Merge pull request #597 from robinmordasiewicz/dev106

adding init

manifests/infrastructure/gpu-operator/HelmRelease.yaml

@@ -0,0 +1,27 @@
+---
+apiVersion: helm.toolkit.fluxcd.io/v2beta1
+kind: HelmRelease
+metadata:
+  name: gpu-operator
+  namespace: cluster-config
+spec:
+  targetNamespace: application
+  releaseName: gpu-operator
+  chart:
+    spec:
+      chart: gpu-operator
+      sourceRef:
+        kind: HelmRepository
+        name: gpu-operator
+        namespace: cluster-config
+  interval: 10m0s
+  install:
+    timeout: 20m
+    remediation:
+      retries: 5
+  upgrade:
+    timeout: 20m
+  values:
+    driver.enabled: false
+    toolkit.enabled: false
+    operator.runtimeClass: "nvidia-container-runtime"

manifests/infrastructure/gpu-operator/HelmRepository.yaml

@@ -0,0 +1,9 @@
+---
+apiVersion: source.toolkit.fluxcd.io/v1beta2
+kind: HelmRepository
+metadata:
+  name: gpu-operator
+  namespace: cluster-config
+spec:
+  interval: 30m
+  url: https://helm.ngc.nvidia.com/nvidia

manifests/infrastructure/gpu-operator/kustomization.yaml

@@ -0,0 +1,6 @@
+---
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - HelmRelease.yaml
+  - HelmRepository.yaml

manifests/infrastructure/kustomization.yaml

@@ -4,3 +4,4 @@ kind: Kustomization
 resources:
   - rbac.yaml
   - ./fortiweb-ingress
+  - ./gpu-operator

terraform/spoke-k8s_cluster.tf

@@ -128,7 +128,6 @@ resource "azurerm_monitor_data_collection_rule" "this" {
   name                = "rule-${azurerm_resource_group.azure_resource_group.name}-${azurerm_resource_group.azure_resource_group.location}"
   resource_group_name = azurerm_resource_group.azure_resource_group.name
   location            = azurerm_resource_group.azure_resource_group.location
-
   destinations {
     log_analytics {
       workspace_resource_id = azurerm_log_analytics_workspace.log_analytics.id
@@ -165,9 +164,9 @@ resource "azurerm_monitor_data_collection_rule_association" "this" {
 }
 
 resource "null_resource" "kube_config" {
-  #triggers = {
-  #  always_run = timestamp()
-  #}
+  triggers = {
+    kube_config_exists = "${fileexists("~/.kube/config")}"
+  }
   depends_on = [azurerm_kubernetes_cluster.kubernetes_cluster]
   provisioner "local-exec" {
     command = "echo \"${azurerm_kubernetes_cluster.kubernetes_cluster.kube_config_raw}\" > ~/.kube/config && chmod 600 ~/.kube/config"