Add NFD rule for Gaudi resource driver (#69)

* add nfd rule Signed-off-by: Oksana Baranova <[email protected]>
intel · Jan 4, 2025 · 48e51ae · 48e51ae
1 parent 5326261
commit 48e51ae
Show file tree

Hide file tree

Showing 7 changed files with 54 additions and 9 deletions.
diff --git a/charts/intel-gaudi-resource-driver/Chart.yaml b/charts/intel-gaudi-resource-driver/Chart.yaml
@@ -3,5 +3,13 @@ name: intel-gaudi-resource-driver
 description: A Helm chart for a Dynamic Resource Allocation (DRA) Intel Gaudi Resource Driver
 
 type: application
-version: 0.2.0
-appVersion: "v0.2.0"
+version: 0.3.0
+appVersion: "v0.3.0"
+home: https://github.com/intel/helm-charts
+
+dependencies:
+  - name: node-feature-discovery
+    alias: nfd
+    version: "0.16.6"
+    condition: nfd.enabled
+    repository: https://kubernetes-sigs.github.io/node-feature-discovery/charts
diff --git a/charts/intel-gaudi-resource-driver/README.md b/charts/intel-gaudi-resource-driver/README.md
@@ -16,7 +16,9 @@ helm repo update
 You can execute `helm search repo intel` command to see pulled charts [optional].
 
 ## Install Helm Chart
+When installing, update the dependencies:
 ```
+helm dependency update
 helm install intel-gaudi-resource-driver intel/intel-gaudi-resource-driver
 ```
 ## Upgrade Chart
@@ -43,7 +45,7 @@ You may also run `helm show values` on this chart's dependencies for additional
 | image.repository | string | `intel` |
 | image.name | string | `"intel-gaudi-resource-driver"` |
 | image.pullPolicy | string | `"IfNotPresent"` |
-| image.tag | string | `"v0.2.0"` |
+| image.tag | string | `"v0.3.0"` |
 
 > [!Note]
-> When upgrading, CRDs from previous version need to be removed manually because Helm supports neither upgrading nor deleting CRDs, see: https://github.com/helm/community/blob/main/hips/hip-0011.md
+> If you change the image tag to be used in Helm chart deployment, ensure that the version of the container image is consistent with deployment YAMLs - they might change between releases.
diff --git a/charts/intel-gaudi-resource-driver/templates/device-class.yaml b/charts/intel-gaudi-resource-driver/templates/device-class.yaml
@@ -1,4 +1,4 @@
-apiVersion: resource.k8s.io/v1alpha3
+apiVersion: resource.k8s.io/v1beta1
 kind: DeviceClass
 metadata:
   name: gaudi.intel.com

diff --git a/charts/intel-gaudi-resource-driver/templates/nfd.yaml b/charts/intel-gaudi-resource-driver/templates/nfd.yaml
@@ -0,0 +1,16 @@
+{{- if .Values.nfd.enabled }}
+apiVersion: nfd.k8s-sigs.io/v1alpha1
+kind: NodeFeatureRule
+metadata:
+  name: intel-gaudi-device-rule
+spec:
+  rules:
+    - name: "intel.gaudi"
+      labels:
+        "intel.feature.node.kubernetes.io/gaudi": "true"
+      matchFeatures:
+        - feature: pci.device
+          matchExpressions:
+            vendor: {op: In, value: ["1da3"]}
+            device: {op: In, value: ["1020", "1030"]}
+{{- end }}
diff --git a/charts/intel-gaudi-resource-driver/templates/resource-driver.yaml b/charts/intel-gaudi-resource-driver/templates/resource-driver.yaml
@@ -73,10 +73,15 @@ spec:
       tolerations:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- if .Values.nfd.enabled }}
+      nodeSelector:
+        intel.feature.node.kubernetes.io/gaudi: "true"
+      {{- else }}
       {{- with .Values.kubeletPlugin.nodeSelector }}
       nodeSelector:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- end }}
       {{- with .Values.kubeletPlugin.affinity }}
       affinity:
         {{- toYaml . | nindent 8 }}

diff --git a/charts/intel-gaudi-resource-driver/templates/validating-admission-policy.yaml b/charts/intel-gaudi-resource-driver/templates/validating-admission-policy.yaml
@@ -7,7 +7,7 @@ spec:
   matchConstraints:
     resourceRules:
     - apiGroups:   ["resource.k8s.io"]
-      apiVersions: ["v1alpha3"]
+      apiVersions: ["v1beta1"]
       operations:  ["CREATE", "UPDATE", "DELETE"]
       resources:   ["resourceslices"]
   matchConditions:

diff --git a/charts/intel-gaudi-resource-driver/values.yaml b/charts/intel-gaudi-resource-driver/values.yaml
@@ -9,7 +9,7 @@ image:
   repository: intel
   name: intel-gaudi-resource-driver
   pullPolicy: IfNotPresent
-  tag: "v0.2.0"
+  tag: "v0.3.0"
 
 serviceAccount:
   create: true
@@ -19,13 +19,27 @@ serviceAccount:
 
 kubeletPlugin:
   podAnnotations: {}
+  nodeSelector: {}
+  # label used when nfd.enabled is true
+    #intel.feature.node.kubernetes.io/gaudi: "true"
   tolerations:
     - key: node-role.kubernetes.io/master
       operator: Exists
       effect: NoSchedule
     - key: node-role.kubernetes.io/control-plane
       operator: Exists
       effect: NoSchedule
-  nodeSelector: {}
-  #node-role.kubernetes.io/control-plane: ""
+      # Refer to the official documentation for Node Feature Discovery (NFD) 
+      # regarding node tainting: 
+      # https://nfd.sigs.k8s.io/usage/customization-guide#node-tainting
+    - key: "intel.feature.node.kubernetes.io/gaudi"
+      operator: "Exists"
+      effect: "NoSchedule"  
   affinity: {}
+
+nfd:
+  enabled: false # change to true to install NFD to the cluster
+  nameOverride: intel-gaudi-nfd
+  # TODO: this deprecated NFD option will be replaced in NFD v0.17 with "featureGates.NodeFeatureAPI" (added in v0.16):
+  # https://kubernetes-sigs.github.io/node-feature-discovery/v0.16/deployment/helm.html#general-parameters
+  enableNodeFeatureApi: true