Alauda GPU Management

Versions and Components

This page lists the maintained Alauda GPU Management version and the product components users need to consider during installation, upgrade, and support.

TOC

Current versionHistorical delivery modelv26.3.3-1 component stackSupported architectureACP integration points

Current version

ComponentProduct versionDelivery modelUpstream baseline
Alauda GPU Managementv26.3.3-1OperatorHub / OLM operatorNVIDIA gpu-operator

Historical delivery model

Earlier releases delivered GPU capabilities as separate cluster plugins. Alauda GPU Management replaces them with a single operator. Use this table to map the previous standalone components to the unified operator.

Previous deliveryVersion lineNow delivered by
Standalone NVIDIA GPU Device Plugin (cluster plugin)v0.18.xAlauda GPU Management device plugin operand
Standalone DCGM-Exporter (cluster plugin)separateAlauda GPU Management DCGM-Exporter operand (still referenceable by name)
Out-of-band driver and container runtime setup—Alauda GPU Management driver and container toolkit operands

The version advances to v26.3.x because the product now tracks the upstream GPU Operator line rather than the device-plugin line. In-place migration from the standalone cluster-plugin model to the operator is not an operator upgrade: uninstall the standalone plugins, then install the operator.

v26.3.3-1 component stack

All components below are included in the product package, except where noted as package-external.

ComponentVersionDefaultUser-visible purpose
Alauda GPU Managementv26.3.3-1EnabledReconciles ClusterPolicy and the managed GPU components.
NVIDIA Container Toolkitv1.19.1-3EnabledConfigures the containerd nvidia runtime handler and CDI.
NVIDIA Device Pluginv0.19.3-1EnabledAdvertises nvidia.com/gpu to Kubernetes.
GPU Feature Discoveryv0.19.3-1EnabledLabels nodes with GPU model and driver facts. Shares the device plugin image.
DCGM-Exporterv4.5.3-482.2EnabledExposes GPU metrics for ACP monitoring.
Driver Managerv0.11.0-1Enabled with the driverCoordinates driver rollout, node drain, and upgrade sequencing.
NVIDIA Driver (precompiled tuple)580.126.09.1 (SUSE Linux Micro 5.5, kernel 5.14.21-150500.55.166)Enabled in precompiled modeProvides the NVIDIA kernel modules and user-space driver on immutable OS nodes.
MIG Managerpackage-externalDisabledOptional MIG geometry manager. Image mirrored by the user from nvcr.io. See MIG resources.
Upstream NVIDIA Driver (mutable OS)package-externalDisabledOptional upstream driver for mutable OS. Image from nvcr.io, mirrored by the user.

Alauda Build of Node Feature Discovery is a shared prerequisite plugin installed separately and is not included in this package. Alauda GPU Management is a controller-based operator that consumes the node labels NFD publishes; it does not deploy NFD itself. Install NFD on the cluster before creating the ClusterPolicy, otherwise GPU nodes are not detected and no GPU components are reconciled.

The precompiled driver tuple shipped in the package targets one (driver version, kernel, OS) combination. Other kernel tuples are imported separately and are not included in the package. See Configure the driver mode.

Supported architecture

ArchitectureNotes
amd64 (x86_64)The product package is amd64.

ACP integration points

IntegrationValue
ACP versionsv4.0-v4.3
Main custom resourceClusterPolicy
Resource keynvidia.com/gpu. Confirm from node allocatable resources.
Runtime integrationruntimeClassName: nvidia, configured by the container toolkit
MonitoringDCGM-Exporter; ACP owns the GPU dashboards
External prerequisiteAlauda Build of Node Feature Discovery installed on the target workload cluster

Do not treat this page as an upstream hardware matrix. Only combinations delivered or supported by Alauda GPU Management are listed here. For the validated hardware and operating systems, see Compatibility.