daily-nightly-jobs.yml

name: Daily nightly jobs
on:
  schedule:
    - cron: "0 0 * * *" # every day at midnight

defaults:
  run:
    # reference: https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#using-a-specific-shell
    shell: bash --noprofile --norc -eo pipefail -x {0}

jobs:
  canary-arm64:
    runs-on: [self-hosted, ubuntu-20.04-arm64, ARM64]
    if: github.repository == 'rook/rook'
    env:
      BLOCK: /dev/sdb

    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup golang
        uses: actions/setup-go@v5
        with:
          go-version: "1.21"

      - name: Install Docker
        run: |
          sudo apt-get update
          sudo apt-get install docker.io -y

      - name: Start Docker service
        run: sudo service docker start

      - name: Check Docker version
        run: docker version

      - name: setup minikube
        run: |
          sudo apt-get install build-essential -y
          curl -LO https://storage.googleapis.com/minikube/releases/latest/minikube-linux-arm64
          sudo install minikube-linux-arm64 /usr/local/bin/minikube
          sudo rm -f minikube-linux-arm64
          curl -LO "https://dl.k8s.io/release/$(curl -L -s https://dl.k8s.io/release/stable.txt)/bin/linux/arm64/kubectl"
          sudo install -o root -g root -m 0755 kubectl /usr/local/bin/kubectl
          minikube start --memory 28g --cpus=12 --driver=docker

      - name: install yq
        run: |
          sudo curl -JL https://github.com/mikefarah/yq/releases/download/3.4.1/yq_linux_arm64  --output /usr/bin/yq
          sudo chmod +x /usr/bin/yq

      - name: print k8s cluster status
        run: tests/scripts/github-action-helper.sh print_k8s_cluster_status

      - name: use local disk and create partitions for osds
        run: |
          export BLOCK="/dev/$(tests/scripts/github-action-helper.sh find_extra_block_dev)"
          tests/scripts/github-action-helper.sh use_local_disk
          tests/scripts/create-bluestore-partitions.sh --disk "$BLOCK" --osd-count 1

      - name: validate-yaml
        run: tests/scripts/github-action-helper.sh validate_yaml

      - name: deploy cluster
        run: |
          # Use the official build images for the nightly arm tests instead of rebuilding
          export USE_LOCAL_BUILD=false
          # removing liveness probes since the env is slow and the probe is killing the daemons
          yq write -d0 -i deploy/examples/cluster-test.yaml "spec.healthCheck.livenessProbe.mon.disabled" true
          yq write -d0 -i deploy/examples/cluster-test.yaml "spec.healthCheck.livenessProbe.mgr.disabled" true
          yq write -d0 -i deploy/examples/cluster-test.yaml "spec.healthCheck.livenessProbe.osd.disabled" true
          tests/scripts/github-action-helper.sh deploy_cluster
          # there are no package for arm64 nfs-ganesha
          kubectl delete -f deploy/examples/nfs-test.yaml

      - name: wait for prepare pod
        run: timeout 900 sh -c 'until kubectl -n rook-ceph logs -f $(kubectl -n rook-ceph get pod -l app=rook-ceph-osd-prepare -o jsonpath='{.items[*].metadata.name}'); do sleep 5; done' || kubectl -n rook-ceph get all && kubectl logs -n rook-ceph deploy/rook-ceph-operator

      - name: wait for ceph to be ready
        run: tests/scripts/github-action-helper.sh wait_for_ceph_to_be_ready "mon,mgr,osd,mds,rgw,rbd_mirror,fs_mirror" 1

      - name: teardown minikube, docker and kubectl
        if: always()
        run: |
          uptime
          minikube delete
          docker system prune --all --force
          sudo service docker stop
          sudo rm -rf  /usr/local/bin/minikube
          sudo rm -rf /usr/local/bin/kubectl

      - name: remove /usr/bin/yq
        if: always()
        run: sudo rm -rf /usr/bin/yq

      - name: collect common logs
        if: always()
        uses: ./.github/workflows/collect-logs
        with:
          name: canary-arm64

