problem
I am experiencing an issue where an NVIDIA H200 GPU allocated to a VM via a Custom Service Offering (Passthrough) is not being correctly tracked by the Management Server. While the VM is Running and the device is attached via vfio-pci on the KVM host, the CloudStack UI reverts the GPU status from "Allocated" to "Free" within 1–2 minutes of VM start.
Environment
CloudStack Version: 4.22.0.0
Hypervisor: KVM (Ubuntu 24.04)
versions
The versions of ACS, hypervisors, storage, network etc..
The steps to reproduce the bug
...
What to do about it?
No response