/azure-reliability

Azure Functions — Reliability Reference

Supported Plans & Zone Redundancy

Assessment Queries

Zone Redundancy Check

az graph query -q "
resources
| where resourceGroup =~ '<rg>'
| where type =~ 'microsoft.web/serverfarms'
| where kind contains 'functionapp' or kind =~ 'linux' or kind =~ 'elastic'
| project name, sku=sku.name, zoneRedundant=properties.zoneRedundant, location
" --subscriptions <sub-id>

Function App Instance Count (Premium)

az functionapp show --name <app> --resource-group <rg> \
  --query "{minInstances:siteConfig.minimumElasticInstanceCount}" -o table

Configure: Zone Redundancy

Flex Consumption (FC1)

# Enable zone redundancy on plan
az resource update \
  --resource-group <rg> \
  --name <plan-name> \
  --resource-type "Microsoft.Web/serverfarms" \
  --set properties.zoneRedundant=true

Premium (EP1/EP2/EP3)

# Enable zone redundancy + set min capacity
az appservice plan update \
  --name <plan-name> \
  --resource-group <rg> \
  --number-of-workers 2

az resource update \
  --resource-group <rg> \
  --name <plan-name> \
  --resource-type "Microsoft.Web/serverfarms" \
  --set properties.zoneRedundant=true

# Set minimum elastic instances per app
az resource update \
  --resource-group <rg> \
  --name <app-name> \
  --resource-type "Microsoft.Web/sites" \
  --set properties.siteConfig.minimumElasticInstanceCount=2

Consumption (Y1) — upgrade path required

Consumption (Y1) plans do not support zone redundancy. The user must upgrade the plan first:

• Recommended: Upgrade to Flex Consumption — similar serverless model, supports ZR, no per-app minimum cost.
• Alternative: Upgrade to Premium (EP1+) — more control, higher base cost (always-ready instances charged 24/7).

⚠️ Inform the user of cost implications before initiating any plan change.

Configure: Health Endpoint

Flex Consumption does NOT support platform health check (healthCheckPath). Instead, add an HTTP endpoint in code:

TypeScript (v4 programming model)

import { app } from "@azure/functions";

app.http('health', {
  methods: ['GET'],
  authLevel: 'anonymous',
  route: 'health',
  handler: async () => ({ status: 200, body: 'OK' })
});

Python (v2 programming model)

import azure.functions as func

app = func.FunctionApp()

@app.route(route="health", methods=["GET"], auth_level=func.AuthLevel.ANONYMOUS)
def health(req: func.HttpRequest) -> func.HttpResponse:
    return func.HttpResponse("OK", status_code=200)

C# (isolated worker)

[Function("Health")]
public IActionResult Health([HttpTrigger(AuthorizationLevel.Anonymous, "get", Route = "health")] HttpRequest req)
{
    return new OkObjectResult("OK");
}

Premium Functions — Platform Health Check

az webapp config set \
  --name <app-name> \
  --resource-group <rg> \
  --generic-configurations '{"healthCheckPath": "/api/health"}'

⚠️ Enabling health check causes an app restart.

IaC Patching: Bicep

App Service Plan (AVM module)

module appServicePlan 'br/public:avm/res/web/serverfarm:0.5.0' = {
  params: {
    skuName: 'FC1'
    reserved: true
    zoneRedundant: true  // ← ADD
  }
}

Premium Plan — extra settings

module appServicePlan 'br/public:avm/res/web/serverfarm:0.5.0' = {
  params: {
    skuName: 'EP1'
    reserved: true
    zoneRedundant: true      // ← ADD
    skuCapacity: 2           // ← ADD (min 2 for ZR)
  }
}

// On the function app resource:
resource functionApp 'Microsoft.Web/sites@2023-12-01' = {
  properties: {
    siteConfig: {
      minimumElasticInstanceCount: 2  // ← ADD
    }
  }
}

IaC Patching: Terraform

resource "azurerm_service_plan" "plan" {
  sku_name               = "FC1"
  os_type                = "Linux"
  zone_balancing_enabled = true  # ← ADD
}

# Premium plan:
resource "azurerm_service_plan" "plan" {
  sku_name               = "EP1"
  os_type                = "Linux"
  zone_balancing_enabled = true  # ← ADD
  worker_count           = 2     # ← ADD
}

resource "azurerm_linux_function_app" "func" {
  site_config {
    minimum_elastic_instance_count = 2  # ← ADD (Premium only)
  }
}

Multi-Region Notes

• Flex Consumption standby costs ~$0 (pay-per-execution) — ideal for active-passive
• Code must be deployed to both regions separately
• Event Hub checkpoints are per-app — secondary starts from its own checkpoint on failover
• Consider Event Hubs Geo-DR for true event replication

Reporting (for the assessment table)

When the parent skill builds the feature-pivoted assessment table, report each Functions resource on the relevant rows:

Additional References

• Reliability in Azure Functions (Microsoft Learn)