AI-200 Event Driven Integration Practice Test 3 — 15 Questions

Question 1

You are designing an asynchronous AI inference processing pipeline using Azure Service Bus Premium. Messages contain large payloads exceeding the standard Service Bus message size limit of 256 KB. You need to store the large payload in Azure Blob Storage while passing a reference in the Service Bus message. Which feature should you implement?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Use the Azure Service Bus Claim Check pattern by storing the payload in Azure Blob Storage and sending only the blob URI in the message body.. The Claim Check pattern is the correct approach because it decouples the large payload from the messaging infrastructure. By storing the payload in Azure Blob Storage and sending only a reference (the blob URI) in the Service Bus message, you keep the message size well under the 256 KB limit (or 1 MB for Premium) while still enabling asynchronous processing. This pattern is explicitly recommended by Microsoft for messages exceeding Service Bus size limits, and it also reduces costs and improves throughput since the messaging tier handles only small control messages.

Answer

Enable compression on the Azure Service Bus queue properties to automatically shrink payloads larger than 256 KB before enqueuing.

Answer

Configure Azure Service Bus auto-forwarding to a secondary queue configured for jumbo messages.

Answer

Upgrade the Service Bus namespace to the Enterprise tier to unlock message payloads up to 100 MB.

Question 2

You are configuring an Azure Event Grid system to route custom application events to an Azure Function webhook. To ensure secure communication, you decide to use Event Grid custom topics with Azure AD authentication. Which configuration step is required on the Azure Function endpoint to authenticate incoming events from Event Grid?

Accepted Answer

Implement validation handshake logic in the Azure Function code to respond to the Event Grid subscription validation event.. When Event Grid creates a webhook subscription, it sends a subscription validation event to the endpoint and expects the endpoint to echo back the validationCode (or validationUrl) to prove ownership. The Azure Function must implement this handshake logic, otherwise Event Grid marks the subscription as failed and stops delivering events.

Answer

Configure the Azure Function app with an Event Grid system-assigned managed identity and grant it the Event Grid Data Sender role.

Answer

Install the Microsoft.Azure.EventGrid NuGet package and configure a SAS token generated from the Function App settings.

Answer

Enable mutual TLS (mTLS) authentication on the Azure Function App and upload the Event Grid root CA certificate.

Question 3

Your enterprise AI application ingests millions of telemetry events per second from IoT devices using Azure Event Hubs. Downstream AI models consume these events via Apache Spark on Azure Databricks. You need to ensure that telemetry from the same device is always processed by the same Spark executor in order to maintain state. How should you partition your Event Hub?

Accepted Answer

Configure a custom partition key containing the device ID when publishing events to the Event Hub.. Event Hubs guarantees ordering and affinity only within a single partition. By setting a PartitionKey that contains the device ID, all events from that device hash to the same partition, and since a Spark executor reads a given partition exclusively, per-device state is preserved. This is the canonical way to achieve per-entity ordering in Event Hubs.

Answer

Increase the number of Consumer Groups on the Event Hub to isolate different device streams.

Answer

Set the Event Hub capture feature to output messages to Azure Data Lake Storage partitioned by device ID.

Answer

Deploy an Azure Service Bus Topic in front of Event Hubs with session IDs enabled.

Question 4

Your organization is building a multi-tenant AI platform. Tenants publish custom events to an Azure Event Grid topic. You need to ensure that tenant A's subscribers only receive events where the event subject starts with "tenant-a/". Which Event Grid feature should you implement on the event subscription?

Accepted Answer

Event Grid Advanced Filters with a StringBeginsWith operator on the subject property.. Event Grid Advanced Filters support operators like StringBeginsWith, StringEndsWith, StringContains, NumberIn, and BoolEquals applied to event properties including subject, eventType, and data fields. Since the requirement is to route only events whose subject starts with 'tenant-a/', a StringBeginsWith advanced filter on the subject property is the native Event Grid mechanism. This filtering happens at the subscription level, so tenant A's subscribers never receive other tenants' events.

