CLZ-270 - Windows VMSS Autoscaling

Service or workflow illustration. The lesson defines the exact resource settings. Open at full size.

Goal

Attach bounded CPU-driven autoscaling to the Windows IIS VMSS and validate capacity as one coherent object with integer minimum, default, and maximum values.

Arrows show the learning sequence, not service traffic. Azure deployment. Open at full size.

Prerequisites

• Complete CLZ-260.
• Quota for the configured maximum capacity.
• Understand Azure Monitor metric aggregation and cooldown periods.

Architecture

Lesson-specific teaching diagram. Read the arrow legend and the accompanying architecture explanation. Azure deployment. On narrow screens, scroll the diagram horizontally. Open at full size.

The shared VMSS module supplies the load-balanced IIS tier and explicit outbound path. An Azure Monitor autoscale setting targets the VMSS, scales out above the CPU threshold, and scales in below it while never crossing the validated capacity bounds.

Resources And Cost

Relative cost: High. The autoscale maximum is the budget ceiling; actual VM, disk, public IP, and data charges vary over time.

Component	Purpose and principal cost driver
Windows IIS VMSS and Load Balancer	Scalable application tier created by the shared module.
Autoscale setting	CPU rules, cooldowns, and bounded capacity profile.
VNet and web NSG	Network and public HTTP teaching path.

Inputs

Input	Required substitution or decision
subscription_id	Required for Azure plan, apply, and destroy; use the exact subscription GUID.
environment	Keep dev for the lab unless testing another environment boundary.
location	Select a supported region with sufficient quota.
name_prefix	Use a short lowercase prefix.
admin_username	Nonreserved scale-set administrator name.
vm_size	Uniform Windows size used by the VMSS.
autoscale_capacity	Required object { minimum, default, maximum }; all values are integers and must satisfy minimum <= default <= maximum.

Keep real subscription IDs, CIDRs, domains, endpoint names, and secrets in untracked local files or environment variables.

Deploy

Review the plan and lesson-specific gates before deliberate execution. Local-only and read-only lessons create no Azure resources. Open at full size.

Run from the repository root in PowerShell 7:

Set-Location CLZ-270-windows-vmss-autoscaling
if (-not (Test-Path terraform.tfvars)) {
  Copy-Item terraform.tfvars.example terraform.tfvars
}
$subscriptionId = az account show --query id --output tsv
terraform init -lockfile=readonly
terraform fmt -check
terraform validate
terraform plan -var "subscription_id=$subscriptionId" -out tfplan
terraform show tfplan
terraform apply tfplan

Validate

The shared VMSS module uses a Manual upgrade policy. An IIS extension added to the model may require explicitly updating existing instances. Review the instance model and extension status before expecting healthy HTTP responses:

$resourceGroup = terraform output -raw resource_group_name
$scaleSetName = az vmss list --resource-group $resourceGroup --query '[0].name' --output tsv
az vmss list-instances --resource-group $resourceGroup --name $scaleSetName --expand instanceView --output json
# Deliberate Azure change: update these lab instances to the latest model after review.
az vmss update-instances --resource-group $resourceGroup --name $scaleSetName --instance-ids '*'

Follow Microsoft's VMSS application deployment guidance. Wait for successful extension execution and probe health before checking HTTP.

$resourceGroup = terraform output -raw resource_group_name
az monitor autoscale show --resource-group $resourceGroup --name (terraform output -raw autoscale_setting_name) --query '{enabled:enabled,capacity:profiles[0].capacity,rules:profiles[0].rules[].metricTrigger.operator}' --output json
(Invoke-WebRequest -Uri (terraform output -raw vmss_load_balancer_url) -UseBasicParsing).StatusCode
terraform plan -var "subscription_id=$subscriptionId" -var 'autoscale_capacity={minimum=3,default=2,maximum=1}'

Expected result:

• Autoscale is enabled with the requested string capacities and two metric rules.
• HTTP returns 200 through the Load Balancer.
• The deliberately invalid capacity ordering fails during Terraform validation without changing Azure.

Troubleshooting

Gather evidence, repair the cause, and rerun verification. Open at full size.

Symptom	Resolution
Autoscale does not react immediately	Metric windows and cooldowns intentionally delay action; inspect metric data and autoscale run history.
The capacity object is rejected	Use whole numbers and preserve minimum <= default <= maximum.
Scale-out hits a subscription limit	Set maximum within verified quota and budget, then reapply the bounds.

Cleanup

Use the lesson state and cleanup instructions. Preserve remote-state dependencies until consumers are removed. Open at full size.

$resourceGroup = terraform output -raw resource_group_name
terraform destroy -var "subscription_id=$subscriptionId"
az group exists --name $resourceGroup

The autoscale setting is removed before the VMSS and supporting network resources.

Production Delta

• Use application and queue metrics where they better represent demand, and test scale behavior with realistic load.
• Add cost alerts, quota headroom, instance warm-up awareness, health-based rollout, and operational dashboards.

Exercises

1. Create an invalid fractional capacity in a temporary tfvars file and explain the validation error.
2. Generate safe test load, watch metrics, and relate the observed delay to time window and cooldown settings.

Navigation

← CLZ-260 - Windows VMSS Manual Scaling | Curriculum | CLZ-280 - NAT Gateway Outbound →