Keep GPU resource-pool names consistent across clustered DDA hosts

Source facts

Microsoft documents dedicating physical GPUs to clustered VMs through DDA as distinct from GPU partitioning. The cluster uses a GPU resource pool to determine placement of VMs assigned to that pool. The PowerShell procedure creates a pool on each server and requires the same pool name on every server. Microsoft Learn.

Applicability

Confirm the supported cluster, GPU, and VM requirements in the current source before preparing pools. Inventory the available hardware on each intended node. Keep a matching pool label separate from proof that every node has suitable hardware and capacity.

DSE recommendation

Create a per-node GPU inventory linked to the intended shared pool name and VM assignments. Have the hardware and virtualization owners review the mapping together. Preserve the host device state and the VM preparation settings before the pilot. Define where the VM should be allowed to start or move and which observed condition will block placement.

Verification

Inspect pool names and membership on every intended host, then assign a representative prepared VM. Verify its actual GPU and workload behavior on the permitted placement paths. Exercise the agreed failover scenario and record the selected node and hardware. Resolve inconsistent names, missing devices, or unexpected placement before assigning more workloads.

Official references

Microsoft Learn: Use GPUs with Discrete Device Assignment in clustered VMs on Windows Server and Azure Local. Source reviewed September 8, 2026.