Platform Support
Overview
Use this page to confirm that your hardware, operating system, and Kubernetes platform are supported by NVIDIA Network Operator before deployment.
Support terms used on this page:
Supported — A configuration that NVIDIA maintains and backs for this release.
Validated — A hardware or software configuration that NVIDIA has tested with this release.
GA (Generally Available) — Production-ready support tier.
Tech Preview — Limited testing; not recommended for production deployments.
Versioning and Lifecycle
NVIDIA Network Operator uses calendar versioning in the form YY.MM.PP (for example, 26.7.0). The first two fields identify the major version and release timeframe; the third field identifies the patch version, used for critical bug and CVE fixes.
When a new major version is released, the previous major version enters a Deprecated state and receives only patch updates for critical fixes. Earlier major versions reach End of Support and no longer receive updates.
Network Operator Version |
Status |
|---|---|
26.7.x |
Supported |
26.4.x |
Deprecated |
26.1.x and lower |
End of Support |
Upgrades are supported within a major release, or to the next major release. For the full support-status table, upgrade procedures, and operational guidance, see Life Cycle Management.
Note
The product lifecycle and versioning are subject to change in future releases.
Important Limitations
Review the following limitations before deployment:
Firmware reset on BMC-controlled platforms — NVIDIA DGX/HGX GB200 NVL72, B200, and B300 systems require additional configuration so that firmware updates apply without a reboot loop. See Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8).
Precompiled DOCA-OFED driver containers are currently unsigned.
Precompiled DOCA-OFED kernel-flavor support — Only the Ubuntu
generickernel flavor is GA. Thenvidia,aws,azure, andoracleflavors are Tech Preview.For per-release known issues, see the Release Notes.
Prerequisites
Component |
Version |
Notes |
|---|---|---|
Kubernetes |
>=1.32 and <=1.36 |
|
Helm |
v3.5+ |
For installation methods, refer to the official Helm website. |
Node Feature Discovery |
>=0.15.6 and <=0.17.0 |
When deploying the Network Operator and GPU Operator on the same cluster, ensure only one instance of Node Feature Discovery (NFD) is installed. We recommend using the version included with the GPU Operator. |
System Requirements
- RDMA-capable NVIDIA network adapters
NVIDIA ConnectX NICs and SuperNICs
NVIDIA BlueField Networking Platforms
NVIDIA GPU Operator v25.3.x or newer – required for workloads that use NVIDIA GPUs and GPUDirect RDMA. See Support for GPUDirect RDMA.
Supported NVIDIA Network Adapters
The following adapters have been tested and validated with NVIDIA Network Operator:
Product Family |
Ethernet |
InfiniBand |
Max Port Speed |
Validated Protocols / Modes |
Notes |
|---|---|---|---|---|---|
NVIDIA ConnectX-6 NIC |
Yes |
Yes |
200 Gb/s |
IB RDMA, RoCE |
— |
NVIDIA ConnectX-6 Dx NIC |
Yes |
No |
200 Gb/s |
RoCE |
— |
NVIDIA ConnectX-7 NIC |
Yes |
Yes |
400 Gb/s |
IB RDMA, RoCE |
— |
NVIDIA ConnectX-8 SuperNIC |
Yes |
Yes |
800 Gb/s |
IB RDMA, RoCE |
— |
NVIDIA ConnectX-9 SuperNIC |
Yes |
Yes |
800 Gb/s |
IB RDMA, RoCE |
— |
NVIDIA BlueField-3 DPU |
Yes |
No |
200 Gb/s |
RoCE |
NIC mode only |
NVIDIA BlueField-3 SuperNIC |
Yes |
No |
400 Gb/s |
RoCE |
NIC mode only |
Supported NVIDIA Data Center Systems
The following NVIDIA Data Center systems have been tested and validated with NVIDIA Network Operator. For the supported versions of each operating system, see Supported Operating Systems and Kubernetes Platforms. On Arm-based systems, only the Arm64 builds of the listed distributions are validated; not every operating system version in that table is available for Arm64.
