Oracle Cloud Infrastructure Documentation

NVIDIA GPU Operator

When you enable the NVIDIA GPU Operator cluster add-on, you can pass the following key/value pairs as arguments.

Configuration Arguments Common to all Cluster Add-ons
Key (API and CLI) Key's Display Name (Console) Description Required/Optional Default Value Example Value
affinity affinity

A group of affinity scheduling rules.

JSON format in plain text or Base64 encoded.

Not used by:
  • Nvidia GPU Operator
Possible equivalents:
  • Node Feature Discovery, use master.affinity
  • NVIDIA Network Operator, use operator.affinity
  • CSI Driver SMB, use contoller.affinity
 Optional null null
nodeSelectors node selectors

You can use node selectors and node labels to control the worker nodes on which add-on pods run.

For a pod to run on a node, the pod's node selector must have the same key/value as the node's label.

Set nodeSelectors to a key/value pair that matches both the pod's node selector, and the worker node's label.

JSON format in plain text or Base64 encoded.

Not used by:
  • NVIDIA GPU Operator
  • CSI Driver SMB
Possible equivalents:
  • Node Feature Discovery, use worker.nodeSelector
  • NVIDIA Network Operator, use operator.nodeSelectors
 Optional null {"foo":"bar", "foo2": "bar2"}

The pod will only run on nodes that have the foo=bar or foo2=bar2 label.
numOfReplicas numOfReplicas The number of replicas of the add-on deployment.
Not used by:
  • AMD GPU Plugin
  • NVIDIA GPU Operator
  • NVIDIA Network Operator
  • CSI Driver SMB
Possible equivalents:
  • CoreDNS, use nodesPerReplica
  • Node Feature Discovery, use master.replicaCount
 Required 1

Creates one replica of the add-on deployment per cluster.

 2

Creates two replicas of the add-on deployment per cluster.
rollingUpdate rollingUpdate

Controls the desired behavior of rolling update by maxSurge and maxUnavailable.

JSON format in plain text or Base64 encoded.

Not used by:
  • Node Feature Discovery
  • NVIDIA Network Operator
  • CSI Driver SMB
Possible equivalents:
  • NVIDIA GPU Operator, use daemonsets.rollingUpdate.maxUnavailable
 Optional null null
tolerations tolerations

You can use taints and tolerations to control the worker nodes on which add-on pods run.

For a pod to run on a node that has a taint, the pod must have a corresponding toleration.

Set tolerations to a key/value pair that matches both the pod's toleration, and the worker node's taint.

JSON format in plain text or Base64 encoded.

Possible equivalents:
  • Node Feature Discovery, use master.tolerations and/or worker.tolerations
  • NVIDIA GPU Operator, use daemonsets.tolerations
  • NVIDIA Network Operator, use operator.tolerations
  • CSI Driver SMB, use controller.tolerations
 Optional null [{"key":"tolerationKeyFoo", "value":"tolerationValBar", "effect":"noSchedule", "operator":"exists"}]

Only pods that have this toleration can run on worker nodes that have the tolerationKeyFoo=tolerationValBar:noSchedule taint.
topologySpreadConstraints topologySpreadConstraints

How to spread matching pods among the given topology.

JSON format in plain text or Base64 encoded.

Not used by:
  • Node Feature Discovery
  • NVIDIA GPU Operator
  • NVIDIA Network Operator
  • CSI Driver SMB
 Optional null null
Configuration Arguments Specific to this Cluster Add-on
Key (API and CLI) Key's Display Name (Console) Description Required/Optional Default Value Example Value
skipNodeFeatureDiscoveryDependencyCheck skipNodeFeatureDiscoveryDependencyCheck Skip Node Feature Discovery dependency check. Optional false
disableNvidiaGpuPlugin disableNvidiaGpuPlugin Disable the existing NvidiaDevicePlugin add-on and related resources. Optional false
cdi.enabled Enable Container Device Interface Flag to enable Container Device Interface. Optional

false for version 25.3.x.

true for version 25.10.x and later.
cdi.default Enable CDI as default Flag to enable the container runtime to use CDI as the default option to perform device injection. Optional

false for version 25.3.x.

