[pull] master from 2i2c-org:master

config/clusters/leap/support.values.yaml

@@ -1,8 +1,3 @@
-nvidiaDevicePlugin:
-  gke:
-    enabled: true
-    version: "latest"
-
 prometheusIngressAuthSecret:
   enabled: true
 

config/clusters/m2lines/support.values.yaml

@@ -1,8 +1,3 @@
-nvidiaDevicePlugin:
-  gke:
-    enabled: true
-    version: "latest"
-
 grafana:
   grafana.ini:
     server:

helm-charts/support/values.schema.yaml

@@ -91,7 +91,6 @@ properties:
     additionalProperties: false
     required:
       - azure
-      - gke
     properties:
       azure:
         type: object
@@ -101,26 +100,6 @@ properties:
         properties:
           enabled:
             type: boolean
-      gke:
-        type: object
-        additionalProperties: false
-        required:
-          - enabled
-          - version
-        properties:
-          enabled:
-            type: boolean
-          version:
-            type: string
-            enum:
-              - stable
-              - latest
-            description: |
-              Install the stable or latest version of nvidia GPU drivers for the node.
-
-              See table in https://cloud.google.com/kubernetes-engine/docs/how-to/gpus#installing_drivers
-              to determine what versions would be installed. Might need to be matched with appropriate
-              version of the CUDA libraries used in the images users use.
 
   prometheusIngressAuthSecret:
     type: object

helm-charts/support/values.yaml

@@ -434,15 +434,11 @@ redirects:
   rules: []
 
 # Enable a daemonset to install nvidia device plugin to GPU nodes
-# AWS does not require this to be set, as eksctl sets this up automatically
+# Not necessary on GCP & AWS don't need this, as it is handled automatically by terraform or eksctl
+# respectively
 nvidiaDevicePlugin:
-  # For Azure-specific image, default to false
   azure:
     enabled: false
-  # For GKE specific image, defaults to false
-  gke:
-    enabled: false
-    version: "stable"
 
 # Setup a separate storageClass specifically for prometheus data
 prometheusStorageClass:

terraform/gcp/cluster.tf

@@ -296,6 +296,10 @@ resource "google_container_node_pool" "notebook" {
       content {
         type  = each.value.gpu.type
         count = each.value.gpu.count
+
+        gpu_driver_installation_config {
+          gpu_driver_version = "DEFAULT"
+        }
       }
 
     }