Check the disk controller before interpreting missing Azure latency metrics

Source facts

Azure’s OS Disk Latency and Data Disk Latency metrics are preview measurements available for SCSI-attached disks, not disks attached through an NVMe controller. The documented temporary-disk latency, queue-depth, read and write metrics are also unavailable for NVMe temporary storage. Disk performance metrics do not record individual create, read, update or delete operations inside a managed disk. Microsoft Learn.

Applicability

Use this check while reviewing a monitoring gap after choosing or changing a VM’s storage controller. Separate an unsupported measurement from an observed value of zero, and do not describe aggregate performance monitoring as a file-operation audit.

DSE recommendation

Check metric availability for the actual controller before treating a missing series as a healthy result. Have the monitoring owner record the controller, disk role and precise requested metric before investigating collection failure. Identify a supported measurement path for the workload’s performance question and a separately authorized audit approach if individual data access is the concern. Keep unavailable signals visibly unresolved instead of filling their chart gaps with assumed healthy values.

Verification

On a controlled VM, confirm the actual disk/controller configuration and inspect which series are emitted. Compare supported observations with the application’s measured behavior during the same interval. Document any excluded metric so a later dashboard consumer knows why it is absent. Do not use a successful VM start or a populated neighboring chart as evidence that an unsupported latency metric has become available.

Official references

Microsoft Learn: Disk performance metrics. Source reviewed September 9, 2026.