Deprecated in version 25.10.x and later.
daemonsets.labels DaemonSets Labels Labels for all GPU Operator DaemonSets. JSON text or Base64-encoded JSON map. Optional null
daemonsets.annotations DaemonSets Annotations Annotations for all GPU Operator DaemonSets. JSON text or Base64-encoded JSON map. Optional null
daemonsets.tolerations DaemonSets Tolerations Tolerations for all GPU Operator DaemonSets. JSON text or Base64-encoded JSON array. Optional null
daemonsets.rollingUpdate.maxUnavailable DaemonSets MaxUnavailable Maximum number of unavailable pods for rolling update on GPU Operator DaemonSets. Optional 1
dcgm.enabled Enable DCGM Enable NVIDIA DCGM Hostengine as a separate pod. Optional false
dcgm.args DCGM Args Arguments for DCGM Hostengine. JSON plain text or Base64-encoded JSON array. Optional null
dcgm.env DCGM Env Environment variables for DCGM Hostengine. JSON plain text or Base64-encoded JSON. Optional null
dcgm.containerResources DCGM Resources Resource requests and limits for the DCGM Hostengine pod. JSON plain text or Base64-encoded JSON. Optional null
dcgmExporter.enabled Enable DCGM Exporter Enable the DCGM Exporter DaemonSet for metrics. Optional true
dcgmExporter.env DCGM Exporter Env Environment variables for DCGM Exporter. JSON plain text or Base64-encoded JSON. Optional null
dcgmExporter.resources DCGM Exporter Resources Resource requests and limits for the DCGM Exporter pod. JSON plain text or Base64-encoded JSON. Optional null
dcgmExporter.service.internalTrafficPolicy Exporter Service Traffic Policy InternalTrafficPolicy for the DCGM Exporter ClusterIP service. Optional Cluster
dcgmExporter.serviceMonitor.enabled Enable ServiceMonitor Enable a ServiceMonitor for NVIDIA DCGM Exporter. Optional false
dcgmExporter.serviceMonitor.interval ServiceMonitor Interval Scrape interval for NVIDIA DCGM Exporter. Supports y, w, d, h, m, s, and ms. Optional 15s
dcgmExporter.serviceMonitor.honorLabels ServiceMonitor Honor Labels Prefer target metric labels on collisions. Optional false
dcgmExporter.serviceMonitor.additionalLabels ServiceMonitor Additional Labels Additional labels for the ServiceMonitor. JSON plain text or Base64-encoded JSON map. Optional null
dcgmExporter.serviceMonitor.relabelings ServiceMonitor Relabelings Relabeling rules for the ServiceMonitor. JSON plain text or Base64-encoded JSON. Optional null
dcgmExporter.config.name Exporter ConfigMap Name Name of the ConfigMap that provides custom metrics configuration for DCGM Exporter. Optional null
migManager.enabled Enable MIG Manager Enable deployment of NVIDIA MIG Manager. Optional false
migManager.resources MIG Manager Resources Resource requests and limits for each MIG Manager pod. JSON plain text or Base64-encoded JSON. Optional null
migManager.env MIG Manager Env Environment variables for MIG Manager. JSON plain text or Base64-encoded JSON. Optional null
migManager.config.name MIG ConfigMap Name ConfigMap name with mig-parted configuration for MIG Manager. Optional null
migManager.gpuClientsConfig.name MIG GPU Clients Config ConfigMap name with gpu-clients configuration for MIG Manager. Optional null
devicePlugin.enabled Enable Device Plugin Enable NVIDIA Kubernetes Device Plugin. Optional true
devicePlugin.args Device Plugin Args Arguments for the Device Plugin. JSON plain text or Base64-encoded JSON array. Optional null
devicePlugin.env Device Plugin Env Environment variables for the Device Plugin. JSON plain text or Base64-encoded JSON. Optional null
devicePlugin.resources Device Plugin Resources Resource requests and limits for each Device Plugin pod. JSON plain text or Base64-encoded JSON. Optional null
devicePlugin.config.name Device Plugin ConfigMap Name ConfigMap name that provides Device Plugin configuration, shared with GFD when applicable. Optional null
devicePlugin.mps.root MPS Root MPS root path on the host. Optional /run/nvidia/mps
toolkit.enabled Enable Toolkit Enable deployment of NVIDIA Container Toolkit. Optional true
toolkit.env Toolkit Env Environment variables for the Container Toolkit. JSON plain text or Base64-encoded JSON. Optional null
toolkit.resources Toolkit Resources Resource requests and limits for each Container Toolkit pod. JSON plain text or Base64-encoded JSON. Optional null
toolkit.installDir Toolkit Install Dir Install directory of Container Toolkit utilities on the host. Optional /usr/local/nvidia
mig.strategy MIG Strategy MIG strategy for GFD and Device Plugin. Valid values are none, single, and mixed. Optional single
operator.logging.level Operator Log Level Zap logging level: debug, info, error, or an integer > 0 for custom verbosity. Optional info
operator.resources Operator Resources Resource requests and limits for the GPU Operator pod. JSON plain text or Base64-encoded JSON. Optional null
hostPaths.rootFS Host RootFS Host root filesystem path. The path must be chroot-able. Optional /
hostPaths.driverInstallDir Driver Install Dir Root directory for NVIDIA driver files on the host. Optional /run/nvidia/driver
driver.rdma.enabled Enable RDMA in Driver Enable RDMA support for the NVIDIA driver. Optional false
driver.rdma.useHostMofed Use Host MOFED Use host-provided Mellanox OFED instead of container-provided Mellanox OFED. Optional false