Add signal-reboot (#814)

* feat: sentinel-command without nsenter by default

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* fix: no readonly mount

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* fix: mount at different folder

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* feat: add signal-reboot

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* feat: make signal configurable and add tests

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* build: rename job

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* cleanup: linter

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* build: also adjust signal manifest

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* test: add e2e-tests

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* fix: small code restructure

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

* fix: adjust version-range

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

---------

Signed-off-by: Christian Kotzbauer <git@ckotzbauer.de>

.github/workflows/on-pr.yaml

@@ -103,8 +103,9 @@ jobs:
   # - Ensure manifests work with the latest versions even with no manifest change
   #     (compared to helm charts, manifests cannot easily template changes based on versions)
   # Helm charts are _trailing_ releases, while manifests are done during development.
-  e2e-manifests:
-    name: End-to-End test with kured with code and manifests from HEAD
+  # This test uses the "command" reboot-method.
+  e2e-manifests-command:
+    name: End-to-End test with kured with code and manifests from HEAD (command)
     runs-on: ubuntu-latest
     strategy:
       fail-fast: false
@@ -181,6 +182,91 @@ jobs:
           ./tests/kind/follow-coordinated-reboot.sh
 
 
+  # This ensures the latest code works with the manifests built from tree.
+  # It is useful for two things:
+  # - Test manifests changes (obviously), ensuring they don't break existing clusters
+  # - Ensure manifests work with the latest versions even with no manifest change
+  #     (compared to helm charts, manifests cannot easily template changes based on versions)
+  # Helm charts are _trailing_ releases, while manifests are done during development.
+  # This test uses the "signal" reboot-method.
+  e2e-manifests-signal:
+    name: End-to-End test with kured with code and manifests from HEAD (signal)
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        kubernetes:
+          - "1.26"
+          - "1.27"
+          - "1.28"
+    steps:
+      - uses: actions/checkout@v3
+      - name: Ensure go version
+        uses: actions/setup-go@v4
+        with:
+          go-version-file: 'go.mod'
+          check-latest: true
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v2
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+      - name: Setup GoReleaser
+        run: make bootstrap-tools
+      - name: Find current tag version
+        run: echo "sha_short=$(git rev-parse --short HEAD)" >> $GITHUB_OUTPUT
+        id: tags
+      - name: Build artifacts
+        run: |
+          VERSION="${{ steps.tags.outputs.sha_short }}" make image
+          VERSION="${{ steps.tags.outputs.sha_short }}" make manifest
+
+      - name: Workaround "Failed to attach 1 to compat systemd cgroup /actions_job/..." on gh actions
+        run: |
+          sudo bash << EOF
+              cp /etc/docker/daemon.json /etc/docker/daemon.json.old
+              echo '{}' > /etc/docker/daemon.json
+              systemctl restart docker || journalctl --no-pager -n 500
+              systemctl status docker
+          EOF
+
+      # Default name for helm/kind-action kind clusters is "chart-testing"
+      - name: Create kind cluster with 5 nodes
+        uses: helm/kind-action@v1.8.0
+        with:
+          config: .github/kind-cluster-${{ matrix.kubernetes }}.yaml
+          version: v0.14.0
+
+      - name: Preload previously built images onto kind cluster
+        run: kind load docker-image ghcr.io/${{ github.repository }}:${{ steps.tags.outputs.sha_short }} --name chart-testing
+
+      - name: Do not wait for an hour before detecting the rebootSentinel
+        run: |
+          sed -i 's/#\(.*\)--period=1h/\1--period=30s/g' kured-ds-signal.yaml
+
+      - name: Install kured with kubectl
+        run: |
+          kubectl apply -f kured-rbac.yaml && kubectl apply -f kured-ds-signal.yaml
+
+      - name: Ensure kured is ready
+        uses: nick-invision/retry@v2.8.3
+        with:
+          timeout_minutes: 10
+          max_attempts: 10
+          retry_wait_seconds: 60
+          # DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE should all be = to cluster_size
+          command: "kubectl get ds -n kube-system kured | grep -E 'kured.*5.*5.*5.*5.*5'"
+
+      - name: Create reboot sentinel files
+        run: |
+          ./tests/kind/create-reboot-sentinels.sh
+
+      - name: Follow reboot until success
+        env:
+          DEBUG: true
+        run: |
+          ./tests/kind/follow-coordinated-reboot.sh
+
+
 
   # This ensures the latest code works with the manifests built from tree.
   # It is useful for two things: