From 414b217ad06367376dcb37e23cbb15147f4b0979 Mon Sep 17 00:00:00 2001
From: Robert Love <terp4life2001@gmail.com>
Date: Tue, 31 Dec 2024 06:52:37 -0500
Subject: [PATCH] nfd-nvidia

---
 .../rules/kustomization.yaml                       |  1 +
 .../node-feature-discovery/rules/nvidia-gpu.yaml   | 14 ++++++++++++++
 .../nvidia-device-plugin/app/helmrelease.yaml      |  3 +--
 .../nvidia-device-plugin/app/kustomization.yaml    |  1 -
 4 files changed, 16 insertions(+), 3 deletions(-)
 create mode 100644 kubernetes/main/apps/kube-system/node-feature-discovery/rules/nvidia-gpu.yaml

diff --git a/kubernetes/main/apps/kube-system/node-feature-discovery/rules/kustomization.yaml b/kubernetes/main/apps/kube-system/node-feature-discovery/rules/kustomization.yaml
index 433f7a0633..743a8ba381 100644
--- a/kubernetes/main/apps/kube-system/node-feature-discovery/rules/kustomization.yaml
+++ b/kubernetes/main/apps/kube-system/node-feature-discovery/rules/kustomization.yaml
@@ -10,3 +10,4 @@ resources:
   - ./intel-gpu.yaml
   - ./vlan-device.yaml
   - ./barcode-device.yaml
+  - ./nvidia-gpu.yaml
diff --git a/kubernetes/main/apps/kube-system/node-feature-discovery/rules/nvidia-gpu.yaml b/kubernetes/main/apps/kube-system/node-feature-discovery/rules/nvidia-gpu.yaml
new file mode 100644
index 0000000000..3b0855b2dd
--- /dev/null
+++ b/kubernetes/main/apps/kube-system/node-feature-discovery/rules/nvidia-gpu.yaml
@@ -0,0 +1,14 @@
+---
+apiVersion: nfd.k8s-sigs.io/v1alpha1
+kind: NodeFeatureRule
+metadata:
+  name: nvidia-gpu-plugin
+spec:
+  rules:
+    - name: nvidia.gpu
+      labels:
+        nvidia.feature.node.kubernetes.io/gpu: "true"
+      matchFeatures:
+        - feature: pci.device
+          matchExpressions:
+            vendor: { op: In, value: ["10de"] }
diff --git a/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/helmrelease.yaml b/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/helmrelease.yaml
index 5d9709f929..345dcf926e 100644
--- a/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/helmrelease.yaml
+++ b/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/helmrelease.yaml
@@ -24,7 +24,6 @@ spec:
       retries: 3
 
   upgrade:
-    cleanupOnFail: true
     crds: CreateReplace
     remediation:
       retries: 3
@@ -32,7 +31,7 @@ spec:
   values:
     runtimeClassName: nvidia
     nodeSelector:
-      nvidia.com/gpu.present: "true"
+      nvidia.feature.node.kubernetes.io/gpu: "true"
     config:
       map:
         default: |-
diff --git a/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/kustomization.yaml b/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/kustomization.yaml
index ee83b23ef3..fad3625bfa 100644
--- a/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/kustomization.yaml
+++ b/kubernetes/main/apps/kube-system/nvidia-device-plugin/app/kustomization.yaml
@@ -1,5 +1,4 @@
 ---
----
 # yaml-language-server: $schema=https://json.schemastore.org/kustomization
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization