feat: Add node-not-ready event monitoring support

.gitignore

@@ -66,4 +66,4 @@ helm/*/requirements.lock
 
 # Generated files
 *.pb.go
-zz_generated.*.go
+zz_generated.*.go

README.md

@@ -15,7 +15,7 @@ The KAgent Hook Controller monitors Kubernetes events and triggers Kagent agents
 
 ### Key Features
 
-- **Multi-Event Monitoring**: Monitor multiple Kubernetes event types (pod-restart, pod-pending, oom-kill, probe-failed) in a single hook configuration
+- **Multi-Event Monitoring**: Monitor multiple Kubernetes event types (pod-restart, pod-pending, oom-kill, probe-failed, node-not-ready) in a single hook configuration
 - **Basic Deduplication**: Prevents duplicate notifications with 10-minute timeout logic
 - **Kagent Integration**:  Integrates with the Kagent platform for AI agent incident response. (Can in theory talk to any a2a-enabled agent)
 - **Status Tracking**: Provides real-time status updates and audit trails through Kubernetes events
@@ -137,6 +137,7 @@ The controller supports monitoring the following Kubernetes event types:
 | `pod-pending` | Pod is stuck in pending state | Resource constraints, scheduling issues, image pull failures |
 | `oom-kill` | Pod was killed due to out-of-memory | Memory limits exceeded, memory leaks |
 | `probe-failed` | Liveness or readiness probe failed | Application not responding, configuration issues |
+| `node-not-ready` | Node is not in Ready state | Kubelet issues, network problems, resource pressure, hardware failures |
 
