Microsoft AZ-204: Developing Solutions for Microsoft Azure

Last checked October 22, 2026 against the official Microsoft exam page.

1. Question 1Develop Azure compute solutions

   A team needs to run a containerised HTTP API that scales to zero when idle and requires no cluster management. Which service fits best?

   1. AAzure Virtual Machines with a manually configured Docker host that stays running overnight and at weekends
   2. BAzure Kubernetes Service with a fixed node pool that must keep running at all times, even when no requests arrive
   3. CAzure Container Apps, which runs containers with managed scaling, including scale to zero for HTTP workloads
   4. DAzure Batch, which is designed for running large parallel compute jobs rather than serving HTTP requests from users
   Show answer and explanationHide answer

   Correct answer: C

   Container Apps provides managed scaling, including scale to zero, for HTTP workloads without cluster management. AKS keeps nodes running, virtual machines need manual management, and Batch targets batch jobs rather than serving requests.

2. Question 2Connect to and consume Azure services and third-party services

   Which Azure API Management policy limits how many calls a client can make during a period?

   1. AThe cache-lookup policy, which returns cached responses but does not restrict how many calls clients are allowed to make
   2. BThe set-header policy, which adds or changes HTTP headers in requests and responses but does not limit how many calls are made
   3. CThe rate-limit or quota policy, which restricts calls per key or subscription over a defined time period
   4. DThe mock-response policy, which returns fixed responses for testing and does not restrict calls by client in production use
   Show answer and explanationHide answer

   Correct answer: C

   Rate-limit and quota policies restrict call volume per key or subscription over time. Header, caching and mock policies serve other purposes.

3. Question 3Develop for Azure storage

   Which Cosmos DB consistency level gives the strongest guarantee that reads return the most recent committed write?

   1. AEventual consistency, which allows reads to return older values for a period and offers the lowest latency of the levels
   2. BStrong consistency, which guarantees that reads return the most recent committed write, at some latency cost
   3. CSession consistency, which guarantees consistency within a single session but is weaker across different sessions of users
   4. DConsistent prefix, which guarantees that reads never see writes out of order but may be behind the latest committed write
   Show answer and explanationHide answer

   Correct answer: B

   Strong consistency guarantees reads return the latest committed write, which is the strongest level. The others trade that guarantee for lower latency or session-scoped behaviour.

4. Question 4Implement Azure security

   Which approach lets an app read Key Vault secrets without storing credentials in its code or configuration?

   1. AA personal access token in the application's configuration file, which the developer renews by hand whenever it expires
   2. BA plain text connection string stored in the code file, which the team rotates manually each quarter of the year
   3. CA shared access signature embedded in the source repository, which expires after one year of use by the team
   4. DA managed identity assigned to the app, granted a Key Vault access role or policy for the secrets it needs
   Show answer and explanationHide answer

   Correct answer: D

   A managed identity lets the app authenticate to Key Vault without stored credentials, and an access role limits what it can read. Embedded connection strings, repository SAS tokens and personal tokens all store credentials.

5. Question 5Monitor and troubleshoot Azure solutions

   Which Application Insights feature runs scheduled requests from several locations to check that an endpoint responds?

   1. AAvailability tests, which run from multiple locations on a schedule and alert when the endpoint fails
   2. BLive Metrics only, which streams the telemetry of running instances but does not send requests itself to endpoints
   3. CProfiler, which samples the running code to find slow methods but does not check the availability of an endpoint
   4. DSmart detection of anomalies in exceptions, which alerts on unusual patterns but does not probe the endpoint itself
   Show answer and explanationHide answer

   Correct answer: A

   Availability tests probe endpoints from several locations on a schedule and alert on failures. Live Metrics, Profiler and smart detection do not probe endpoints in that way.