Add dynamic timewait in integ tests

.github/workflows/helm-integ-test.yml

@@ -75,4 +75,116 @@ jobs:
           retry_wait_seconds: 5
           command: |
             cd integration-tests/terraform/helm
+            terraform destroy --auto-approve
+
+  HelmChartsIntegrationTestWindows-2022:
+    name: HelmChartsIntegrationTestWindows-2022
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      - name: Generate testing id
+        run: echo TESTING_ID="${{ github.run_id }}-${{ github.run_number }}" >> $GITHUB_ENV
+
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v2
+        with:
+          role-to-assume: ${{ env.TERRAFORM_AWS_ASSUME_ROLE }}
+          aws-region: ${{ env.AWS_DEFAULT_REGION }}
+
+      #  local directory to store the kubernetes config
+      - name: Create kubeconfig directory
+        run: mkdir -p ${{ github.workspace }}/../../../.kube
+
+      - name: Set KUBECONFIG environment variable
+        run: echo KUBECONFIG="${{ github.workspace }}/../../../.kube/config" >> $GITHUB_ENV
+
+      - name: Verify Terraform version
+        run: terraform --version
+
+      - name: Terraform apply
+        uses: nick-fields/retry@v2
+        with:
+          max_attempts: 1
+          timeout_minutes: 60 # EKS takes about 20 minutes to spin up a cluster and service on the cluster
+          retry_wait_seconds: 5
+          command: |
+            cd integration-tests/terraform/helm-windows
+            terraform init
+            if terraform apply -auto-approve \
+                -var="windows_os_version=WINDOWS_CORE_2022_x86_64" -var="kube_dir=${{ github.workspace }}/../../../.kube"; then
+              terraform destroy -auto-approve
+            else
+              terraform destroy -auto-approve && exit 1
+            fi
+
+      - name: Terraform destroy
+        if: ${{ cancelled() || failure() }}
+        uses: nick-fields/retry@v2
+        with:
+          max_attempts: 3
+          timeout_minutes: 8
+          retry_wait_seconds: 5
+          command: |
+            cd integration-tests/terraform/helm-windows
+            terraform destroy --auto-approve
+
+  HelmChartsIntegrationTestWindows-2019:
+    name: HelmChartsIntegrationTestWindows-2019
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+    steps:
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      - name: Generate testing id
+        run: echo TESTING_ID="${{ github.run_id }}-${{ github.run_number }}" >> $GITHUB_ENV
+
+      - name: Configure AWS Credentials
+        uses: aws-actions/configure-aws-credentials@v2
+        with:
+          role-to-assume: ${{ env.TERRAFORM_AWS_ASSUME_ROLE }}
+          aws-region: ${{ env.AWS_DEFAULT_REGION }}
+
+      #  local directory to store the kubernetes config
+      - name: Create kubeconfig directory
+        run: mkdir -p ${{ github.workspace }}/../../../.kube
+
+      - name: Set KUBECONFIG environment variable
+        run: echo KUBECONFIG="${{ github.workspace }}/../../../.kube/config" >> $GITHUB_ENV
+
+      - name: Verify Terraform version
+        run: terraform --version
+
+      - name: Terraform apply
+        uses: nick-fields/retry@v2
+        with:
+          max_attempts: 1
+          timeout_minutes: 60 # EKS takes about 20 minutes to spin up a cluster and service on the cluster
+          retry_wait_seconds: 5
+          command: |
+            cd integration-tests/terraform/helm-windows
+            terraform init
+            if terraform apply -auto-approve \
+                -var="windows_os_version=WINDOWS_CORE_2019_x86_64" -var="kube_dir=${{ github.workspace }}/../../../.kube"; then
+              terraform destroy -auto-approve
+            else
+              terraform destroy -auto-approve && exit 1
+            fi
+
+      - name: Terraform destroy
+        if: ${{ cancelled() || failure() }}
+        uses: nick-fields/retry@v2
+        with:
+          max_attempts: 3
+          timeout_minutes: 8
+          retry_wait_seconds: 5
+          command: |
+            cd integration-tests/terraform/helm-windows
             terraform destroy --auto-approve

integration-tests/terraform/helm-windows/main.tf

@@ -128,10 +128,10 @@ resource "aws_eks_node_group" "node_group_windows" {
     min_size     = 1
   }
 
-  ami_type       = "WINDOWS_CORE_2022_x86_64"
+  ami_type       = var.windows_os_version
   capacity_type  = "ON_DEMAND"
   disk_size      = 50
-  instance_types = ["t3a.medium"]
+  instance_types = ["t3.large"]
 
   depends_on = [
     aws_iam_role_policy_attachment.node_CloudWatchAgentServerPolicy,
@@ -206,9 +206,24 @@ resource "helm_release" "this" {
   chart      = "${var.helm_dir}"
 }
 
+resource "null_resource" "deployment_wait" {
+  depends_on = [
+    helm_release.this,
+  ]
+  provisioner "local-exec" {
+    command = <<-EOT
+      curl -LO "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/amd64/kubectl"
+      chmod +x kubectl
+      ./kubectl rollout status daemonset fluent-bit-windows -n amazon-cloudwatch --timeout 600s
+      ./kubectl rollout status daemonset cloudwatch-agent-windows -n amazon-cloudwatch --timeout 600s
+    EOT
+  }
+}
+
 resource "null_resource" "validator" {
   depends_on = [
-    helm_release.this
+    helm_release.this,
+    null_resource.deployment_wait
   ]
   provisioner "local-exec" {
     command = "go test ${var.test_dir} -v --tags=windowslinux"

integration-tests/terraform/helm-windows/variables.tf

@@ -31,3 +31,8 @@ variable "cluster_name" {
   type    = string
   default = "cwagent-operator-helm-integ"
 }
+
+variable "windows_os_version" {
+  type    = string
+  default = "WINDOWS_CORE_2022_x86_64"
+}