Answer

Service Bus Topic Filters and Actions with SQL filter syntax.

Answer

Event Hubs Capture filtering rules based on partition keys.

Answer

Azure API Management policies applied to the Event Grid custom topic ingress endpoint.

Question 5

You are troubleshooting an Azure Event Grid subscription where events published to a custom topic are failing to reach an Azure Storage Queue destination. Dead-lettering is enabled on the subscription, and failed events are appearing in the storage account container. Upon inspecting the dead-lettered events, you notice the error indicates an unauthorized access response. What is the most likely root cause and remediation?

Accepted Answer

The Event Grid system-assigned managed identity lacks the Storage Queue Data Contributor role on the target storage account.. When Event Grid delivers events to a Storage Queue using managed identity authentication, the Event Grid system-assigned identity must have the 'Storage Queue Data Contributor' role on the target storage account. An 'unauthorized access' error in dead-lettered events indicates the identity lacks this RBAC assignment, so the fix is to grant the role.

Answer

The event schema version configured on the subscription does not match the storage queue message schema specification.

Answer

The Azure Storage Queue firewall is blocking Event Grid IPs, so you must add the EventGrid service tag to the storage account firewall rules.

Answer

The Azure Storage Queue message TTL has expired before Event Grid could establish a TCP handshake.

Question 6

You are architecting a resilient, high-throughput AI telemetry ingestion pipeline using Azure Event Hubs. The pipeline must support multiple independent downstream AI consumer applications reading the same telemetry stream at different speeds without interfering with one another. Which TWO architectural components or features must you configure? (Choose two)

Accepted Answer

Multiple Consumer Groups. Option A (Multiple Consumer Groups) is correct because Event Hubs consumer groups give each independent downstream AI consumer application its own view of the same telemetry stream, with separate offset and checkpoint tracking, so consumers reading at different speeds do not interfere with one another. Option C (Event Hubs Capture) is correct because it is the built-in Event Hubs feature that automatically persists the streaming telemetry to Azure Blob Storage or Azure Data Lake Storage in Avro format, enabling durable, high-throughput retention of the same stream for downstream analytics or replay. Option B is incorrect because Azure Service Bus Topics with subscriptions is a separate messaging service, not an Event Hubs feature, and would change the architecture away from Event Hubs. Option D is incorrect because Geo-Disaster Recovery is a metadata replication/aliasing feature for regional failover, not a partitioning or capture mechanism. Option E is incorrect because Event Grid system topics deliver resource events (for example, blob created events) and are not the telemetry ingestion or consumer-isolation mechanism required here.

Answer

Azure Service Bus Topics with subscriptions

Answer

Capture partitioning via Geo-Disaster Recovery

Answer

Event Grid system topics

Question 7

You are designing an event-driven AI inference system that uses Azure Service Bus topics and subscriptions. A subscription receives messages from multiple publishers. You need to ensure that messages from a specific publisher are processed only by a dedicated worker pool, while other messages go to a different worker pool. You also need to ensure that messages that cannot be processed after several attempts are moved to a separate queue for manual inspection. Which two actions should you perform? (Choose two.)

Accepted Answer

Set the MaxDeliveryCount property on the subscription to a low value to quickly move failed messages to the dead-letter queue.. To route messages from a specific publisher to a dedicated worker pool, use a SQL filter on the subscription to match a custom property. To move unprocessable messages to a separate queue for manual inspection, set the MaxDeliveryCount to a low value so that after a few failed attempts, messages are automatically moved to the subscription's dead-letter queue. These two actions together meet the requirements.

Answer

Enable dead-lettering on the subscription and configure a separate queue as the dead-letter destination.

Answer

Create a separate topic for each publisher and have workers subscribe to the respective topics.

Answer

Use a correlation filter to match messages based on a unique identifier.

Question 8

You are building an AI-powered image moderation service on Azure. When a new image is uploaded to Azure Blob Storage, an Azure Event Grid subscription triggers an Azure Function that calls a third-party moderation API. The third-party API is frequently unavailable for short bursts, and the Function currently throws an exception and loses the event. You need the Function to retry the event delivery automatically with exponential backoff without writing custom retry code. What should you configure?

