charts/charts/stable/intel-gpu-plugin

intel-gpu-plugin

The Intel GPU plugin facilitates offloading the processing of computation intensive workloads to GPU hardware

This chart is not maintained by the upstream project and any issues with the chart should be raised here

Source Code

• https://github.com/intel/intel-device-plugins-for-kubernetes/blob/master/cmd/gpu_plugin

Requirements

Dependencies

Repository	Name	Version

TL;DR

helm repo add k8s-at-home https://k8s-at-home.com/charts/
helm repo update
helm install intel-gpu-plugin k8s-at-home/intel-gpu-plugin

Installing the Chart

To install the chart with the release name intel-gpu-plugin

helm install intel-gpu-plugin k8s-at-home/intel-gpu-plugin

Uninstalling the Chart

To uninstall the intel-gpu-plugin deployment

helm uninstall intel-gpu-plugin

The command removes all the Kubernetes components associated with the chart including persistent volumes and deletes the release.

Configuration

Read through the values.yaml file. It has several commented out suggested values. Other values may be used from the values.yaml from the common library.

Specify each parameter using the --set key=value[,key=value] argument to helm install.

helm install intel-gpu-plugin \
  --set env.TZ="America/New York" \
    k8s-at-home/intel-gpu-plugin

Alternatively, a YAML file that specifies the values for the above parameters can be provided while installing the chart.

helm install intel-gpu-plugin k8s-at-home/intel-gpu-plugin -f values.yaml

Custom configuration

Node Feature Discovery

If your cluster runs Node Feature Discovery, you can deploy the device plugin only on nodes with Intel GPU by specifying the desired nodeSelector or affinity in your values. For example (make sure to update to your exact feature label):

affinity:
  nodeAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
        - matchExpressions:
            - key: feature.node.kubernetes.io/pci-0300_8086.present
              operator: In
              values:
                - "true"

Values

Important: When deploying an application Helm chart you can add more values from our common library chart here

Key	Type	Default	Description
affinity	object	{}
fullnameOverride	string	""
image.pullPolicy	string	"IfNotPresent"	Image pull policy
image.repository	string	"intel/intel-gpu-plugin"	Image repository
image.tag	string	"0.20.0"	Image tag. Possible values listed here
imagePullSecrets	list	[]
nameOverride	string	""
nodeSelector	object	{}
podAnnotations	object	{}	Key-value pairs to add as pod annotations
podSecurityContext	object	{}
resources	object	{}
securityContext	object	{}
serviceAccount.annotations	object	{}
serviceAccount.create	bool	true
serviceAccount.name	string	""
sharedDeviceNumber	int	1	Number of containers sharing the same GPU device
strategyType	string	"Recreate"	Specifies the strategy used to replace old Pods by new ones
tolerations	list	[]

Changelog

All notable changes to this application Helm chart will be documented in this file but does not include changes from our common library. To read those click here.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

1.2.0

Fixed

• Renamed the shared_device_number value to sharedDeviceNumber.

1.1.0

Added

• This chart now supports the shared-dev-num value. This allows setting the number of containers sharing the same GPU device.

Changed

• Bumped the image version to 0.19.0.

Fixed

• Changelog versions

1.0.1

Added

• N/A

Changed

• use helm-docs

Removed

• N/A

Support

• See the Docs
• Open an issue
• Ask a question
• Join our Discord community

---

Autogenerated from chart metadata using helm-docs v1.5.0