System |
CPU Architecture |
GPU Architecture |
Network Adapter(s) |
Operating System(s) |
Notes |
|---|---|---|---|---|---|
NVIDIA Grace ARM Server |
Arm (NVIDIA Grace) |
NVIDIA Hopper |
BlueField-3 (NIC Mode) |
Ubuntu / Red Hat OpenShift / SUSE Linux Enterprise Server |
GA (RoCE only, without GPUDirect RDMA) |
NVIDIA IGX Orin |
Arm (NVIDIA Orin) |
NVIDIA Ampere |
ConnectX-7 |
Ubuntu |
GA (RoCE only, without GPUDirect RDMA) |
NVIDIA RTX PRO 6000 Blackwell Server |
x86 |
NVIDIA Blackwell |
BlueField-3 SuperNIC (NIC mode) / ConnectX-8 |
Ubuntu 22.04 / 24.04 (x86) / Red Hat OpenShift |
GA |
NVIDIA DGX/HGX B200 |
x86 |
NVIDIA Blackwell |
BlueField-3 SuperNIC (NIC mode) / ConnectX-7 |
Ubuntu 22.04 / 24.04 (x86) / Red Hat OpenShift |
GA. Firmware reset required (see Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8)) |
NVIDIA DGX/HGX B300 |
x86 |
NVIDIA Blackwell |
ConnectX-8 SuperNIC |
Ubuntu 22.04 / 24.04 (x86) / Red Hat OpenShift |
GA. Firmware reset required (see Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8)) |
NVIDIA DGX/HGX GB200 NVL72 |
Arm (NVIDIA Grace) |
NVIDIA Blackwell |
ConnectX-7 |
Ubuntu 24.04 (Arm64) / Red Hat OpenShift |
GA. Firmware reset required (see Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8)) |
NVIDIA DGX/HGX GB300 NVL72 |
Arm (NVIDIA Grace) |
NVIDIA Blackwell |
ConnectX-8 |
Ubuntu 24.04 (Arm64) / Red Hat OpenShift |
GA. Firmware reset required (see Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8)) |
NVIDIA DGX GB300 Workstation |
Arm (NVIDIA Grace) |
NVIDIA Blackwell |
ConnectX-8 SuperNIC |
Ubuntu 24.04, 22.04 (Arm64) |
GA (RoCE only, without InfiniBand) |
NVIDIA Vera Rubin NVL72 |
Arm (NVIDIA Vera) |
NVIDIA Rubin |
ConnectX-9 SuperNIC |
Ubuntu 24.04, 22.04 (Arm64) |
Tech Preview — system validation is in progress. Uses ConnectX-9 SuperNIC, which is generally available for RoCE and InfiniBand workloads; see Supported NVIDIA Network Adapters. Firmware reset required (see Platforms with external BMC (DGX or HGX GB200/GB300, B200/B300, Vera Rubin NVL72, Rubin NVL8)) |
Supported Deployment Options
NVIDIA Network Operator supports the deployment models below, and supports the DOCA-OFED driver being either pre-installed on the host or deployed as a container.
Cluster Deployment Models
Deployment Model |
Description |
Guide |
|---|---|---|
Bare metal |
Worker nodes run directly on physical hosts. Network Operator manages the NICs on those hosts. |
|
Virtual machine with PCIe passthrough |
A full NIC or an individual Virtual Function is passed through to the virtual machine. Network Operator runs inside the virtual machine and manages the passed-through device. |
|
Virtual machine with SR-IOV |
Network Operator runs in the host cluster and allocates Virtual Functions to KubeVirt or Red Hat OpenShift Virtualization virtual machines through VFIO PCI passthrough. |
KubeVirt SR-IOV Integration and OpenShift Virtualization SR-IOV Integration |
DOCA-OFED Driver Delivery
Every NVIDIA Network Operator release supports both the latest DOCA-OFED GA driver and the latest DOCA-OFED LTS driver. For this release those are doca3.5.0-26.07-0.7.7.0-0 (GA) and doca3.2.2-25.10-2.4.1.0-4 (LTS).
Both delivery methods are equally supported. The DOCA-OFED driver may be pre-installed on the host from the DOCA-Host package, or deployed as the containerized DOCA-OFED driver by Network Operator.
Delivery Method |
Installed and Upgraded By |
Supported Operating Systems |
|---|---|---|
DOCA-Host (pre-installed) |
The host provisioning process, before Kubernetes and Network Operator are installed. Outside Network Operator scope — see the NVIDIA DOCA Installation Guide for Linux. |
See Supported Host OS per DOCA-Host Installation Profile in the NVIDIA DOCA documentation. |
DOCA-OFED driver container ( |
Network Operator, through |
Supported Operating Systems and Kubernetes Platforms on this page. |
Important
DOCA-Host (host-installed) and the containerized DOCA-OFED driver are mutually exclusive. Choose one per node; do not deploy both.