Accepted Answer

Enable the Azure Function's built-in retry policy using the retry attribute on the Event Grid trigger binding.. The failure occurs inside the Azure Function when the external moderation API is temporarily unavailable, so the retry must happen at the function invocation layer. The Event Grid trigger's built-in retry policy re-invokes the function with exponential backoff on unhandled exceptions, satisfying the requirement without custom code. Delivery-layer settings and architectural rewrites do not address this specific need.

Answer

Move the moderation call to an Azure Durable Functions orchestrator with a RetryOptions policy on the activity function.

Answer

Configure the Event Grid subscription's retry policy and event time-to-live (TTL) settings on the subscription.

Answer

Increase the Event Grid subscription's max delivery attempts and set the event TTL to 24 hours.

Question 9

You are building a serverless document-classification pipeline. When a new PDF is uploaded to Azure Blob Storage, an Azure Event Grid system topic raises a BlobCreated event. Your Azure Function must enrich the event with an AI-generated document category before the result is written to a downstream system. The enrichment call is idempotent but occasionally takes 90 seconds. You need the Function to receive the original Event Grid event payload, including the subject and eventType, and to release the event back to Event Grid for retry when the enrichment call fails. Which binding configuration should you use?

Accepted Answer

Configure an Event Grid trigger on the Function and set maxEventBatchSize to 1, then throw an exception to force a retry.. The Event Grid trigger is the only trigger that surfaces the original Event Grid event payload, including subject and eventType, to the Function. Handling one event per invocation and throwing an exception on failure causes Event Grid to retry delivery under the subscription's retry policy. Using a Blob Storage trigger loses the event metadata, and host.json retry policies do not apply to Event Grid deliveries.

Answer

Configure an Event Grid trigger and set the Function's retry policy in host.json to exponential backoff with a maximum retry count of 5.

Answer

Configure a Blob Storage trigger on the Function and use the blob path to reconstruct the Event Grid event properties.

Answer

Configure an Event Grid trigger and write the enriched event to a Service Bus queue, then have a second Function process the queue for retries.

Question 10

You are developing a real-time AI sentiment analysis pipeline that consumes messages from an Azure Service Bus queue. The messages contain large text payloads, and you need to ensure that the consumer application can handle messages that exceed the standard Service Bus message size limit. You also want to minimize latency and cost. Which TWO actions should you take? (Choose two.)

Accepted Answer

Store the large text payload in Azure Blob Storage and send a reference to the blob in the Service Bus message.. To handle messages larger than the Service Bus limit, you can either use a Premium tier namespace, which raises the limit to 100 MB, or implement the claim-check pattern by storing the payload in Blob Storage and sending a reference. Both actions directly address the large payload requirement, while other options do not affect the size limit or are less reliable.

Answer

Partition the Service Bus queue to increase the effective message size limit.

Answer

Compress the text payload using GZip before sending it in the Service Bus message.

Answer

Enable the Service Bus message sessions feature to group large messages together.

Question 11

You are implementing a real-time AI telemetry pipeline that uses Azure Event Hubs with 32 partitions. A consumer group named ai-scoring is used by a Java application running on four Azure Kubernetes Service pods. You observe that after a pod restart, some telemetry records are processed twice, and the checkpoint store shows offsets advancing in an uneven pattern. You need to minimize duplicate processing while maintaining throughput. What should you do?

Accepted Answer

Enable the EventProcessorClient's checkpointing after each event and set the maximum wait time to a low value.. Duplicate processing after a pod restart is expected with at-least-once delivery. The EventProcessorClient reduces duplicates by checkpointing frequently, and checkpointing after each event shrinks the re-delivery window to a single event. Producer-side duplicate detection and client changes do not address consumer checkpoint timing, and adding pods does not change the at-least-once contract.

Answer

Enable duplicate detection on the Event Hubs namespace and set the time window to 10 minutes.

Answer

Switch the consumer to use the EventHubConsumerClient with a single receiver for all partitions.

