talos/.github/workflows/integration-aws-nvidia-oss-cron.yaml

# THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
#
# Generated on 2024-10-08T12:03:35Z by kres 34e72ac.

name: integration-aws-nvidia-oss-cron
concurrency:
  group: ${{ github.head_ref || github.run_id }}
  cancel-in-progress: true
"on":
  schedule:
    - cron: 30 7 * * *
jobs:
  default:
    runs-on:
      - self-hosted
      - generic
    steps:
      - name: gather-system-info
        id: system-info
        uses: kenchan0130/actions-system-info@v1.3.0
        continue-on-error: true
      - name: print-system-info
        run: |
          MEMORY_GB=$((${{ steps.system-info.outputs.totalmem }}/1024/1024/1024))

          OUTPUTS=(
            "CPU Core: ${{ steps.system-info.outputs.cpu-core }}"
            "CPU Model: ${{ steps.system-info.outputs.cpu-model }}"
            "Hostname: ${{ steps.system-info.outputs.hostname }}"
            "NodeName: ${NODE_NAME}"
            "Kernel release: ${{ steps.system-info.outputs.kernel-release }}"
            "Kernel version: ${{ steps.system-info.outputs.kernel-version }}"
            "Name: ${{ steps.system-info.outputs.name }}"
            "Platform: ${{ steps.system-info.outputs.platform }}"
            "Release: ${{ steps.system-info.outputs.release }}"
            "Total memory: ${MEMORY_GB} GB"
          )

          for OUTPUT in "${OUTPUTS[@]}";do
            echo "${OUTPUT}"
          done          
        continue-on-error: true
      - name: checkout
        uses: actions/checkout@v4
      - name: Unshallow
        run: |
          git fetch --prune --unshallow          
      - name: Set up Docker Buildx
        id: setup-buildx
        uses: docker/setup-buildx-action@v3
        with:
          driver: remote
          endpoint: tcp://buildkit-amd64.ci.svc.cluster.local:1234
        timeout-minutes: 10
      - name: Mask secrets
        run: |
          echo "$(sops -d .secrets.yaml | yq -e '.secrets | to_entries[] | "::add-mask::" + .value')"          
      - name: Set secrets for job
        run: |
          sops -d .secrets.yaml | yq -e '.secrets | to_entries[] | .key + "=" + .value' >> "$GITHUB_ENV"          
      - name: Download artifacts
        if: github.event_name != 'schedule'
        uses: actions/download-artifact@v4
        with:
          name: talos-artifacts
          path: _out
      - name: Fix artifact permissions
        if: github.event_name != 'schedule'
        run: |
          xargs -a _out/executable-artifacts -I {} chmod +x {}          
      - name: ci-temp-release-tag
        if: github.event_name != 'schedule'
        run: |
          make ci-temp-release-tag          
      - name: generate
        if: github.event_name == 'schedule'
        run: |
          make generate          
      - name: uki-certs
        if: github.event_name == 'schedule'
        env:
          PLATFORM: linux/amd64
        run: |
          make uki-certs          
      - name: build
        if: github.event_name == 'schedule'
        env:
          IMAGE_REGISTRY: registry.dev.siderolabs.io
          PLATFORM: linux/amd64
          PUSH: "true"
        run: |
          make talosctl-linux-amd64 kernel sd-boot sd-stub initramfs installer imager talos _out/integration-test-linux-amd64          
      - name: talosctl-cni-bundle
        if: github.event_name == 'schedule'
        run: |
          make talosctl-cni-bundle          
      - name: images-essential
        if: github.event_name == 'schedule'
        env:
          IMAGE_REGISTRY: registry.dev.siderolabs.io
          PLATFORM: linux/amd64
        run: |
          make images-essential          
      - name: checkout extensions
        uses: actions/checkout@v4
        with:
          path: _out/extensions
          ref: release-1.8
          repository: siderolabs/extensions
      - name: set variables
        run: |
          cat _out/talos-metadata >> "$GITHUB_ENV"          
      - name: build extensions
        env:
          PLATFORM: linux/amd64
          PUSH: "true"
          REGISTRY: registry.dev.siderolabs.io
        run: |
          make nvidia-container-toolkit-production nvidia-open-gpu-kernel-modules-production zfs extensions-metadata -C _out/extensions          
      - name: e2e-aws-prepare
        env:
          E2E_AWS_TARGET: nvidia-oss
          EXTENSIONS_METADATA_FILE: _out/extensions/_out/extensions-metadata
          IMAGE_REGISTRY: registry.dev.siderolabs.io
        run: |
          make e2e-aws-prepare          
      - name: checkout contrib
        uses: actions/checkout@v4
        with:
          path: _out/contrib
          ref: main
          repository: siderolabs/contrib
      - name: setup tf
        uses: hashicorp/setup-terraform@v3
        with:
          terraform_wrapper: "false"
      - name: tf apply
        env:
          TF_E2E_ACTION: apply
          TF_E2E_TEST_TYPE: aws
          TF_SCRIPT_DIR: _out/contrib
        run: |
          make e2e-cloud-tf          
      - name: e2e-aws-nvidia-oss
        env:
          EXTRA_TEST_ARGS: -talos.extensions.nvidia
          TEST_NUM_NODES: "4"
        run: |
          make e2e-aws          
      - name: tf destroy
        if: always()
        env:
          TF_E2E_ACTION: destroy
          TF_E2E_TEST_TYPE: aws
          TF_SCRIPT_DIR: _out/contrib
        run: |
          make e2e-cloud-tf