feat: Implement agent heartbeat with mTLS and node status tracking

2025-05-17 13:32:05 -04:00
parent dad5586339
commit 3408e7801e
7 changed files with 639 additions and 16 deletions
--- a/internal/api/node_status_handler.go
+++ b/internal/api/node_status_handler.go
@ -0,0 +1,108 @@
+package api
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"log"
+	"net/http"
+	"strings"
+	"time"
+
+	"git.dws.rip/dubey/kat/internal/store"
+)
+
+// NodeStatusRequest represents the data sent by an agent in a heartbeat
+type NodeStatusRequest struct {
+	NodeName    string    `json:"nodeName"`
+	NodeUID     string    `json:"nodeUID"`
+	Timestamp   time.Time `json:"timestamp"`
+	Resources   struct {
+		Capacity    map[string]string `json:"capacity"`
+		Allocatable map[string]string `json:"allocatable"`
+	} `json:"resources"`
+	WorkloadInstances []struct {
+		WorkloadName  string `json:"workloadName"`
+		Namespace     string `json:"namespace"`
+		InstanceID    string `json:"instanceID"`
+		ContainerID   string `json:"containerID"`
+		ImageID       string `json:"imageID"`
+		State         string `json:"state"`
+		ExitCode      int    `json:"exitCode"`
+		HealthStatus  string `json:"healthStatus"`
+		Restarts      int    `json:"restarts"`
+	} `json:"workloadInstances,omitempty"`
+	OverlayNetwork struct {
+		Status       string `json:"status"`
+		LastPeerSync string `json:"lastPeerSync"`
+	} `json:"overlayNetwork"`
+}
+
+// NewNodeStatusHandler creates a handler for node status updates
+func NewNodeStatusHandler(stateStore store.StateStore) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		// Extract node name from URL path
+		pathParts := strings.Split(r.URL.Path, "/")
+		if len(pathParts) < 4 {
+			http.Error(w, "Invalid URL path", http.StatusBadRequest)
+			return
+		}
+		nodeName := pathParts[len(pathParts)-2] // /v1alpha1/nodes/{nodeName}/status
+
+		log.Printf("Received status update from node: %s", nodeName)
+
+		// Read and parse the request body
+		body, err := io.ReadAll(r.Body)
+		if err != nil {
+			log.Printf("Failed to read request body: %v", err)
+			http.Error(w, "Failed to read request body", http.StatusBadRequest)
+			return
+		}
+		defer r.Body.Close()
+
+		var statusReq NodeStatusRequest
+		if err := json.Unmarshal(body, &statusReq); err != nil {
+			log.Printf("Failed to parse status request: %v", err)
+			http.Error(w, "Failed to parse status request", http.StatusBadRequest)
+			return
+		}
+
+		// Validate that the node name in the URL matches the one in the request
+		if statusReq.NodeName != nodeName {
+			log.Printf("Node name mismatch: %s (URL) vs %s (body)", nodeName, statusReq.NodeName)
+			http.Error(w, "Node name mismatch", http.StatusBadRequest)
+			return
+		}
+
+		// Store the node status in etcd
+		nodeStatusKey := fmt.Sprintf("/kat/nodes/status/%s", nodeName)
+		nodeStatus := map[string]interface{}{
+			"lastHeartbeat": time.Now().Unix(),
+			"status":        "Ready",
+			"resources":     statusReq.Resources,
+			"network":       statusReq.OverlayNetwork,
+		}
+
+		// Add workload instances if present
+		if len(statusReq.WorkloadInstances) > 0 {
+			nodeStatus["workloadInstances"] = statusReq.WorkloadInstances
+		}
+
+		nodeStatusData, err := json.Marshal(nodeStatus)
+		if err != nil {
+			log.Printf("Failed to marshal node status: %v", err)
+			http.Error(w, "Failed to marshal node status", http.StatusInternalServerError)
+			return
+		}
+
+		log.Printf("Storing node status in etcd at key: %s", nodeStatusKey)
+		if err := stateStore.Put(r.Context(), nodeStatusKey, nodeStatusData); err != nil {
+			log.Printf("Failed to store node status: %v", err)
+			http.Error(w, "Failed to store node status", http.StatusInternalServerError)
+			return
+		}
+
+		log.Printf("Successfully stored status update for node: %s", nodeName)
+		w.WriteHeader(http.StatusOK)
+	}
+}