Recommended delivery method by host operating system type:
Host OS Type |
Recommended |
Rationale |
|---|---|---|
Mutable — Ubuntu, Red Hat Enterprise Linux, SUSE Linux Enterprise Server |
Either |
Choose DOCA-Host to install and version drivers separately from the cluster; it is a prerequisite completed before Kubernetes and Network Operator are installed. Choose the driver container for a cloud-native workflow that keeps the driver on the same lifecycle as the rest of the stack. |
Immutable — Red Hat CoreOS with Red Hat OpenShift |
Driver container |
Host packages cannot be installed, so DOCA-Host is not directly available. The driver container is the only way to deliver DOCA kernel drivers. |
Note
Precompiled driver container images are available for a subset of operating systems and kernel flavors — see Supported Precompiled Container Images for DOCA-OFED Drivers.
Supported Operating Systems and Kubernetes Platforms
NVIDIA Network Operator has been validated on the following OS and platform combinations for Ethernet (RoCE) and InfiniBand (IB RDMA):
Note
Kubernetes support for the NVIDIA Spectrum-X Reference Architecture (RA) is limited to a subset of OS and platform combinations. For details, refer to the NVIDIA Spectrum-X documentation.
Operating System |
Upstream Kubernetes |
Red Hat OpenShift |
Rancher RKE2 |
Canonical MicroK8s |
Nutanix NKP |
Container Runtime |
Notes |
|---|---|---|---|---|---|---|---|
Ubuntu 26.04 LTS |
1.32–1.36 |
— |
— |
1.32–1.36 |
2.12–2.15 |
Containerd |
— |
Ubuntu 24.04 LTS |
1.32–1.36 |
— |
— |
1.32–1.36 |
2.12–2.15 |
Containerd |
— |
Ubuntu 22.04 LTS |
1.32–1.36 |
— |
— |
1.32–1.36 |
2.12–2.15 |
Containerd |
RT kernel support |
Red Hat CoreOS |
— |
4.17–4.22 |
— |
— |
— |
CRI-O |
RT kernel support |
Red Hat Enterprise Linux 10.2 / 10.0 / 9.8 / 9.6 / 9.4 |
1.32–1.36 |
— |
— |
— |
— |
Containerd, CRI-O |
RT kernel support |
Red Hat Enterprise Linux 8.10 / 8.8 |
1.32–1.36 |
— |
— |
— |
— |
Containerd, CRI-O |
RT kernel support |
SUSE Linux Enterprise Server 15 SP7 |
1.32–1.36 |
— |
1.32–1.36 |
— |
— |
Containerd |
Supported for Kubernetes and Rancher deployments |
Supported Container Runtimes
NVIDIA Network Operator has been validated in the following scenarios:
Operating System |
Containerd |
CRI-O |
Notes |
|---|---|---|---|
Ubuntu 26.04 LTS |
Yes |
No |
|
Ubuntu 24.04 LTS |
Yes |
No |
|
Ubuntu 22.04 LTS |
Yes |
No |
|
Red Hat CoreOS |
No |
Yes |
|
Red Hat Enterprise Linux 9 |
Yes |
Yes |
|
Red Hat Enterprise Linux 8 |
Yes |
Yes |
|
Red Hat Enterprise Linux 10 |
Yes |
Yes |
|
SUSE Linux Enterprise Server 15 SP7 |
Yes |
No |
NVIDIA Spectrum-X Ethernet Networking Platform
NVIDIA Network Operator has been validated with the following NVIDIA Spectrum-X Reference Architecture (RA) versions:
Note
For details on supported topologies, NIC hardware, software components, and version-specific notes, refer to the NVIDIA Spectrum-X documentation.
