install the published helm oci chart in e2e tests

.github/workflows/e2e-tests.yaml

@@ -90,6 +90,13 @@ jobs:
         with:
           name: values-overrides
           path: ${{ github.workspace }}
+      - name: Set up Helm
+        uses: azure/setup-helm@9bc31f4ebc9c6b171d7bfbaa5d006ae7abdb4310 # v5.0.1
+      - name: Verify the published Helm chart is available
+        env:
+          HELM_CHART: ${{ needs.publish-helm-oci-chart.outputs.chart_reference }}
+          HELM_CHART_VERSION: ${{ needs.publish-helm-oci-chart.outputs.chart_version }}
+        run: helm show chart "${HELM_CHART}" --version "${HELM_CHART_VERSION}"
       - name: Set up Holodeck
         uses: NVIDIA/holodeck@e0f3932cf284d92421a55536839fa821a14116aa # v0.3.7
         with:
@@ -124,6 +131,8 @@ jobs:
         env:
           OPERATOR_VERSION: ${{ needs.variables.outputs.operator_version }}
           OPERATOR_IMAGE: ${{ needs.variables.outputs.operator_image }}
+          HELM_CHART: ${{ needs.publish-helm-oci-chart.outputs.chart_reference }}
+          HELM_CHART_VERSION: ${{ needs.publish-helm-oci-chart.outputs.chart_version }}
           GPU_PRODUCT_NAME: "Tesla-T4"
           SKIP_LAUNCH: "true"
           CONTAINER_RUNTIME: "containerd"
@@ -158,6 +167,13 @@ jobs:
         with:
           name: values-overrides
           path: ${{ github.workspace }}
+      - name: Set up Helm
+        uses: azure/setup-helm@9bc31f4ebc9c6b171d7bfbaa5d006ae7abdb4310 # v5.0.1
+      - name: Verify the published Helm chart is available
+        env:
+          HELM_CHART: ${{ needs.publish-helm-oci-chart.outputs.chart_reference }}
+          HELM_CHART_VERSION: ${{ needs.publish-helm-oci-chart.outputs.chart_version }}
+        run: helm show chart "${HELM_CHART}" --version "${HELM_CHART_VERSION}"
       - name: Set up Holodeck
         uses: NVIDIA/holodeck@e0f3932cf284d92421a55536839fa821a14116aa # v0.3.7
         with:
@@ -192,6 +208,8 @@ jobs:
         env:
           OPERATOR_VERSION: ${{ needs.variables.outputs.operator_version }}
           OPERATOR_IMAGE: ${{ needs.variables.outputs.operator_image }}
+          HELM_CHART: ${{ needs.publish-helm-oci-chart.outputs.chart_reference }}
+          HELM_CHART_VERSION: ${{ needs.publish-helm-oci-chart.outputs.chart_version }}
           GPU_PRODUCT_NAME: "Tesla-T4"
           SKIP_LAUNCH: "true"
           CONTAINER_RUNTIME: "containerd"

.github/workflows/publish-helm-oci-chart.yaml

@@ -23,6 +23,13 @@ on:
       operator_version:
         required: true
         type: string
+    outputs:
+      chart_reference:
+        description: "The OCI reference of the published Helm chart"
+        value: ${{ jobs.publish-helm-chart.outputs.chart_reference }}
+      chart_version:
+        description: "The version of the published Helm chart"
+        value: ${{ jobs.publish-helm-chart.outputs.chart_version }}
 
 permissions:
   contents: read
@@ -36,6 +43,9 @@ jobs:
   publish-helm-chart:
     runs-on: ubuntu-latest
     timeout-minutes: 10
+    outputs:
+      chart_reference: ${{ steps.publish.outputs.chart_reference }}
+      chart_version: ${{ steps.package.outputs.chart_version }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
         name: Check out code
@@ -89,6 +99,7 @@ jobs:
           echo "chart_version=${CHART_VERSION}" >> "${GITHUB_OUTPUT}"
 