      - name: upload canary test result
        uses: actions/upload-artifact@v4
        if: always()
        with:
          name: canary-arm64
          path: test

  smoke-suite-quincy-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephSmokeSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          SKIP_CLEANUP_POLICY=false CEPH_SUITE_VERSION="quincy-devel" go test -v -timeout 1800s -run TestCephSmokeSuite github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="smoke-ns"
          export OPERATOR_NAMESPACE="smoke-ns-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-smoke-suite-quincy-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  smoke-suite-reef-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephSmokeSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          SKIP_CLEANUP_POLICY=false CEPH_SUITE_VERSION="reef-devel" go test -v -timeout 1800s -run TestCephSmokeSuite github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="smoke-ns"
          export OPERATOR_NAMESPACE="smoke-ns-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-smoke-suite-reef-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  smoke-suite-ceph-main:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephSmokeSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          SKIP_CLEANUP_POLICY=false CEPH_SUITE_VERSION=main go test -v -timeout 1800s -run TestCephSmokeSuite github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="smoke-ns"
          export OPERATOR_NAMESPACE="smoke-ns-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-smoke-suite-master-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  object-suite-quincy-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephObjectSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          SKIP_CLEANUP_POLICY=false CEPH_SUITE_VERSION="quincy-devel" go test -v -timeout 1800s -failfast -run TestCephObjectSuite github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="object-ns"
          export OPERATOR_NAMESPACE="object-ns-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-object-suite-quincy-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  object-suite-ceph-main:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephObjectSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          SKIP_CLEANUP_POLICY=false CEPH_SUITE_VERSION=main go test -v -timeout 1800s -failfast -run TestCephObjectSuite github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="object-ns"
          export OPERATOR_NAMESPACE="object-ns-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-object-suite-master-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  upgrade-from-reef-stable-to-reef-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephUpgradeSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          go test -v -timeout 1800s -failfast -run TestCephUpgradeSuite/TestUpgradeCephToReefDevel github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="upgrade"
          export OPERATOR_NAMESPACE="upgrade-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-upgrade-suite-reef-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  upgrade-from-quincy-stable-to-quincy-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - name: checkout
        uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: setup cluster resources
        uses: ./.github/workflows/integration-test-config-latest-k8s
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          kubernetes-version: "1.28.4"

      - name: TestCephUpgradeSuite
        run: |
          export DEVICE_FILTER=$(tests/scripts/github-action-helper.sh find_extra_block_dev)
          go test -v -timeout 1800s -failfast -run TestCephUpgradeSuite/TestUpgradeCephToQuincyDevel github.com/rook/rook/tests/integration

      - name: collect common logs
        if: always()
        run: |
          export LOG_DIR="/home/runner/work/rook/rook/tests/integration/_output/tests/"
          export CLUSTER_NAMESPACE="upgrade"
          export OPERATOR_NAMESPACE="upgrade-system"
          tests/scripts/collect-logs.sh

      - name: Artifact
        uses: actions/upload-artifact@v4
        if: failure()
        with:
          name: ceph-upgrade-suite-quincy-artifact
          path: /home/runner/work/rook/rook/tests/integration/_output/tests/

  rgw-multisite-test-with-ceph-devel:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    strategy:
      matrix:
        ceph-image-tag:
          ["latest-main-devel", "latest-quincy-devel", "latest-reef-devel"]

    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: run RGW multisite test
        uses: ./.github/workflows/rgw-multisite-test
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          ceph-image: quay.io/ceph/daemon-base:${{ matrix.ceph-image-tag }}

      - name: upload test result
        uses: actions/upload-artifact@v4
        if: always()
        with:
          name: rgw-multisite-testing-ceph-${{ matrix.ceph-image-tag }}
          path: test

  encryption-pvc-kms-ibm-kp:
    if: github.repository == 'rook/rook'
    runs-on: ubuntu-20.04
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: consider debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}

      - name: run encryption KMS IBM Key Protect
        uses: ./.github/workflows/encryption-pvc-kms-ibm-kp
        if: "env.IBM_KP_SERVICE_INSTANCE_ID != '' && env.IBM_KP_SERVICE_API_KEY != ''"
        env:
          IBM_KP_SERVICE_INSTANCE_ID: ${{ secrets.IBM_INSTANCE_ID }}
          IBM_KP_SERVICE_API_KEY: ${{ secrets.IBM_SERVICE_API_KEY }}
        with:
          ibm-instance-id: ${{ secrets.IBM_INSTANCE_ID }}
          ibm-service-api-key: ${{ secrets.IBM_SERVICE_API_KEY }}
          github-token: ${{ secrets.GITHUB_TOKEN }}

      - name: upload test result
        uses: actions/upload-artifact@v4
        if: always()
        with:
          name: encryption-pvc-kms-ibm-kp
          path: test