Monitor and maintain an Azure Virtual Desktop infrastructure
Turn on telemetry first
AVD does not send diagnostics anywhere by default. Create diagnostic settings on the host pool, workspace, and application group to send checkpoints (Connection, HostRegistration, AgentHealth) and error/warning events to a Log Analytics workspace. Only then do the AVD Insights workbook, custom queries, and alerts have data. Add VM Insights for CPU, memory, and disk on the session hosts.
Day-two operations
Use drain mode before patching or image updates so no new sessions land on a host, then replace it from a new Compute Gallery image version. Scaling plans define ramp-up, peak, ramp-down, and off-peak schedules with capacity thresholds; behavior differs between pooled and personal pools. Alert on host availability, CPU sustained above threshold, and sessions approaching the max session limit.
Troubleshooting with checkpoints
User connection diagnostics shows where a failed connection stopped: the service, the gateway, or the session host. "No session hosts available" usually means every host is in drain mode or at max sessions. Agent health "Unavailable" means the agent cannot reach the service - check outbound 443, DNS, proxy, and time sync. A black screen after connect points to graphics, agent, or profile issues.
Traps
Enabling diagnostics on only one resource type leaves gaps in the workbook; Insights without a Log Analytics workspace shows nothing; scaling plans must be assigned to the host pool; and an alert rule without an action group notifies nobody.
Worked example: alert when session count exceeds 90% of (max session limit × healthy hosts) for 15 minutes, routed to an action group that opens a ticket and adds an N+1 host.