BootID support for KMM

Author: sriram-30

internal/controllers/mock_nmc_reconciler.go

@@ -49,6 +49,7 @@ type NMCReconciler struct {
 	helper     nmcReconcilerHelper
 	nodeAPI    node.Node
 	podManager pod.WorkerPodManager
+	bootIDMap  map[string]string
 }
 
 func NewNMCReconciler(
@@ -66,6 +67,7 @@ func NewNMCReconciler(
 		helper:     helper,
 		nodeAPI:    nodeAPI,
 		podManager: podManager,
+		bootIDMap:  make(map[string]string),
 	}
 }
 
@@ -110,6 +112,10 @@ func (r *NMCReconciler) Reconcile(ctx context.Context, req reconcile.Request) (r
 
 	errs := make([]error, 0, len(nmcObj.Spec.Modules)+len(nmcObj.Status.Modules))
 	var readyLabelsToRemove []string
+	prevBootID, exists := r.bootIDMap[nmcObj.Name]
+	currentBootID := node.Status.NodeInfo.BootID
+	bootIDChanged := exists && prevBootID != currentBootID
+
 	for _, mod := range nmcObj.Spec.Modules {
 		moduleNameKey := mod.Namespace + "/" + mod.Name
 
@@ -121,7 +127,7 @@ func (r *NMCReconciler) Reconcile(ctx context.Context, req reconcile.Request) (r
 			delete(statusMap, moduleNameKey)
 			continue
 		}
-		if err := r.helper.ProcessModuleSpec(ctrl.LoggerInto(ctx, logger), &nmcObj, &mod, statusMap[moduleNameKey], &node); err != nil {
+		if err := r.helper.ProcessModuleSpec(ctrl.LoggerInto(ctx, logger), &nmcObj, &mod, statusMap[moduleNameKey], &node, bootIDChanged); err != nil {
 			errs = append(
 				errs,
 				fmt.Errorf("error processing Module %s: %v", moduleNameKey, err),
@@ -133,6 +139,8 @@ func (r *NMCReconciler) Reconcile(ctx context.Context, req reconcile.Request) (r
 		delete(statusMap, moduleNameKey)
 	}
 
+	r.bootIDMap[nmcObj.Name] = currentBootID
+
 	// We have processed all module specs.
 	// Now, go through the remaining, "orphan" statuses that do not have a corresponding spec; those must be unloaded.
 
@@ -217,7 +225,7 @@ func GetContainerStatus(statuses []v1.ContainerStatus, name string) v1.Container
 type nmcReconcilerHelper interface {
 	GarbageCollectInUseLabels(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig) error
 	GarbageCollectWorkerPods(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig) error
-	ProcessModuleSpec(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig, spec *kmmv1beta1.NodeModuleSpec, status *kmmv1beta1.NodeModuleStatus, node *v1.Node) error
+	ProcessModuleSpec(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig, spec *kmmv1beta1.NodeModuleSpec, status *kmmv1beta1.NodeModuleStatus, node *v1.Node, bootIDChanged bool) error
 	ProcessUnconfiguredModuleStatus(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig, status *kmmv1beta1.NodeModuleStatus, node *v1.Node) error
 	RemovePodFinalizers(ctx context.Context, nodeName string) error
 	SyncStatus(ctx context.Context, nmc *kmmv1beta1.NodeModulesConfig) error
@@ -344,6 +352,7 @@ func (h *nmcReconcilerHelperImpl) ProcessModuleSpec(
 	spec *kmmv1beta1.NodeModuleSpec,
 	status *kmmv1beta1.NodeModuleStatus,
 	node *v1.Node,
+	bootIDChanged bool,
 ) error {
 	podName := pod.WorkerPodName(nmcObj.Name, spec.Name)
 
@@ -374,7 +383,7 @@ func (h *nmcReconcilerHelperImpl) ProcessModuleSpec(
 			return h.podManager.CreateLoaderPod(ctx, nmcObj, spec)
 		}
 
-		if h.nodeAPI.NodeBecomeReadyAfter(node, status.LastTransitionTime) {
+		if bootIDChanged || h.nodeAPI.NodeBecomeReadyAfter(node, status.LastTransitionTime) {
 			logger.Info("node has been rebooted and become ready after kernel module was loaded; creating loader Pod")
 			return h.podManager.CreateLoaderPod(ctx, nmcObj, spec)
 		}

internal/controllers/nmc_reconciler.go

@@ -298,9 +298,9 @@ var _ = Describe("NodeModulesConfigReconciler_Reconcile", func() {
 			wh.EXPECT().SyncStatus(ctx, nmc),
 			kubeClient.EXPECT().Get(ctx, types.NamespacedName{Name: nmc.Name}, &node).Return(nil),
 			nm.EXPECT().IsNodeSchedulable(&node, nil).Return(true),
-			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec0, &status0, &node),
+			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec0, &status0, &node, false),
 			nm.EXPECT().IsNodeSchedulable(&node, nil).Return(true),
-			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec1, nil, &node),
+			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec1, nil, &node, false),
 			wh.EXPECT().ProcessUnconfiguredModuleStatus(contextWithValueMatch, nmc, &status2, &node),
 			wh.EXPECT().GarbageCollectInUseLabels(ctx, nmc),
 			wh.EXPECT().GarbageCollectWorkerPods(ctx, nmc),
@@ -380,7 +380,7 @@ var _ = Describe("NodeModulesConfigReconciler_Reconcile", func() {
 			wh.EXPECT().SyncStatus(ctx, nmc).Return(nil),
 			kubeClient.EXPECT().Get(ctx, types.NamespacedName{Name: nmc.Name}, &node).Return(nil),
 			nm.EXPECT().IsNodeSchedulable(&node, nil).Return(true),
-			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec0, &status0, &node).Return(errors.New(errorMeassge)),
+			wh.EXPECT().ProcessModuleSpec(contextWithValueMatch, nmc, &spec0, &status0, &node, false).Return(errors.New(errorMeassge)),
 			wh.EXPECT().ProcessUnconfiguredModuleStatus(contextWithValueMatch, nmc, &status2, &node).Return(errors.New(errorMeassge)),
 			wh.EXPECT().GarbageCollectInUseLabels(ctx, nmc).Return(errors.New(errorMeassge)),
 			wh.EXPECT().GarbageCollectWorkerPods(ctx, nmc).Return(errors.New(errorMeassge)),