Nebius Managed Kubernetes provides the cluster foundation and optional managed GPU settings. Your platform team still selects the node-group shape, driver mode, storage, interconnect, identity, labels, taints, and scaling behavior that define each capacity class.

See the Nebius Managed Kubernetes for the upstream configuration and requirements.

Prerequisites

• A Nebius Managed Kubernetes cluster with a GPU node group and reviewed NVIDIA driver installation mode.
• The required interconnect, storage, registry, DNS, service account, labels, taints, and autoscaling configuration.
• The cluster connected to Polyaxon by installing Polyaxon Agent (compute cluster), with namespace access and a queue or preset selecting that node group.

Connect the cluster

Follow Connect Kubernetes to Polyaxon to install Polyaxon Agent (compute cluster), configure storage and queues, and submit the shared GPU visibility example. Use the same namespace and queue that will run your workloads.

The example only checks GPU allocation and driver visibility. Validate storage, identity, and multi-node networking separately using the provider requirements below.

Provider requirements

• Record the node-group shape, driver mode, labels, taints, storage, and interconnect as one validated capacity class.
• Keep cluster, node-group, compute-cluster connection, queue, and preset configuration in version control.
• Revalidate the GPU, storage, identity, DNS, and network path after node-image, driver, or cluster upgrades.
• Run collective communication tests only on the exact node group and interfaces used by distributed workloads.

For distributed training, continue with NCCL and RCCL validation.

For help choosing a provider, see Choosing GPU infrastructure for Polyaxon.

Troubleshooting

The GPU resource is missing

Check node-group readiness, the selected driver installation mode, GPU operator or device-plugin status, and nvidia.com/gpu in node capacity.

Storage or interconnect devices are unavailable

Verify that the selected node group supports the requested feature, then inspect CSI or network components, topology constraints, mounts, and pod placement.

The connected compute cluster cannot submit or collect logs

Verify the Polyaxon Agent deployment, namespace RBAC, service-account permissions, registry access, queue routing, network policy, and control-plane connectivity.

References

• Nebius Managed Kubernetes — Provider documentation for cluster, node group, networking, and workload configuration.
• NVIDIA CUDA container — Published CUDA container tags used for the GPU visibility check.
• Nebius node groups — Node-group sizing, labels, taints, scaling, lifecycle, and GPU settings.
• Nebius GPU setup — Supported GPU driver installation modes and configuration choices.
• Polyaxon job runtime — Job containers, resources, connections, initialization, and lifecycle settings.
• Polyaxon resource scheduling — Kubernetes requests, limits, custom resources, queues, and placement controls.

---