EXPERIMENT / WORK-IN-PROGRESS

This repo contains a helm chart to create a mirrored repository to support the mirroring & caching of RPMs to build the driver for the gpu-operator. Periodically, the repos at cdn.redhat.com are not available at time of building the gpu driver. This internally hosted repository provides high availability.

Architecture Diagram

Prerequisites

• entitlement certificate and key in .pem format
• you have connectivity to cdn.redhat.com from within the cluster
• the cluster can be either cluster-entitled or pod entitlements
• the helm chart is deployed into gpu-operator-resources
• the gpu-operator is configured to use the service (mirror-repo) as the RPM repo [1]

[1] https://github.com/NVIDIA/gpu-operator/blob/85f0d1972a2402d549db99d536e85a7fd57bd4b1/deployments/gpu-operator/values.yaml#L64

In the link above, configure "configMapName" and "destintationDir". configMapName as repo-config and destinationDir as /etc/yum.repos.d.

Download entitlement

TBD

Enable entitled builds on OpenShift with a pod in a namespace

https://www.openshift.com/blog/how-to-use-entitled-image-builds-to-build-drivercontainers-with-ubi-on-openshift

Create a machineset that will disable the automounting of subscriptions. By default, CRI-O will automount certificates already on the host. UBI image contains all of the files needed for pod-entitled builds.

Apply the following to your cluster. It will be applied to all worker nodes. If your requirement is on a different machineconfiguration, change the "worker" value.

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker
  name: 50-disable-secret-automount
spec:
  config:
    ignition:
      version: 2.2.0
    storage:
      files:
      - contents:
          source: data:text/plain;charset=utf-8;base64,Cg==
        filesystem: root
        mode: 0644
        path: /etc/containers/mounts.conf

To validate that the change was successful, you can oc debug node/<worker> and cat /host/etc/containers/mounts.conf. After a successful rollout, the file should exist and be empty. In a failed MachineConfig rollout, the file would not exist.