Microsoft · AI-500
Reliability, availability, and tracing in Foundry
AI-500 · Lesson 13 of 16 · 1:45:13
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-500 exam (Microsoft Certified: Multi-Agent AI Solutions Expert). In this lesson: a reliability monitoring strategy (agent health, workflow failures, trace correlation, drift, quality regressions, remediation), platform availability, service-level objectives, and why preview features carry no SLA, tracing in Foundry with OpenTelemetry, correlation IDs, protected sensitive content, the Monitor tab, and Trace Replay (preview), and alerting with severity tiers, dynamic thresholds, the Foundry Control Plane, and blameless postmortems. Includes three original practice questions.
Preview the topics
- Contoso promised its corporate customers 99.9% availability, but its booking agent relies on a feature that is still in preview. What should the architect flag?
- Contoso’s support engineers must investigate failed runs in Foundry traces, but only the privacy team may read prompts and responses. What should the team configure?
- Contoso’s on-call engineers receive two hundred alerts a night, most of them harmless, and they have started ignoring them. What should the team change?