verify-operator-boot.yml

name: "Verify Operator Boot"

# Verifies that the operator binary builds and connects to a live kind cluster.
# Acceptance criteria:
#   1. cargo build --release succeeds with zero errors
#   2. ./target/release/stellar-operator run starts against a kind cluster
#   3. Log line "Connected to Kubernetes cluster" appears
#
# Uses shared composite actions for Rust setup and log collection.

on:
  push:
    branches: [main]
  workflow_dispatch:

concurrency:
  group: verify-boot-${{ github.ref }}
  cancel-in-progress: true

env:
  CARGO_TERM_COLOR: always
  RUST_BACKTRACE: 1
  CLUSTER_NAME: stellar-verify
  OPERATOR_NAMESPACE: stellar-system

jobs:
  verify-operator-boot:
    name: Verify Operator Boot & Cluster Connection
    runs-on: ubuntu-latest
    timeout-minutes: 30

    steps:
      - uses: actions/checkout@v7

      # ── Scope check: skip on PRs that don't touch operator code ──────────
      - name: Decide whether boot verification should run
        id: scope
        run: |
          set -euo pipefail
          if [[ "${{ github.event_name }}" != "pull_request" ]]; then
            echo "run=true" >> "$GITHUB_OUTPUT"
            exit 0
          fi
          BASE_SHA="${{ github.event.pull_request.base.sha }}"
          HEAD_SHA="${{ github.event.pull_request.head.sha }}"
          CHANGED=$(git diff --name-only "$BASE_SHA" "$HEAD_SHA" || true)
          echo "$CHANGED"
          if echo "$CHANGED" | grep -Eq '^(src/|config/crd/|Cargo\.toml|Cargo\.lock|build\.rs|Makefile)'; then
            echo "run=true" >> "$GITHUB_OUTPUT"
          else
            echo "run=false" >> "$GITHUB_OUTPUT"
          fi

      # ── Build (AC #1) — shared composite action (issue #1136) ──────────────
      - name: Build operator binary
        if: steps.scope.outputs.run == 'true'
        uses: ./.github/actions/build-operator
        with:
          cache-key: "verify-boot"
          binary-only: "true"

      - name: Confirm binary exists
        if: steps.scope.outputs.run == 'true'
        run: |
          ls -lh target/release/stellar-operator
          echo "✅ cargo build --release succeeded"

      # ── Create kind cluster (AC #2) ───────────────────────────────────────
      - name: Install kind
        if: steps.scope.outputs.run == 'true'
        uses: helm/kind-action@v1.14.0
        with:
          install_only: true

      - name: Create kind cluster
        if: steps.scope.outputs.run == 'true'
        run: |
          kind create cluster --name ${{ env.CLUSTER_NAME }} --wait 60s
          kubectl cluster-info --context kind-${{ env.CLUSTER_NAME }}
          echo "✅ kind cluster '${{ env.CLUSTER_NAME }}' is ready"

      # ── Install CRD ───────────────────────────────────────────────────────
      - name: Install StellarNode CRD
        if: steps.scope.outputs.run == 'true'
        run: |
          kubectl apply -f config/crd/stellarnode-crd.yaml
          kubectl wait --for=condition=established --timeout=30s \
            crd/stellarnodes.stellar.org || true

      # ── Run operator and verify log line (AC #2 + AC #3) ─────────────────
      - name: Run operator and verify cluster connection
        if: steps.scope.outputs.run == 'true'
        run: |
          env -u GITHUB_REPOSITORY ./target/release/stellar-operator run \
            --namespace ${{ env.OPERATOR_NAMESPACE }} \
            > /tmp/operator.log 2>&1 &
          OPERATOR_PID=$!
          echo "Operator PID: $OPERATOR_PID"

          TIMEOUT=60
          ELAPSED=0
          FOUND=false
          while [[ $ELAPSED -lt $TIMEOUT ]]; do
            if grep -q "Connected to Kubernetes cluster" /tmp/operator.log 2>/dev/null; then
              FOUND=true
              break
            fi
            sleep 1
            ELAPSED=$((ELAPSED + 1))
          done

          echo "--- Operator log ---"
          cat /tmp/operator.log || true
          echo "--------------------"

          kill "$OPERATOR_PID" 2>/dev/null || true

          if [[ "$FOUND" == "true" ]]; then
            echo "✅ Log line 'Connected to Kubernetes cluster' confirmed"
          else
            echo "❌ Log line 'Connected to Kubernetes cluster' NOT found within ${TIMEOUT}s"
            exit 1
          fi

      # ── Runtime error check ───────────────────────────────────────────────
      - name: Report runtime errors
        if: always() && steps.scope.outputs.run == 'true'
        run: |
          echo "--- Runtime error check ---"
          if grep -iE "error|panic|missing env" /tmp/operator.log 2>/dev/null; then
            echo "⚠️  Errors found in operator log (see above)"
          else
            echo "✅ No runtime errors detected"
          fi

      # ── Upload log artifact (always-on boot log) ──────────────────────────
      - name: Upload operator log
        if: always() && steps.scope.outputs.run == 'true'
        uses: actions/upload-artifact@v7
        with:
          name: operator-boot-log-${{ github.run_id }}
          path: /tmp/operator.log
          retention-days: 7

      # ── Unified failure diagnostics bundle (issue #1151) ─────────────────
      - name: Upload failure diagnostics bundle
        if: failure() && steps.scope.outputs.run == 'true'
        uses: ./.github/actions/collect-failure-diagnostics
        with:
          artifact-name: ci-diagnostics-verify-operator-boot-${{ github.run_id }}
          operator-namespace: stellar-system
          extra-paths: |
            /tmp/operator.log
          retention-days: "14"

      # ── Cleanup ───────────────────────────────────────────────────────────
      - name: Delete kind cluster
        if: always() && steps.scope.outputs.run == 'true'
        run: kind delete cluster --name ${{ env.CLUSTER_NAME }} || true

      - name: Skip message
        if: steps.scope.outputs.run != 'true'
        run: echo "Skipping verify-operator-boot — no operator/runtime files changed."