Spectrum-X RA Version |
Support Level |
Topologies and NIC Hardware |
Kubernetes Versions |
Operating Systems |
Notes |
|---|---|---|---|---|---|
2.3 |
GA on 26.7.x |
|
Upstream Kubernetes (1.35–1.36) |
|
Hardware Multiplane (recommended) and Software Multiplane |
2.1 |
Tech Preview on 26.4.x (GA on 26.1.x) |
|
Upstream Kubernetes (1.32–1.36) |
|
Small RA, Software Multi-Plane |
Note
NVIDIA ConnectX-9 SuperNIC is not part of Spectrum-X RA 2.3. Network Operator accepts ConnectX-9 SuperNIC for Spectrum-X NIC configuration as a Tech Preview — available for evaluation, but not an RA-validated or production-ready Spectrum-X configuration, and it requires a Spectrum-X profile covering ConnectX-9 SuperNIC, available from NVIDIA Support. For general RoCE and InfiniBand workloads, ConnectX-9 SuperNIC is fully supported; see the NIC table above.
Support for KubeVirt SR-IOV Passthrough
NVIDIA Network Operator supports attaching SR-IOV Virtual Functions (VFs) to KubeVirt and Red Hat OpenShift Virtualization virtual machines via VFIO PCI passthrough. For deployment instructions, see KubeVirt SR-IOV Integration on upstream Kubernetes, or OpenShift Virtualization SR-IOV Integration on Red Hat OpenShift.
Operating System |
Kubernetes |
KubeVirt |
Red Hat OpenShift Virtualization |
|---|---|---|---|
Ubuntu 26.04 LTS |
1.32–1.36 |
v1.8.2+ |
— |
Ubuntu 24.04 LTS |
1.32–1.36 |
v1.8.2+ |
— |
Ubuntu 22.04 LTS |
1.32–1.36 |
v1.8.2+ |
— |
Red Hat CoreOS |
— |
— |
4.17–4.22 |
Key limitations:
VFIO passthrough requires IOMMU-capable hardware (Intel VT-d or AMD-Vi) with IOMMU enabled at boot.
Live migration is not supported for VMs that use SR-IOV VFIO passthrough.
Host-side RDMA is not available for VFIO-passed VFs; RDMA inside the guest is provided by the
mlx5_coredriver shipped with the guest image.
Support for GPUDirect RDMA
NVIDIA Network Operator enables GPUDirect RDMA between NVIDIA GPUs and supported NICs/SuperNICs when deployed together with NVIDIA GPU Operator. For deployment instructions, see Network Operator Deployment for GPUDirect Workloads. For GPU Operator-side support details, see NVIDIA GPU Operator Platform Support — Support for GPUDirect RDMA.
Requirements:
NVIDIA GPU Operator v25.3.x or newer
NVIDIA DOCA-OFED Driver doca3.1.0-25.07-0.9.7.0-0 or newer
nvidia_peermemkernel module (auto-loaded by recent NVIDIA GPU drivers)Supported NVIDIA GPU (Ampere, Hopper, Blackwell, or Rubin)
Supported NVIDIA NICs (NVIDIA ConnectX or BlueField)
Operating System |
Kubernetes |
|---|---|
Ubuntu 26.04 LTS |
1.32–1.36 |
Ubuntu 24.04 LTS |
1.32–1.36 |
Ubuntu 22.04 LTS |
1.32–1.36 |
Red Hat Enterprise Linux 10.2 / 10.0 / 9.8 / 9.6 / 9.4 |
1.32–1.36 |
Red Hat Enterprise Linux 8.10 / 8.8 |
1.32–1.36 |
Red Hat CoreOS |
OpenShift 4.17–4.22 |
For background on the underlying technology, see the NVIDIA GPUDirect RDMA documentation.
Support for GPUDirect Storage
NVIDIA Network Operator provides the RDMA networking fabric (RoCE and InfiniBand) required by GPUDirect Storage (GDS). GDS-compatible storage clients operate over the RDMA fabric provisioned by Network Operator.
Warning
With the DOCA-OFED driver container and ENABLE_NFSRDMA=true, host-side NVMe-over-RDMA cannot run concurrently — DOCA-OFED’s ib_core conflicts with the inbox nvme_rdma module. The DOCA-Host package install path is unaffected, as is local-NVMe storage.
For configuration details (including the ENABLE_NFSRDMA setting and the NVMe inbox kernel module workaround), see DOCA-OFED Driver Container. For background on the underlying technology and the list of GDS-compatible storage clients and version requirements, see the NVIDIA GPUDirect Storage documentation.
Supported Precompiled Container Images for DOCA-OFED Drivers
Overview
To save startup time and operational effort, precompiled DOCA-OFED driver container images are available for common OS/flavor/kernel/architecture variants.
