Reliability pillar principles in Azure - Commands & Configuration

Introduction

Reliability means making sure your cloud services keep working well even if something goes wrong. It helps avoid downtime and keeps your users happy by handling failures smoothly.

When you want your website to stay online even if a server crashes

When you need to recover quickly from unexpected errors in your app

When you want to test how your system behaves under failure conditions

When you want to automatically fix problems without manual intervention

When you want to monitor your services to catch issues early

Commands

This command creates an alert in Azure Monitor to notify you if the CPU usage on a virtual machine goes above 80%. It helps detect problems early to keep your service reliable.

Terminal

az monitor metrics alert create --name HighCPUAlert --resource-group example-rg --scopes /subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/virtualMachines/example-vm --condition "avg Percentage CPU > 80" --description "Alert when CPU usage is high"

Expected Output

{"id":"/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/microsoft.insights/metricAlerts/HighCPUAlert","name":"HighCPUAlert","type":"microsoft.insights/metricalerts","location":"global","tags":{},"properties":{"description":"Alert when CPU usage is high","enabled":true,"scopes":["/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/virtualMachines/example-vm"],"evaluationFrequency":"PT1M","windowSize":"PT5M","criteria":{"allOf":[{"criterionType":"StaticThresholdCriterion","metricName":"Percentage CPU","metricNamespace":"Microsoft.Compute/virtualMachines","operator":"GreaterThan","threshold":80,"timeAggregation":"Average"}]},"autoMitigate":true}}

→

--condition - Defines the metric and threshold to trigger the alert

→

--scopes - Specifies the resource to monitor

This command creates an availability set for virtual machines. It spreads VMs across different hardware to reduce downtime if one server fails.

Terminal

az vm availability-set create --name example-avset --resource-group example-rg --platform-fault-domain-count 2 --platform-update-domain-count 5

Expected Output

{"id":"/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/availabilitySets/example-avset","name":"example-avset","type":"Microsoft.Compute/availabilitySets","location":"eastus","properties":{"platformFaultDomainCount":2,"platformUpdateDomainCount":5}}

→

--platform-fault-domain-count - Number of fault domains to spread VMs across

→

--platform-update-domain-count - Number of update domains to spread VMs across

This command creates a virtual machine inside the availability set to ensure it benefits from fault and update domain protection.

Terminal

az vm create --resource-group example-rg --name example-vm1 --image UbuntuLTS --availability-set example-avset --admin-username azureuser --generate-ssh-keys

Expected Output

{

  "fqdns": "",

  "id": "/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/virtualMachines/example-vm1",

  "location": "eastus",

  "name": "example-vm1",

  "powerState": "VM running",

  "resourceGroup": "example-rg",

  "zones": null

}

→

--availability-set - Assigns the VM to the availability set for reliability

This command creates a virtual machine scale set that automatically updates and keeps multiple VM instances running for high availability.

Terminal

az vmss create --resource-group example-rg --name example-vmss --image UbuntuLTS --upgrade-policy-mode automatic --admin-username azureuser --generate-ssh-keys --instance-count 2

Expected Output

{

  "id": "/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/virtualMachineScaleSets/example-vmss",

  "location": "eastus",

  "name": "example-vmss",

  "provisioningState": "Succeeded",

  "resourceGroup": "example-rg"

}

→

--upgrade-policy-mode - Controls how VMs are updated automatically

→

--instance-count - Number of VM instances to run

This command sets up autoscaling to add or remove VM instances automatically based on demand, helping keep the service reliable under changing loads.

Terminal

az monitor autoscale create --resource-group example-rg --resource example-vmss --resource-type Microsoft.Compute/virtualMachineScaleSets --name example-autoscale --min-count 2 --max-count 5 --count 2

Expected Output

{"id":"/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/microsoft.insights/autoscalesettings/example-autoscale","name":"example-autoscale","type":"microsoft.insights/autoscalesettings","location":"global","properties":{"enabled":true,"targetResourceUri":"/subscriptions/00000000-0000-0000-0000-000000000000/resourceGroups/example-rg/providers/Microsoft.Compute/virtualMachineScaleSets/example-vmss","profiles":[{"name":"Auto created profile","capacity":{"minimum":"2","maximum":"5","default":"2"},"rules":[]}],"notifications":[]}}

→

--min-count - Minimum number of VM instances

→

--max-count - Maximum number of VM instances

Key Concept

If you remember nothing else from this pattern, remember: design your cloud resources to detect problems early and recover automatically to keep your service running smoothly.

Common Mistakes

Not setting up alerts to monitor resource health

Without alerts, you won't know when something goes wrong until users complain

Create metric alerts to get notified immediately about issues

Deploying VMs without availability sets or scale sets

A single VM failure can cause downtime if not spread across fault domains

Use availability sets or scale sets to distribute VMs for higher reliability

Not configuring autoscaling for variable workloads

Your service may become slow or crash under heavy load without enough resources

Set up autoscale rules to add or remove instances automatically

Summary

Create alerts to monitor important metrics and get notified of issues early.

Use availability sets or scale sets to spread resources and avoid single points of failure.

Configure autoscaling to adjust resources automatically based on demand.