       - name: Publish Helm OCI chart
+        id: publish
         env:
           GITHUB_REPOSITORY_OWNER: ${{ github.repository_owner }}
           CHART_PACKAGE: ${{ steps.package.outputs.chart_package }}
@@ -100,4 +111,6 @@ jobs:
           CHART_REPOSITORY="oci://ghcr.io/${LOWERCASE_REPO_OWNER}/gpu-operator/charts"
 
           helm push "${CHART_PACKAGE}" "${CHART_REPOSITORY}"
+
+          echo "chart_reference=${CHART_REPOSITORY}/gpu-operator" >> "${GITHUB_OUTPUT}"
           echo "::notice::Published Helm chart to ${CHART_REPOSITORY}/gpu-operator:${CHART_VERSION}"

tests/local.sh

@@ -42,4 +42,6 @@ remote \
     GPU_MODE="${GPU_MODE}" \
     NGC_API_KEY="${NGC_API_KEY}" \
     VALUES_FILE="${VALUES_FILE:-}" \
+    HELM_CHART="${HELM_CHART:-}" \
+    HELM_CHART_VERSION="${HELM_CHART_VERSION:-}" \
         ${TEST_CASE}

tests/scripts/.definitions.sh

@@ -26,3 +26,9 @@ TERRAFORM="terraform -chdir=${TERRAFORM_DIR}"
 : ${PRIVATE_REGISTRY:="nvcr.io"}
 
 : ${GPU_MODE:="gpu"}
+
+CHART_REFERENCE="${HELM_CHART:-${PROJECT_DIR}/deployments/gpu-operator}"
+HELM_CMD_ARGS=("${CHART_REFERENCE}")
+if [[ -n "${HELM_CHART_VERSION:-}" ]]; then
+    HELM_CMD_ARGS+=(--version "${HELM_CHART_VERSION}")
+fi

tests/scripts/end-to-end-nvidia-driver.sh

@@ -7,14 +7,14 @@ test_nvidiadriver_helm_render_options() {
     local render_file
     render_file=$(mktemp)
 
-    ${HELM} template gpu-operator "${PROJECT_DIR}/deployments/gpu-operator" \
+    ${HELM} template gpu-operator "${HELM_CMD_ARGS[@]}" \
         -n "${TEST_NAMESPACE}" \
         --set driver.nvidiaDriverCRD.enabled=true \
         --set driver.nvidiaDriverCRD.deployDefaultCR=true > "${render_file}"
     grep -q "kind: NVIDIADriver" "${render_file}"
     grep -q "default: true" "${render_file}"
 
-    ${HELM} template gpu-operator "${PROJECT_DIR}/deployments/gpu-operator" \
+    ${HELM} template gpu-operator "${HELM_CMD_ARGS[@]}" \
         -n "${TEST_NAMESPACE}" \
         --set driver.nvidiaDriverCRD.enabled=true \
         --set driver.nvidiaDriverCRD.deployDefaultCR=false > "${render_file}"
@@ -23,7 +23,7 @@ test_nvidiadriver_helm_render_options() {
         exit 1
     fi
 
-    ${HELM} template gpu-operator "${PROJECT_DIR}/deployments/gpu-operator" \
+    ${HELM} template gpu-operator "${HELM_CMD_ARGS[@]}" \
         -n "${TEST_NAMESPACE}" \
         --set driver.nvidiaDriverCRD.enabled=false \
         --set driver.nvidiaDriverCRD.deployDefaultCR=true > "${render_file}"

tests/scripts/install-operator.sh

@@ -105,15 +105,15 @@ echo "Operator image: ${OPERATOR_IMAGE}:${OPERATOR_VERSION}"
 
 if [[ "${USE_VALUES_FILE}" == "true" ]]; then
 	echo "Using values file approach: ${VALUES_FILE}"
-	${HELM} install ${PROJECT_DIR}/deployments/gpu-operator --generate-name \
+	${HELM} install "${HELM_CMD_ARGS[@]}" --generate-name \
 		-n "${TEST_NAMESPACE}" \
 		${EXTRA_VALUES_FILES} \
 		-f "${VALUES_FILE}" \
 		${OPERATOR_OPTIONS} \
 		--wait
 else
 	echo "Using --set flags approach"
-	${HELM} install ${PROJECT_DIR}/deployments/gpu-operator --generate-name \
+	${HELM} install "${HELM_CMD_ARGS[@]}" --generate-name \
 		-n "${TEST_NAMESPACE}" \
 		${OPERATOR_OPTIONS} \
 		${TOOLKIT_CONTAINER_OPTIONS} \

tests/scripts/migrate-clusterpolicy-to-nvidiadriver.sh

@@ -174,7 +174,7 @@ if [[ -z "${operator_name}" ]]; then
 fi
 
 echo "Migrating Helm release/${operator_name} from ClusterPolicy driver management to NVIDIADriver"
-${HELM} upgrade "${operator_name}" "${PROJECT_DIR}/deployments/gpu-operator" \
+${HELM} upgrade "${operator_name}" "${HELM_CMD_ARGS[@]}" \
     -n "${TEST_NAMESPACE}" \
     --reuse-values \
     ${OPERATOR_OPTIONS:-} \