Skip to content

Fix fmpm verif #476

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mikemckiernan merged 1 commit into from
Aug 21, 2026
Merged

Fix fmpm verif #476

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
43 changes: 35 additions & 8 deletions gpu-operator/gpu-operator-kubevirt-dra.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -508,18 +508,45 @@ advertises the required ``partitionN`` attribute.

#. Optional: Verification that relies on additional software.

#. Refer to the `Fabric-Manager-Client <https://github.com/NVIDIA/Fabric-Manager-Client>`__ repository
for information about building the `fmpm` utility.
#. Build and run the ``list-partitions`` binary.

You can use the `fmpm` utility to list the partitions and verify the partition status.
Perform the build on a Linux system with Go 1.24 or later and the same CPU
architecture as the target GPU node.

.. code-block:: console
#. Clone the `go-nvfm <https://github.com/NVIDIA/go-nvfm/tree/main/examples/list-partitions>`__ repository and build the example:

.. code-block:: console

$ git clone --depth 1 https://github.com/NVIDIA/go-nvfm.git
$ cd go-nvfm
$ make example-list-partitions

#. Identify the driver pod on the node that owns the allocated GPUs:

.. code-block:: console

$ kubectl get pods --namespace gpu-operator \
--selector app.kubernetes.io/component=nvidia-driver \
--field-selector spec.nodeName=<node-name>

#. Copy the ``list-partitions`` executable to the driver container:

.. code-block:: console

$ kubectl cp --namespace gpu-operator \
--container nvidia-driver-ctr \
./list-partitions <driver-pod>:/run/nvidia/list-partitions

#. List the Fabric Manager partitions:

.. code-block:: console

$ /run/nvidia/fmpm \
--unix-domain-socket /run/nvidia-fabricmanager/socket \
-l
$ kubectl exec --namespace gpu-operator \
--container nvidia-driver-ctr \
<driver-pod> -- bash -c \
"NVFM_UNIX_SOCKET_PATH=/run/nvidia-fabricmanager/socket /run/nvidia/list-partitions"

Confirm that the partition selected for the claim reports ``isActive: 1``.
Confirm that the partition selected for the claim reports ``"isActive": 1``.

#. For an additional connectivity demonstration, install a guest-compatible build of
`nvbandwidth v0.8 <https://github.com/NVIDIA/nvbandwidth/releases/tag/v0.8>`__ and run:
Expand Down