Skip to content

Install Dependencies

An OpenShift installation of Run:AI has third-party dependencies that must be pre-installed before installing Run:AI itself. The following document provides instructions for installing and configuring these dependencies.


You must have Cluster Administrator rights to install these dependencies.

Install Operators

NVIDIA provided detailed documentation on how to install the required Operators so as to suppport GPUs on OpenShift.

Disable the NVIDIA Device Plugin and DCGM Exporter

Verify that the GPU Operator is installed by running:

oc get pods -n gpu-operator-resources

After successful verification,

(1) Disabled the GPU Operator by running:

kubectl scale --replicas=0 -n openshift-operators deployment gpu-operator

(1) Disable the NVIDIA DCGM exporter by running:

oc -n gpu-operator-resources patch daemonset nvidia-dcgm-exporter \
   -p '{"spec": {"template": {"spec": {"nodeSelector": {"non-existing": "true"}}}}}'

(2) Replace the NVIDIA Device Plug-in with the Run:AI version:

oc patch daemonsets.apps -n gpu-operator-resources nvidia-device-plugin-daemonset \
   -p '{"spec":{"template":{"spec":{"containers":[{"name":"nvidia-device-plugin-ctr","image":""}]}}}}'
oc create clusterrolebinding --clusterrole=admin \
  --serviceaccount=gpu-operator-resources:nvidia-device-plugin nvidia-device-plugin-crb

Next Steps

Continue with installing the Run:AI Backend.

Last update: January 20, 2022