AI-200 practice: Implement event-driven scaling by using Kubernetes Event‑driven Autoscaling (KEDA) in Container Apps

Question 1

A document-classification workload runs in Azure Container Apps and reads from a Service Bus queue. Volume varies from zero overnight to several thousand messages during business hours. Security does not want the Service Bus connection string stored as plain text anywhere in the Container App definition.

Which combination should you configure?

Accepted Answer

A KEDA azure-servicebus scale rule with queue-length trigger, TriggerAuthentication via secret or managed identity, and minReplicas set to 0

Answer

A CPU utilization scale rule with minReplicas set to 1, storing the Service Bus connection string directly in a Container App environment variable value

Answer

A KEDA cron scale rule that increases replica count during business hours only, ignoring actual Service Bus queue depth when scaling decisions run

Answer

An HTTP scale rule targeting the Service Bus REST management endpoint, with the connection string placed in the scale rule metadata fields

Question 2

You define a KEDA-based scale rule for a Container App that should scale out based on the depth of an Azure Storage Queue:
```json
{
  "scale": {
    "minReplicas": 0,
    "maxReplicas": 10,
    "rules": [
      {
        "name": "queue-scale-rule",
        "custom": {
          "type": "azure-queue",
          "metadata": {
            "queueName": "orders-queue",
            "queueLength": "5"
          },
          "auth": [
            {
              "secretRef": "storage-connection-string",
              "triggerParameter": "connection"
            }
          ]
        }
      }
    ]
  }
}
```
The app never scales beyond 1 replica even though the queue has 50 messages. What is the most likely cause?

Accepted Answer

The secretRef named storage-connection-string does not exist in the Container App's secrets array, or the connection string is invalid, causing the scaler to silently fail authentication and default to minimal scaling.

Answer

queueLength must be specified as an integer, not a string, in Container Apps scale rules.

Answer

KEDA queue scalers do not support Azure Storage Queues, only Service Bus.

Answer

minReplicas: 0 prevents the KEDA scaler from ever activating; it must be set to at least 1.

Question 3

You need an AI inference container in Azure Container Apps to scale out based on the depth of an Azure Service Bus queue rather than CPU usage. What should you configure?

Accepted Answer

KEDA azure-servicebus scale rule

Answer

HTTP scale rule on request count

Answer

Manual scale fixed at max replicas

Answer

Monitor autoscale on CPU percent

Question 4

A Container Apps-hosted inference API needs to scale to zero when idle to reduce cost, then scale up on the first incoming HTTP request. Which configuration supports this?

Accepted Answer

minReplicas 0 with an HTTP scale rule

Answer

minReplicas 1 with a CPU scale rule

Answer

Disable autoscaling on the Container App

Answer

AKS Cluster Autoscaler without KEDA rules

Question 5

You are deploying a Python-based AI microservice to Azure Container Apps. The service needs to scale to zero when there are no messages in an Azure Service Bus queue and scale up rapidly during bursts. Which KEDA scaler configuration should you implement in your container app manifest?

Accepted Answer

type: azure-service-bus

Answer

type: azure-queue-storage

Answer

type: azure-event-hubs

Answer

type: azure-blob-storage

Question 6

You are configuring KEDA for an Azure Container App to scale based on CPU usage. You notice the app is scaling up and down too frequently (flapping). Which KEDA setting helps smooth out the scaling behavior?

Accepted Answer

stabilizationWindowSeconds

Answer

cooldownPeriodForScaleDown

Answer

pollingIntervalForMetrics

Answer

minReplicaCountThreshold

Question 7

A Container App processes jobs from the Service Bus queue `jobs` in namespace `sb-prod`. It must scale out on queue depth, and policy forbids storing any Service Bus connection string or SAS key in the app, including as a Container Apps secret. The app has a user-assigned managed identity with an appropriate Service Bus data role. How should you define the scale rule?

Accepted Answer

A custom azure-servicebus rule with queueName and namespace metadata that authenticates with the managed identity (for example --scale-rule-identity <identity resource ID>)

Answer

A custom azure-servicebus rule whose auth block references a Container Apps secret that holds the namespace connection string

Answer

An HTTP scale rule with concurrentRequests set to the expected queue length

Answer

A CPU scale rule at 70% utilization, because KEDA can't authenticate to Service Bus without a connection string

Question 8

A Container App has this scale configuration: `minReplicas: 0`, `maxReplicas: 10`, and one `azure-servicebus` rule with `messageCount: "20"`. A burst puts 300 active messages in the queue. Ignoring timing effects, how many replicas does the app settle on?

Accepted Answer

10, because the target of 15 replicas is capped by maxReplicas

Answer

1, because KEDA only activates a single replica when scaling from zero

Answer

15, because 300 messages divided by 20 messages per replica is 15

Answer

20, because messageCount is the number of replicas to add for each batch

Question 9

A team reports two things about a queue-driven Container App with default scale settings: new messages sometimes wait several seconds before a replica starts, and replicas keep running for minutes after the queue empties. Which two statements correctly explain this default behavior? Select two.

Accepted Answer

Container Apps polls each scale rule's event source on an interval (30 seconds by default), so new work can wait until the next poll

Accepted Answer

After the last event, Container Apps waits a cooldown period (300 seconds by default) before scaling the app to zero

Answer

KEDA scales to zero immediately when the queue length reaches 0, so lingering replicas point to a bug in the app

Answer

Setting minReplicas to 0 disables the scale rules until an HTTP request arrives

Answer

Service Bus pushes messages to KEDA, so any delay must come from Service Bus rather than polling

Question 10

Which Azure CLI command configures a KEDA-based custom scale rule (based on Service Bus queue length) on an existing Container App?

Accepted Answer

az containerapp update --name myApp --resource-group myRG --scale-rule-name queue-rule --scale-rule-type azure-servicebus --scale-rule-metadata queueName=myqueue --scale-rule-auth trigger=connection

Answer

az containerapp revision copy --name myApp --resource-group myRG --from-revision myApp--0001 --scale-min-replicas 1 --scale-max-replicas 5 --image myregistry.azurecr.io/myapp:v2

Answer

az monitor autoscale create --resource <appResourceId> --name queueAutoscale --min-count 1 --max-count 5 --condition "avg Percentage CPU > 70" --resource-group myRG

Answer

az aks nodepool update --resource-group myRG --cluster-name myCluster --name nodepool1 --enable-cluster-autoscaler --min-count 1 --max-count 5 --update-cluster-autoscaler

11 questions