Merge pull request #3670 from k8s-infra-cherrypick-robot/cherry-pick-3645-to-release-1.8

[release-1.8] Update default GPU VM type to Standard_NV12s_v3

Author: k8s-ci-robot

templates/test/ci/cluster-template-prow-nvidia-gpu.yaml

@@ -1,6 +1,6 @@
 - op: replace
   path: /spec/template/spec/osDisk/managedDisk/storageAccountType
-  value: "${AZURE_GPU_NODE_STORAGE_TYPE:=Standard_LRS}"
+  value: "${AZURE_GPU_NODE_STORAGE_TYPE:=Premium_LRS}"
 - op: replace
   path: /spec/template/spec/vmSize
-  value: "${AZURE_GPU_NODE_MACHINE_TYPE:=Standard_NV6}"
+  value: "${AZURE_GPU_NODE_MACHINE_TYPE:=Standard_NV12s_v3}"

templates/test/ci/prow-nvidia-gpu/patches/node-storage-type.yaml

@@ -469,7 +469,7 @@ var _ = Describe("Workload cluster creation", func() {
 	// ci-e2e.sh and Prow CI skip this test by default, since N-series GPUs are relatively expensive
 	// and may require specific quota limits on the subscription.
 	// To include this test, set `GINKGO_SKIP=""`.
-	// You can override the default SKU `Standard_NV6` and `Standard_LRS` storage by setting
+	// You can override the default SKU `Standard_NV12s_v3` and `Premium_LRS` storage by setting
 	// the `AZURE_GPU_NODE_MACHINE_TYPE` and `AZURE_GPU_NODE_STORAGE_TYPE` environment variables.
 	// See https://azure.microsoft.com/en-us/pricing/details/virtual-machines/linux/ for pricing.
 	Context("Creating a GPU-enabled cluster [OPTIONAL]", func() {