrics
  3. We will create an ExternalMetric resource with name appgw-request-count-metric. This will instruct the metric adapter to expose AvgRequestCountPerHealthyHost metric for myApplicationGateway resource in myResourceGroup resource group. You can use the filter field to target a specific backend pool and backend http setting in the Application Gateway. Copy paste this YAML content in external-metric.yaml and apply with kubectl apply -f external-metric.yaml.

    kind: ExternalMetric
    name: appgw-request-count-metric
      type: azuremonitor
        resourceGroup: myResourceGroup # replace with your application gateway's resource group name
        resourceName: myApplicationGateway # replace with your application gateway's name
        resourceProviderNamespace: Microsoft.Network
        resourceType: applicationGateways
        metricName: AvgRequestCountPerHealthyHost
        aggregation: Average
        filter: BackendSettingsPool eq '<backend-pool-name>~<backend-http-setting-name>' # optional

You can now make a request to the metric server to see if our new metric is getting exposed:

kubectl get --raw "/apis/"
# Sample Output
# {
#   "kind": "ExternalMetricValueList",
#   "apiVersion": "",
#   "metadata":
#     {
#       "selfLink": "/apis/",
#     },
#   "items":
#     [
#       {
#         "metricName": "appgw-request-count-metric",
#         "metricLabels": null,
#         "timestamp": "2019-11-05T00:18:51Z",
#         "value": "30",
#       },
#     ],
# }

Using the new metric to scale up our deployment

Once we are able to expose appgw-request-count-metric through the metric server, We are ready to use Horizontal Pod Autoscaler to scale up our target deployment.

In following example, we will target a sample deployment aspnet. We will scale up Pods when appgw-request-count-metric > 200 per Pod upto a max of 10 Pods.

Replace your target deployment name and apply the following auto scale configuration. Copy paste this YAML content in autoscale-config.yaml and apply with kubectl apply -f autoscale-config.yaml.

apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
  name: deployment-scaler
    apiVersion: apps/v1
    kind: Deployment
    name: aspnet # replace with your deployment's name
  minReplicas: 1
  maxReplicas: 10
  - type: External
      metricName: appgw-request-count-metric
      targetAverageValue: 200

Test your configuration by using a load test tools like apache bench:

ab -n10000 http://<application-gateway-ip-address>/