packaging.yaml

name: Packaging

on:
  pull_request:
  workflow_dispatch:
  push:
    branches:
      - "main"
    tags:
      - "v*"

permissions:
  contents: read
  packages: write
  attestations: write
  id-token: write

jobs:
  helm:
    name: Helm Chart
    runs-on: ubuntu-latest
    permissions:
      packages: write
    outputs:
      generated-semver: ${{ steps.semantic-version.outputs.generated-semver }}
    steps:
    - uses: actions/checkout@v6
    - uses: actions/setup-python@v6
    - name: Install yq
      run: |
        pip install yq
    - name: Generate SemVer
      id: semantic-version
      run: |
        CHART_VERSION=$(yq -r '.version' helm/Chart.yaml)
        if [ "${{ github.event_name }}" = "pull_request" ]; then
          LOCAL_SEGMENT=+pr-${{ github.event.pull_request.number }}
        elif [ "${{ github.ref_type }}" = "tag" ]; then
          LOCAL_SEGMENT=""
        else
          SHORT_SHA=$(echo "${{ github.sha }}" | cut -c1-7)
          LOCAL_SEGMENT=+${SHORT_SHA}
        fi
        GENERATED_VERSION=${CHART_VERSION}${LOCAL_SEGMENT}
        yq -Y -i ".version = \"$GENERATED_VERSION\"" helm/Chart.yaml
        echo "generated-semver=$GENERATED_VERSION" >> $GITHUB_OUTPUT
    - name: Chart | Push
      uses: appany/helm-oci-chart-releaser@v0.5.0
      with:
        name: climate-ref-aft
        repository: climate-ref/charts
        tag: ${{ steps.semantic-version.outputs.generated-semver }}
        path: helm
        registry: ghcr.io
        registry_username: ${{ github.actor }}
        registry_password: ${{ secrets.GITHUB_TOKEN }}
        update_dependencies: 'true'

  test-minimal:
    name: Test Minimal Deployment
    runs-on: ubuntu-latest
    timeout-minutes: 10
    needs: [helm]
    steps:
    - uses: actions/checkout@v6
    - name: Start minikube
      uses: medyagh/setup-minikube@latest
    - name: Set up Helm
      uses: azure/setup-helm@v4.3.1
    - name: Install Chart (minimal)
      run: |
        helm install test oci://ghcr.io/climate-ref/charts/climate-ref-aft \
          --version=${{ needs.helm.outputs.generated-semver }} \
          -f helm/ci/minimal-values.yaml
    - name: Wait for pods
      run: |
        echo "Waiting for deployments to be ready..."
        kubectl wait --for=condition=available deployment --all --timeout=120s
        kubectl get pods
    - name: Verify all pods healthy
      run: |
        # Check no pods are in CrashLoopBackOff or Error state
        UNHEALTHY=$(kubectl get pods --no-headers | grep -E 'CrashLoopBackOff|Error|ImagePullBackOff' || true)
        if [ -n "$UNHEALTHY" ]; then
          echo "Unhealthy pods found:"
          echo "$UNHEALTHY"
          echo ""
          echo "=== Pod logs ==="
          for pod in $(echo "$UNHEALTHY" | awk '{print $1}'); do
            echo "--- $pod ---"
            kubectl logs "$pod" --tail=50 || true
          done
          exit 1
        fi
        echo "All pods healthy."
    - name: Capture logs on failure
      if: failure()
      run: |
        kubectl get pods
        kubectl describe pods
        for pod in $(kubectl get pods --no-headers | awk '{print $1}'); do
          echo "=== $pod ==="
          kubectl logs "$pod" --tail=100 || true
        done

  test:
    name: Test Helm Deployment
    runs-on: ubuntu-latest
    timeout-minutes: 30
    needs: [helm]
    steps:
    - uses: actions/checkout@v6
    - name: Cache Sample Data (Restore)
      id: cache-sample-data-restore
      uses: actions/cache/restore@v5
      with:
        path: ${{ github.workspace }}/cache/ref-config
        key: ${{ runner.os }}-sample-data
        enableCrossOsArchive: true
    - name: Set permissions for cached data
      run: |
        sudo install -d --owner=1000 --group=1000 ${GITHUB_WORKSPACE}/cache/ref-config
    - name: Start minikube
      uses: medyagh/setup-minikube@latest
      with:
        mount-path: '${{ github.workspace }}/cache/ref-config:/cache/ref-config'
    - name: Set up Helm
      uses: azure/setup-helm@v4.3.1
    - name: Install Chart
      run: |
        helm install test oci://ghcr.io/climate-ref/charts/climate-ref-aft \
          --version=${{ needs.helm.outputs.generated-semver }} \
          -f helm/ci/gh-actions-values.yaml

        sleep 60
        kubectl get pods
        echo ""
        kubectl describe pod -l app.kubernetes.io/component=pmp
        echo ""
        kubectl logs -l app.kubernetes.io/component=pmp
    - name: Run Migrations
      run: |
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref config list
    - name: Initialize Providers
      run: |
        # Imports ilamb3 which tries to create /home/app/.config/ilamb3 on import, no way to tell it to live somewhere else
        # First, set up all providers without fetching data (handles conda envs)
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref providers setup --skip-data --skip-validate
        # Fetch data for providers except esmvaltool (ERA5 data is too large for CI)
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref providers setup --provider pmp
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref providers setup --provider ilamb
    - name: Fetch Test Data
      run: |
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref datasets fetch-data --registry sample-data --output-directory /ref/sample-data

    - name: Cache Sample Data (Save)
      uses: actions/cache/save@v5
      with:
        path: ${{ github.workspace }}/cache/ref-config
        key: ${{ runner.os }}-sample-data

    - name: Ingest Test Data (CMIP6)
      run: |
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref -v datasets ingest --source-type cmip6 /ref/sample-data/CMIP6
    - name: Ingest Test Data (obs4mips)
      run: |
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref -v datasets ingest --source-type obs4mips /ref/sample-data/obs4REF
    - name: Simple Solve
      run: |
        # Use a fixed set of fast diagnostics to keep CI times predictable
        kubectl exec deployment/test-climate-ref-aft-orchestrator -- ref -v solve --timeout 720 --one-per-provider \
          --diagnostic global-mean-timeseries \
          --diagnostic annual-cycle \
          --diagnostic gpp-wecann \
          --provider esmvaltool \
          --provider pmp \
          --provider ilamb
    - name: Capture Worker Logs on Failure
      if: failure()
      run: |
        echo "=== PMP Worker Logs ==="
        kubectl logs -l app.kubernetes.io/component=pmp --tail=500 || true
        echo ""
        echo "=== ESMValTool Worker Logs ==="
        kubectl logs -l app.kubernetes.io/component=esmvaltool --tail=500 || true
        echo ""
        echo "=== ILAMB Worker Logs ==="
        kubectl logs -l app.kubernetes.io/component=ilamb --tail=500 || true
        echo ""
        echo "=== Orchestrator Worker Logs ==="
        kubectl logs -l app.kubernetes.io/component=orchestrator --tail=500 || true
        echo ""
        echo "=== Flower Logs ==="
        kubectl logs -l app.kubernetes.io/component=flower --tail=500 || true
        echo ""
        echo "=== Dragonfly Logs ==="
        kubectl logs -l app.kubernetes.io/name=dragonfly --tail=500 || true