AI-300 Practice Question 57: Throttling Metric Alert
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 6 · question 7 of 10
An operations team wants to be notified as soon as a model deployment in Microsoft Foundry starts returning throttling responses, before users complain. Which approach should they use?
Show answer and explanation
Answer:
Foundry Models publish request metrics to Azure Monitor, and the request-count metric can be split by the StatusCode dimension. A metric alert rule that filters this metric to 429 fires when throttling starts, so the team learns about it in near real time (option B).
- A is wrong because the Audit category records administrative operations such as configuration changes. It does not record throttled inference calls. - C is wrong because latency can change for many reasons, and a throttled request returns quickly with an error. It is not a reliable signal for 429 responses. - D is wrong because budget alerts track spend and can arrive days later. They say nothing about rate limits being hit right now.
Reference: Monitor Foundry Models