Goal

• Deploy the upstream kubernetes-sigs/dra-example-driver on OpenShift as a hardware-independent reference DRA driver for certifying DRA features
• Contribute OpenShift support (SCC, UBI image, deployment docs) upstream

Why is this important?

• Decouples DRA feature certification from third-party drivers and hardware availability
• The dra-example-driver is the upstream sig-node reference implementation — it simulates GPU devices via environment variables, no real hardware needed
• Enables repeatable DRA testing in CI without GPU nodes
• We can still test on third-party drivers (e.g., NVIDIA), but this gives a reliable baseline that is not tied to vendor release cycles

Scenarios

1. Engineer deploys dra-example-driver via Helm on an OCP 4.21+ cluster and verifies kubelet plugin pods are running on all nodes
2. Engineer creates ResourceClaims, DeviceClasses, and validates device allocation/scheduling using mock GPU devices
3. CI job deploys the driver and runs DRA e2e tests on every PR/nightly without requiring GPU hardware

Acceptance Criteria

• CI - MUST be running successfully with tests automated
• Release Technical Enablement - Provide necessary release enablement details and documents
• dra-example-driver deploys and runs on OCP 4.21+ (K8s 1.35) with DRA enabled by default
• Helm chart includes OpenShift SCC for the kubelet plugin DaemonSet
• Container image built on UBI base image
• Core DRA workflows validated: ResourceClaim creation, device allocation, pod scheduling
• Changes contributed upstream to kubernetes-sigs/dra-example-driver

Dependencies (internal and external)

1. DRA enabled by default in OCP 4.21 (OCPNODE-3895 — already merged)
2. kubernetes-sigs/dra-example-driver upstream repo: https://github.com/kubernetes-sigs/dra-example-driver
3. cert-manager (optional, only if webhook validation is needed)

Previous Work

1. OCPNODE-4079: Implement partitionable devices support in dra-example-driver
2. OCPNODE-4043: e2e tests to validate downstream DRA APIs with NVIDIA GPU
3. https://github.com/openshift/api/pull/2498 — DRA feature gate enabled by default

Open questions

1. Should we create a new OCPNODE component for dra-example-driver or reuse an existing one?
2. Webhook: use cert-manager or OpenShift service-serving certificates?

Done Checklist

• CI - CI is running, tests are automated and merged.
• Release Enablement - <link to Feature Enablement Presentation>
• DEV - Upstream code and tests merged: <link to PR on kubernetes-sigs/dra-example-driver>
• DEV - Upstream documentation merged: <link>
• QE - Test plans in Polarion: <link>
• QE - Automated tests merged: <link>
• DOC - Downstream documentation merged: <link>