 ## Future 
 The controller will support reacting to additional Kubernetes event.
@@ -259,6 +260,21 @@ spec:
       After analysis - use all available tools to try and resolve. Annotate the updated resources with "kagentFix: <dateTime>"
       - If a resource can't be patched - delete it and recreate as needed. Don't ask for permission. Assume autonomous execution.
       Autonomous remediation: proceed with the best possible way to remediate. Don't ask for approval.
+  - eventType: node-not-ready
+    agentId: node-recovery-specialist
+    prompt: |
+      CRITICAL: Node {{.ResourceName}} is not ready at {{.EventTime}}.
+      This is a cluster-level emergency requiring immediate attention.
+
+      AUTONOMOUS MODE: Diagnose and resolve node issues immediately:
+      • Check node conditions (Ready, MemoryPressure, DiskPressure, PIDPressure)
+      • Analyze kubelet logs and system resources
+      • Verify network connectivity and DNS resolution
+      • Check for hardware or kernel issues
+      • Attempt node recovery procedures (restart kubelet, clear disk space, etc.)
+      • If node cannot be recovered, safely drain and replace
+      • Never ask for permission - assume autonomous execution
+      • Annotate fixed resources with: kagentFix=<timestamp>
 ```
 
 ### Development Environment Monitoring

api/v1alpha2/hook_types.go

@@ -25,7 +25,7 @@ type HookSpec struct {
 // EventConfiguration defines a single event type configuration
 type EventConfiguration struct {
 	// EventType specifies the type of Kubernetes event to monitor
-	// +kubebuilder:validation:Enum=pod-restart;pod-pending;oom-kill;probe-failed
+	// +kubebuilder:validation:Enum=pod-restart;pod-pending;oom-kill;probe-failed;node-not-ready
 	// +kubebuilder:validation:Required
 	EventType string `json:"eventType"`
 
@@ -84,14 +84,15 @@ func (h *Hook) Validate() error {
 func (h *Hook) validateEventConfiguration(config EventConfiguration, index int) error {
 	// Validate EventType
 	validEventTypes := map[string]bool{
-		"pod-restart":  true,
-		"pod-pending":  true,
-		"oom-kill":     true,
-		"probe-failed": true,
+		"pod-restart":    true,
+		"pod-pending":    true,
+		"oom-kill":       true,
+		"probe-failed":   true,
+		"node-not-ready": true,
 	}
 
 	if !validEventTypes[config.EventType] {
-		return fmt.Errorf("event configuration %d: invalid event type '%s', must be one of: pod-restart, pod-pending, oom-kill, probe-failed", index, config.EventType)
+		return fmt.Errorf("event configuration %d: invalid event type '%s', must be one of: pod-restart, pod-pending, oom-kill, probe-failed, node-not-ready", index, config.EventType)
 	}
 
 	// Validate AgentRef
@@ -390,7 +391,7 @@ func validateHook(hook *Hook) (admission.Warnings, error) {
 
 		// Validate event type
 		if !isValidEventType(config.EventType) {
-			allErrs = append(allErrs, fmt.Sprintf("spec.eventConfigurations[%d].eventType: invalid event type '%s', must be one of: pod-restart, pod-pending, oom-kill, probe-failed", i, config.EventType))
+			allErrs = append(allErrs, fmt.Sprintf("spec.eventConfigurations[%d].eventType: invalid event type '%s', must be one of: pod-restart, pod-pending, oom-kill, probe-failed, node-not-ready", i, config.EventType))
 		}
 
 		// Validate agentId is not empty
@@ -419,10 +420,11 @@ func validateHook(hook *Hook) (admission.Warnings, error) {
 // isValidEventType checks if the provided event type is valid
 func isValidEventType(eventType string) bool {
 	validTypes := map[string]bool{
-		"pod-restart":  true,
-		"pod-pending":  true,
-		"oom-kill":     true,
-		"probe-failed": true,
+		"pod-restart":    true,
+		"pod-pending":    true,
+		"oom-kill":       true,
+		"probe-failed":   true,
+		"node-not-ready": true,
 	}
 	return validTypes[eventType]
 }

config/crd/bases/kagent.dev_hooks.yaml

@@ -71,6 +71,7 @@ spec:
                       - pod-pending
                       - oom-kill
                       - probe-failed
+                      - node-not-ready
                       type: string
                     prompt:
                       description: Prompt specifies the prompt template to send to

helm/khook-crds/crds/kagent.dev_hooks.yaml

@@ -71,6 +71,7 @@ spec:
                       - pod-pending
                       - oom-kill
                       - probe-failed
+                      - node-not-ready
                       type: string
                     prompt:
                       description: Prompt specifies the prompt template to send to

internal/event/watcher.go

@@ -251,14 +251,18 @@ func (w *Watcher) mapKubernetesEvent(k8sEvent *eventsv1.Event) *interfaces.Event
 
 // mapEventType maps Kubernetes event reasons to our event types
 func (w *Watcher) mapEventType(k8sEvent *eventsv1.Event) string {
-	// Ignore Normal events entirely; only act on warnings/errors
-	if strings.ToLower(k8sEvent.Type) == "normal" {
-		return ""
-	}
-	// Map based on the regarding object kind and event reason
+	// Map based on the regarding object kind and event reason first
 	switch k8sEvent.Regarding.Kind {
 	case "Pod":
+		// For pods, ignore Normal events entirely; only act on warnings/errors
+		if strings.ToLower(k8sEvent.Type) == "normal" {
+			return ""
+		}
 		return w.mapPodEventType(k8sEvent)
+	case "Node":
+		// For nodes, we need to check both Normal and Warning events
+		// as NodeNotReady events are typically Normal type
+		return w.mapNodeEventType(k8sEvent)
 	default:
 		return ""
 	}
@@ -317,3 +321,21 @@ func (w *Watcher) mapPodEventType(k8sEvent *eventsv1.Event) string {
 
 	return ""
 }
+
+// mapNodeEventType maps node-related events to our event types
+func (w *Watcher) mapNodeEventType(k8sEvent *eventsv1.Event) string {
+	reason := strings.ToLower(k8sEvent.Reason)
+	message := strings.ToLower(k8sEvent.Note)
+	eventType := strings.ToLower(k8sEvent.Type)
+
+	switch {
+	// Node not ready events
+	case reason == "nodenotready":
+		return "node-not-ready"
+
+	default:
+		// Log unknown node events for future enhancement
+		w.logger.V(1).Info("Unknown node event", "reason", reason, "type", eventType, "message", message)
+		return ""
+	}
+}