AI-103: Developing AI Apps and Agents on Azure

By Microsoft · Question Mode

QuestionQ30

VerifiedPlan and manage an Azure AI solution

You have a Microsoft Foundry project containing an internal Q&A agent.

Users report the following issues when they ask the agent questions:

  • An increasing number of "No relevant information found" responses.
  • Periodic HTTP 429 rate limit exceeded errors during peak hours.

You need to determine whether each issue results from model unavailability, resource limits, or inference failures. What should you do?

Select
Metrics to enable:
Diagnostic log to collect:
Explanation

Enabling the Model Availability Rate metric shows drops in model uptime that explain inference/unavailability-driven failures such as a rise in "No relevant information found" responses, while the Provisioned Utilization metric shows how much of the deployment's provisioned capacity is being consumed, which explains HTTP 429 rate-limit errors during peak load caused by resource limits. Collecting the RequestResponse diagnostic log records each call's operation name, duration, and HTTP status code, letting you correlate specific failed or throttled calls (including 429s) back to the metric-level signal for root-cause analysis; the Audit and Trace log categories do not provide this per-call status/duration detail, and AllMetrics is a metrics export category rather than a diagnostic log.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!