Answer

Increase the number of pods to 32 so that each pod owns exactly one partition.

Question 12

An AI inference service publishes results to an Azure Service Bus topic named inference-results. A downstream subscription named audit-sub must receive every result for compliance, while a subscription named alert-sub should only receive results where the model confidence is below 0.5. The publisher sets a custom application property named confidence with a double value. You need to configure alert-sub so that only low-confidence messages are delivered, without modifying the publisher. What should you do?

Accepted Answer

Create a SQL filter on alert-sub with the expression confidence < 0.5 and remove the default True filter from the subscription.. Service Bus SQL filters support range comparisons on custom application properties, which are referenced by name without the sys prefix. The default True filter on a new subscription matches all messages, so it must be removed for the custom filter to take effect. Correlation filters only support equality and cannot express a less-than condition, and sys-prefixed names refer to system properties.

Answer

Create a correlation filter on alert-sub with the property confidence set to less than 0.5.

Answer

Create a SQL filter on alert-sub with the expression confidence < 0.5 and remove the default True filter.

Answer

Create a SQL filter on alert-sub with the expression sys.confidence < 0.5 and remove the default True filter.

Question 13

You are designing an event-driven pipeline where an Azure Function processes AI training completion events from an Azure Event Hubs instance. The Function must process events in order within each partition and must not lose events if the Function host crashes mid-batch. You need to configure the Event Hubs trigger binding to meet these requirements. Which TWO settings or behaviors should you implement? (Choose two.)

Accepted Answer

Use a single consumer group dedicated to this Function and ensure no other consumers read from the same consumer group.. In-order processing within a partition requires a single consumer per partition, which the cardinality setting to one enforces. A dedicated consumer group prevents other readers from advancing offsets or stealing leases, which is essential for both ordering and no-loss guarantees. Automatic trigger checkpointing and larger batch sizes do not provide the required guarantees, and the trigger does not expose a per-event checkpoint strategy.

Answer

Enable the Event Hubs trigger's built-in checkpointing and rely on it to commit offsets after each batch completes.

Answer

Configure the trigger to use the EventProcessorClient with a blob storage checkpoint store and set the checkpoint strategy to after each event.

Answer

Set the trigger's maxEventBatchSize to the maximum allowed value to reduce the number of checkpoints.

Question 14

Your team is building an asynchronous order processing pipeline. If a downstream AI service fails to process a message from a Service Bus queue after 5 delivery attempts, the message must be automatically removed from active processing and preserved for manual inspection. What should you configure?

Accepted Answer

Configure the Max Delivery Count property on the queue and enable dead-lettering on message expiration or failure.. Configuring Max Delivery Count on the queue (for example, 5) combined with dead-lettering on message expiration or delivery failure ensures that after five unsuccessful delivery attempts the message is moved to the dead-letter queue (DLQ) rather than being retried indefinitely or discarded. The DLQ preserves the message payload and failure metadata for manual inspection and reprocessing. This is the native Service Bus mechanism for poison-message handling.

Answer

Enable Duplicate Detection with a history window of 5 attempts.

Answer

Set the AutoDeleteOnIdle property on the queue to 5 minutes.

Answer

Implement Event Grid dead-lettering pointing to an Azure Blob Storage container.

Question 15

Your enterprise application uses Azure Service Bus Topics to distribute orders to inventory and billing services. A transient database outage occurs in the billing service. You must ensure that billing messages are not lost and can be retried without interfering with the inventory service. What feature should you configure?

Accepted Answer

Configure separate subscriptions for each service under the Service Bus Topic.. Configuring separate subscriptions for each service under the Service Bus Topic is correct because each subscription receives its own copy of the message, allowing the billing service to process independently and retry on failure without affecting the inventory service. This provides isolation and ensures messages are not lost during transient outages. Other options do not provide the required decoupling and retry capability.

Answer

Enable partitioned queues on the sender application.

Answer

Use Event Grid filters to duplicate the messages into two separate storage accounts.

Answer

Implement Azure Relay to tunnel the database connection.