Free AI-200 Practice Questions

Your team pushes code to a Git repository that produces a Dockerfile-based image. You want Azure Container Registry to automatically rebuild and push the image whenever the base image or the source commit changes, without a separate CI system.

What should you configure?

1. A.An ACR Task with a Git commit trigger and a base image update trigger
2. B.A scheduled ACR Task that rebuilds nightly regardless of changes
3. C.An Azure DevOps pipeline that calls docker build and docker push manually
4. D.A Logic App that polls the Git repo every hour and calls the ACR REST API

You need to roll out a new container image to 10% of production traffic while the previous version continues serving the other 90%, so you can monitor error rates before a full cutover.

Which Azure Container Apps feature achieves this?

1. A.Multiple revisions with traffic-splitting weights configured between the active revisions
2. B.A single revision with a manually toggled feature flag in application code
3. C.Deploying the new image to a second, separate Container Apps environment
4. D.Scaling the existing revision to zero and back up with the new image

A Container Apps job processes messages from an Azure Storage queue. Volume swings from zero to several thousand messages unpredictably, and you want replica count to track queue depth automatically, including scaling to zero when idle.

Which scaling configuration is correct?

1. A.A KEDA Azure Queue Storage scale rule referencing the queue length, with minReplicas set to 0
2. B.A fixed replica count set to the historical peak to avoid scaling delays
3. C.An Azure Automation runbook on a 5-minute schedule that adjusts replica count
4. D.HTTP scaling rules, since Container Apps only autoscales on inbound requests

After deploying a new manifest to AKS, pods repeatedly restart and enter CrashLoopBackOff. You need to determine whether the application is failing its health check or crashing on startup before it is ready to serve traffic.

What should you inspect first?

1. A.The readiness and liveness probe configuration and recent pod events
2. B.The AKS cluster's node pool VM size
3. C.The Azure Container Registry access policy
4. D.The cluster's Kubernetes version compatibility matrix

You are deploying a container to Azure App Service. The container needs a database connection string that must never appear in plaintext in App Service configuration or deployment logs.

How should you supply the connection string?

1. A.Store it as a plain App Setting environment variable
2. B.Store it in Azure Key Vault and reference it from App Service using a Key Vault reference
3. C.Bake it into the container image at build time
4. D.Pass it as a command-line argument in the container start command

A Cosmos DB for NoSQL container stores large JSON documents with a nested "rawPayload" field that is never queried but is included in every write. RU consumption on writes is higher than expected.

What should you change to reduce RU cost without affecting query results?

1. A.Exclude the rawPayload path from the indexing policy
2. B.Switch the consistency level from Session to Strong
3. C.Increase the provisioned throughput (RU/s) on the container
4. D.Enable a change feed processor on the container

A globally distributed order-processing system needs the lowest possible read latency across regions and can tolerate reading slightly stale data, as long as reads never go backward in time for a single client session.

Which consistency level fits this requirement?

1. A.Session
2. B.Strong
3. C.Eventual
4. D.Bounded Staleness with a very large staleness window

You are storing document embeddings in Cosmos DB for NoSQL and need to run a similarity search that returns the nearest vectors to a query embedding, with acceptable latency at scale.

What must you configure on the container to support this?

1. A.A vector embedding policy and a vector index defined on the container, then query using the vector distance function
2. B.A composite index on the embedding array property
3. C.A stored procedure that computes cosine similarity in application code for every item
4. D.A Cosmos DB trigger that recalculates similarity on every write

Whenever an order document is inserted or updated in Cosmos DB, a downstream service must be notified within seconds to trigger a fulfillment workflow, without polling the container on a timer.

Which approach is correct?

1. A.Implement a change feed processor with a lease container to react to inserts and updates in near real time
2. B.Poll the container every few seconds using a timestamp filter in the WHERE clause
3. C.Enable TTL on the container to trigger a notification on expiration
4. D.Use a stored procedure executed manually after each write

An Azure Database for PostgreSQL table with pgvector stores 10 million embeddings. Exact nearest-neighbor search has become too slow for the application's latency budget, and a small accuracy trade-off is acceptable.

What should you do?

1. A.Create an approximate nearest-neighbor index (such as HNSW) on the vector column instead of relying on a sequential scan
2. B.Increase the column's storage size to VARCHAR(MAX)
3. C.Partition the table by primary key range only
4. D.Switch the query to use a LIKE pattern match instead of vector distance

A RAG application queries PostgreSQL/pgvector for similarity search, but the same handful of queries repeat thousands of times per minute during peak traffic, adding load to the database.

What is the most appropriate use of Azure Managed Redis here?

