How Do You Reduce Azure Container Apps Costs?

// Bicep template — Container App with scale-to-zero
resource containerApp 'Microsoft.App/containerApps@2023-05-01' = {
  name: 'my-api'
  location: location
  properties: {
    environmentId: environment.id
    configuration: {
      ingress: {
        external: false          // internal-only saves request charges
        targetPort: 8080
      }
    }
    template: {
      containers: [
        {
          name: 'api'
          image: 'myregistry.azurecr.io/api:latest'
          resources: {
            cpu: json('0.25')    // minimum allocation
            memory: '0.5Gi'
          }
        }
      ]
      scale: {
        minReplicas: 0           // scale to zero when idle
        maxReplicas: 10
        rules: [
          {
            name: 'http-scaling'
            http: {
              metadata: {
                concurrentRequests: '50'
              }
            }
          }
        ]
      }
    }
  }
}

# Azure CLI — update container resource allocation
az containerapp update \
  --name my-api \
  --resource-group rg-production \
  --cpu 0.25 \
  --memory 0.5Gi

# Verify current allocation
az containerapp show \
  --name my-api \
  --resource-group rg-production \
  --query "properties.template.containers[0].resources"

// Bicep — autoscale with multiple rules
scale: {
  minReplicas: 0
  maxReplicas: 20
  rules: [
    {
      name: 'http-rule'
      http: {
        metadata: {
          concurrentRequests: '100'   // scale at 100 concurrent, not default 10
        }
      }
    }
    {
      name: 'queue-rule'
      custom: {
        type: 'azure-servicebus'
        metadata: {
          queueName: 'orders'
          namespace: 'sb-production'
          messageCount: '50'          // scale per 50 messages
        }
        auth: [
          {
            secretRef: 'sb-connection'
            triggerParameter: 'connection'
          }
        ]
      }
    }
  ]
}

// Bicep — KEDA cron scaler for business-hours scaling
scale: {
  minReplicas: 0
  maxReplicas: 10
  rules: [
    {
      name: 'business-hours'
      custom: {
        type: 'cron'
        metadata: {
          timezone: 'America/New_York'
          start: '0 8 * * 1-5'       // Mon-Fri 8 AM
          end: '0 18 * * 1-5'        // Mon-Fri 6 PM
          desiredReplicas: '2'        // keep 2 warm during business hours
        }
      }
    }
    {
      name: 'http-rule'
      http: {
        metadata: {
          concurrentRequests: '50'
        }
      }
    }
  ]
}

# KEDA ScaledObject — tuned for cost optimization
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: order-processor
spec:
  scaleTargetRef:
    name: order-processor
  pollingInterval: 30              # check every 30s, not default 10s
  cooldownPeriod: 300              # wait 5 min before scale-down
  minReplicaCount: 0
  maxReplicaCount: 10
  triggers:
    - type: azure-servicebus
      metadata:
        queueName: orders
        namespace: sb-production
        messageCount: "50"
      authenticationRef:
        name: sb-trigger-auth

// KQL — identify over-provisioned Container Apps
ContainerAppConsoleLogs_CL
| where TimeGenerated > ago(7d)
| summarize
    AvgCPU = avg(CpuUsage_d),
    MaxCPU = max(CpuUsage_d),
    AvgMemoryMB = avg(MemoryUsage_d),
    MaxMemoryMB = max(MemoryUsage_d)
  by ContainerAppName_s, ContainerName_s
| extend
    AllocatedCPU = 1.0,      // adjust to your actual allocation
    AllocatedMemoryMB = 2048  // adjust to your actual allocation
| extend
    CPUUtilPct = round(MaxCPU / AllocatedCPU * 100, 1),
    MemUtilPct = round(MaxMemoryMB / AllocatedMemoryMB * 100, 1)
| where CPUUtilPct < 50 or MemUtilPct < 50
| project
    ContainerAppName_s,
    ContainerName_s,
    CPUUtilPct,
    MemUtilPct,
    RecommendedCPU = case(
      MaxCPU < 0.125, 0.25,
      MaxCPU < 0.375, 0.5,
      MaxCPU < 0.75, 1.0,
      2.0
    )
| order by CPUUtilPct asc

# List all workload profiles and their utilization
az containerapp env workload-profile list \
  --resource-group rg-production \
  --name my-environment \
  --output table

# Move a container app to a different workload profile
az containerapp update \
  --name my-api \
  --resource-group rg-production \
  --workload-profile-name "Consumption"  # move back to consumption

# Remove unused dedicated workload profile
az containerapp env workload-profile delete \
  --resource-group rg-production \
  --name my-environment \
  --workload-profile-name "dedicated-d4"

# Set ingress to internal-only
az containerapp ingress update \
  --name background-worker \
  --resource-group rg-production \
  --type internal

# Disable ingress entirely for queue-driven workers
az containerapp ingress disable \
  --name queue-processor \
  --resource-group rg-production

# Check current storage mounts for a container app
az containerapp show \
  --name my-api \
  --resource-group rg-production \
  --query "properties.template.volumes"

# Update Azure Files share to use Standard tier
az storage share-rm update \
  --resource-group rg-production \
  --storage-account mystorageaccount \
  --name app-logs \
  --access-tier "TransactionOptimized"  # or "Cool" for infrequent access

START
  |
  v
[Does the app have minReplicas > 0?]
  |                              |
 YES                             NO
  |                              |
  v                              v
[Is it latency-sensitive       [Check resource allocation]
 production API with SLA?]       |
  |              |               v
 YES            NO            [Is CPU/memory > 2x peak usage?]
  |              |               |              |
  v              v              YES             NO
[Keep min=1,  [Set min=0,       |              |
 add cron      immediate        v              v
 scaling]      savings]      [Right-size     [Review autoscale
                              allocation]     thresholds]
                                               |
                                               v
                                            [Are KEDA thresholds
                                             below 50 messages?]
                                               |           |
                                              YES          NO
                                               |           |
                                               v           v
                                            [Increase    [Check ingress
                                             threshold,   configuration]
                                             extend
                                             cooldown]

START
  |
  v
[Does the workload need GPU?]
  |              |
 YES             NO
  |              |
  v              v
[Use           [Does it need > 4 vCPU or > 8 GiB memory?]
 Dedicated       |              |
 (GPU)]         YES             NO
                 |              |
                 v              v
              [Use           [Is sustained utilization > 60%?]
               Dedicated       |              |
               (D-series)]    YES             NO
                               |              |
                               v              v
                            [Use           [Does it need custom VNET?]
                             Dedicated,      |              |
                             consider       YES             NO
                             savings        |              |
                             plan]          v              v
                                         [Use           [Use Consumption
                                          Dedicated      with scale-to-zero]
                                          with VNET]