The container image tag pattern used for common variants is: driver_ver-container_ver-kernel_ver-flavor-os-arch. For example: 24.07-0.6.1.0-0-6.8.0-49-generic-ubuntu24.04-amd64
Note
For the generic flavor of Ubuntu, the default kernel version is used for precompiling (for example, 6.8.0-31 for Ubuntu 24.04). For all other flavors, the latest kernel version available at the time of DOCA packaging and release is used.
Supported Operating Systems
Currently precompiled DOCA-OFED driver container images are provided for the following operating systems:
Ubuntu 24.04 (amd64/arm64)
Ubuntu 22.04 (amd64/arm64)
Limitations
NVIDIA supports precompiled driver containers for the most recently released DOCA-OFED GA drivers.
NVIDIA builds precompiled driver containers for
generic,nvidia,aws,azure, andoraclekernel flavors.Precompiled driver containers are currently unsigned.
If your hosts use a different kernel variant, you can create a custom precompiled driver container and host it in your own container registry. Please refer to the Precompiled Container Build Instructions for NVIDIA DOCA-OFED Driver Container section.
Warning
Only
generickernel variant is tested and supported as a GA.nvidia,aws,azure, andoraclekernel variants are supported as a Tech Preview and have limited testing.
Additional Supported Tools and Integrations
Container management tools:
Deployment tooling:
NVIDIA Kubernetes Launch Kit — CLI that discovers cluster hardware and generates Network Operator manifests. See NVIDIA Kubernetes Launch Kit.
Orchestration & resource scheduling:
Note
Run:ai requires the NVIDIA Network Operator as a prerequisite. To configure NVIDIA Network Operator refer to the Run:ai cluster requirements documentation for more information.
Network Operator Component Matrix
The following components are shipped and versioned with NVIDIA Network Operator:
Operators
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
NVIDIA (OSS) |
nvcr.io/nvidia/cloud-native |
network-operator |
v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
network-operator-init-container |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
nic-configuration-operator |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
nic-configuration-operator-daemon |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
maintenance-operator |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
spectrum-x-operator |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
sriov-network-operator |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
sriov-network-operator-webhook |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
sriov-network-operator-config-daemon |
network-operator-v26.7.0 |
Yes |
Drivers and Services
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
NVIDIA (EULA) |
nvcr.io/nvidia/mellanox |
doca-driver |
doca3.5.0-26.07-0.7.7.0-0 |
Yes |
LTS version: doca3.2.2-25.10-2.4.1.0-4 |
|
NVIDIA (EULA) |
nvcr.io/nvidia/doca |
xplane |
3.5.0041 |
Yes |
Deployed for Spectrum-X Hardware Multiplane. The version tracks the DOCA release, not the Network Operator release. |
|
NVIDIA (EULA) |
nvcr.io/nvidia/doca |
doca_telemetry |
1.25.5-doca3.4.0-host |
Yes |
Networking Plugins
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
Community (OSS) |
nvcr.io/nvidia/mellanox |
multus-cni |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
plugins |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
sriov-cni |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
ib-sriov-cni |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
ipoib-cni |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
rdma-cni |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
ovs-cni-plugin |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
nvidia-k8s-ipam |
network-operator-v26.7.0 |
Yes |
||
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
ib-kubernetes |
network-operator-v26.7.0 |
Yes |
Device Plugins and Resource Allocation
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
k8s-rdma-shared-dev-plugin |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
sriov-network-device-plugin |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
dra-driver-sriov |
network-operator-v26.7.0 |
No |
Tech Preview. See DRA SR-IOV Driver. |
Discovery
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
NVIDIA (OSS) |
nvcr.io/nvidia/mellanox |
nic-feature-discovery |
network-operator-v26.7.0 |
Yes |
||
Community (OSS) |
nvcr.io/nvidia/mellanox |
node-feature-discovery |
network-operator-v26.7.0 |
Yes |
Optionally deployed. May already be present in the cluster with proper configuration. |
Tools
Component |
Origin |
Repository |
Image Name |
Tag |
NVAIE |
Notes |
|---|---|---|---|---|---|---|
NVIDIA (OSS) |
nvcr.io/nvidia/cloud-native |
k8s-launch-kit |
v26.7.0 |
Yes |
CLI tool that generates manifests and installs Network Operator; not deployed by the operator. See NVIDIA Kubernetes Launch Kit. |