1. A.Cache frequent query embeddings and their similarity results in Redis with an appropriate expiration, using Redis as a fast layer in front of PostgreSQL — not as the system of record
2. B.Replace PostgreSQL entirely and store all embeddings only in Redis with no expiration
3. C.Use Redis purely as a message queue between the app and PostgreSQL
4. D.Disable caching entirely and scale PostgreSQL compute instead

A background processor keeps failing to process a specific message due to malformed data, and Service Bus keeps redelivering it after each failure, consuming worker capacity.

What should happen once the message exceeds its maximum delivery count?

1. A.Service Bus automatically moves it to the dead-letter sub-queue, where it can be inspected and handled separately without blocking the main queue
2. B.The message is silently discarded by Service Bus
3. C.The consuming application must manually track delivery count using a custom database table
4. D.Service Bus pauses the entire queue until an administrator intervenes

Multiple event types flow through a single Event Grid topic, but a specific subscriber should only receive events where the subject starts with "orders/" and the event type is "OrderCreated".

How should you configure the subscription?

1. A.Add subject-begins-with and event-type filters directly on the Event Grid subscription
2. B.Create a separate topic per event type and subject prefix
3. C.Have the subscriber receive all events and filter them in application code
4. D.Use a Logic App as an intermediary to inspect every event before forwarding

You need a function that fires whenever a message arrives on a Service Bus queue, processes it, and writes the result as a new document in Cosmos DB — with minimal boilerplate code.

Which combination is correct?

1. A.A Service Bus trigger on the function, with a Cosmos DB output binding
2. B.An HTTP trigger that polls the Service Bus queue manually
3. C.A Timer trigger that checks the queue every minute
4. D.A Cosmos DB trigger that fires on the Service Bus queue

A business event ("InvoiceApproved") must be delivered to several independent downstream systems, each of which needs a durable, ordered subscription and the ability to reprocess a message if its own consumer was briefly offline.

Which service and construct best fits this requirement?

1. A.Azure Service Bus topics and subscriptions, giving each consumer its own durable subscription with independent message retrieval
2. B.Azure Event Grid with a single webhook endpoint shared by all consumers
3. C.Azure Queue Storage with one shared queue for all consumers
4. D.A single Service Bus queue polled by all consumers simultaneously

A database password stored in Key Vault must be rotated every 90 days automatically, and a Function App reading the secret must always retrieve the current version without requiring a redeploy after each rotation.

What should you configure?

1. A.A Key Vault rotation policy on the secret, with the Function App referencing the secret without pinning a specific version
2. B.A hard-coded secret version in the Function App configuration, updated manually every 90 days
3. C.A scheduled Function that redeploys the entire app every 90 days
4. D.Storing the password directly in application settings instead of Key Vault

An application needs both non-sensitive settings (feature flags, API endpoints) managed centrally and sensitive secrets (API keys) stored securely, with a single configuration read path in code.

What is the correct architecture?

1. A.Store non-sensitive settings in Azure App Configuration, and store secrets in Key Vault, with App Configuration holding Key Vault references so the app reads from one place
2. B.Store everything, including secrets, directly in App Configuration as plain values
3. C.Store everything in Key Vault, including feature flags and non-sensitive settings
4. D.Store all configuration in environment variables set manually per environment

A request flows from a Container Apps front end through an Azure Function to Cosmos DB, and overall latency is high. You need to see which hop in the chain contributes the most time.

What should you implement?

1. A.Instrument each service with the OpenTelemetry SDK and correlate the resulting distributed traces by trace ID across hops
2. B.Add console.log timestamps in each service and manually compare log timestamps after the fact
3. C.Enable verbose logging only on the Cosmos DB SDK
4. D.Increase the timeout setting on the Container Apps ingress

You need to find the three most frequent exception types logged by an application in the last 24 hours, using data already flowing into Log Analytics.

Which KQL query structure is correct?

1. A.exceptions | where timestamp > ago(24h) | summarize count() by type | top 3 by count_
2. B.exceptions | select type | filter timestamp = today()
3. C.SELECT type, COUNT(*) FROM exceptions WHERE timestamp > NOW() - 1 GROUP BY type LIMIT 3
4. D.exceptions | project type | distinct type | take 3

A Container Apps job's managed identity only needs to retrieve the current value of specific secrets at runtime. It must not be able to list all secrets in the vault or modify access policies.

Which access configuration is correct?

1. A.Assign the identity an RBAC role scoped to secret Get permission only (or an access policy limited to Get), not broader management permissions
2. B.Assign the identity the Key Vault Administrator role
3. C.Assign the identity Owner on the Key Vault's resource group
4. D.Grant the identity List and Get on all secret, key, and certificate types by default