package main

import (
	"context"
	"encoding/json"
	"fmt"
	"io/ioutil"
	"log"
	"net/http"
	"os"
	"strconv"
	"strings"
	"time"
	"golang.org/x/crypto/ssh"
	v1 "k8s.io/api/core/v1"
        "k8s.io/apimachinery/pkg/types"
	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
	"k8s.io/client-go/kubernetes"
	"k8s.io/client-go/rest"
	metrics "k8s.io/metrics/pkg/client/clientset/versioned"
	"gopkg.in/yaml.v3"
)

// Node and NodePool structs
type Node struct {
	Name        string  `yaml:"name" json:"name"`
	IP          string  `yaml:"ip" json:"ip"`
	Status      string  `yaml:"status" json:"status"`
	CPU         int     `yaml:"cpu" json:"cpu"`
	Memory      int     `yaml:"memory" json:"memory"`
	Role        string  `yaml:"role" json:"role"`
	Cluster     string  `yaml:"cluster" json:"cluster"`
	LastActive  string  `yaml:"last_active" json:"last_active"`
	Pods        int     `yaml:"pods,omitempty" json:"pods,omitempty"`
	Temperature float64 `yaml:"temperature,omitempty" json:"temperature,omitempty"`
}

type NodePool struct {
	Nodes []Node `yaml:"nodes" json:"nodes"`
}

// --- ENV helpers ---
func mustIntEnv(name string, def int) int {
	if val := os.Getenv(name); val != "" {
		if i, err := strconv.Atoi(val); err == nil {
			return i
		}
	}
	return def
}

func mustEnv(key, def string) string {
        if val := os.Getenv(key); val != "" {
                return val
        }
        return def
}

// --- YAML load/save ---
func loadPool(file string) (*NodePool, error) {
	data, err := ioutil.ReadFile(file)
	if err != nil {
		return nil, err
	}
	var pool NodePool
	err = yaml.Unmarshal(data, &pool)
	if err != nil {
		return nil, err
	}
	return &pool, nil
}

func savePool(file string, pool *NodePool) error {
	data, err := yaml.Marshal(pool)
	if err != nil {
		return err
	}
	return ioutil.WriteFile(file, data, 0644)
}

func initNodePool(clientset *kubernetes.Clientset, poolFile string) error {
    ctx := context.Background()
    // Load pool from file — fail if cannot read
    pool, err := loadPool(poolFile)
    if err != nil {
        return fmt.Errorf("cannot load existing node pool file %q: %w", poolFile, err)
    }

    // Map current cluster nodes for quick lookup
    clusterNodes, err := clientset.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
    if err != nil {
        return fmt.Errorf("failed to list nodes: %w", err)
    }

    clusterMap := map[string]v1.Node{}
    for _, n := range clusterNodes.Items {
        clusterMap[n.Name] = n
    }

    // 🟧 Update pool in place, preserving order
    for i := range pool.Nodes {
        n := &pool.Nodes[i]
        if kn, ok := clusterMap[n.Name]; ok {
            // Update only cluster nodes
            n.IP = nodeIP(&kn)
            n.Status = "online"
            n.Cluster = os.Getenv("CLUSTER_NAME")
            n.LastActive = time.Now().Format(time.RFC3339)
            if isControlPlane(&kn) {
                n.Role = "microk8s-controlplane"
            } else {
                n.Role = "worker"
            }
            // Preserve CPU, Memory, Pods, Temperature if previously set
        }
        // else: node not in cluster → leave everything untouched
    }

    // Save back to disk
    return savePool(poolFile, pool)
}
// --- SSH ---
func runSSH(host, user, pass, cmd string) (string, error) {
	config := &ssh.ClientConfig{
		User:            user,
		Auth:            []ssh.AuthMethod{ssh.Password(pass)},
		HostKeyCallback: ssh.InsecureIgnoreHostKey(),
		Timeout:         10 * time.Second,
	}
	client, err := ssh.Dial("tcp", fmt.Sprintf("%s:22", host), config)
	if err != nil {
		return "", err
	}
	defer client.Close()
	session, err := client.NewSession()
	if err != nil {
		return "", err
	}
	defer session.Close()
	out, err := session.CombinedOutput(cmd)
	return string(out), err
}

// --- Node helpers ---
func isControlPlane(n *v1.Node) bool {
	if _, ok := n.Labels["node.kubernetes.io/microk8s-controlplane"]; ok {
		return true
	}
	if _, ok := n.Labels["node-role.kubernetes.io/control-plane"]; ok {
		return true
	}
	return false
}

func nodeIP(n *v1.Node) string {
	for _, addr := range n.Status.Addresses {
		if addr.Type == v1.NodeInternalIP {
			return addr.Address
		}
	}
	return ""
}

// --- Utilization ---

// --- Utilization ---
// updateNodePoolupdates CPU, Memory, Pods, Temperature, Status, and Role
// only for nodes present in the current Kubernetes cluster.
// Nodes not in the cluster remain untouched.
func updateNodePool(clientset *kubernetes.Clientset, metricsClient *metrics.Clientset, poolFile, sshUser, sshPass string) error {
    ctx := context.Background()

    // 1️⃣ Load the pool file from disk
    pool, err := loadPool(poolFile)
    if err != nil {
        return fmt.Errorf("failed to load node pool: %w", err)
    }

    // 2️⃣ List cluster nodes
    clusterNodes, err := clientset.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
    if err != nil {
        return fmt.Errorf("failed to list cluster nodes: %w", err)
    }

    // Map cluster nodes for quick lookup
    clusterMap := map[string]*v1.Node{}
    for i := range clusterNodes.Items {
        clusterMap[clusterNodes.Items[i].Name] = &clusterNodes.Items[i]
    }

    // 3️⃣ List metrics
    metricsList, err := metricsClient.MetricsV1beta1().NodeMetricses().List(ctx, metav1.ListOptions{})
    if err != nil {
        return fmt.Errorf("failed to get node metrics: %w", err)
    }

    usageCPU := map[string]int64{}
    usageMem := map[string]int64{}
    for _, m := range metricsList.Items {
        usageCPU[m.Name] = m.Usage.Cpu().MilliValue()
        usageMem[m.Name] = m.Usage.Memory().Value()
    }

    // Node capacities
    capCPU := map[string]int64{}
    capMem := map[string]int64{}
    for _, n := range clusterNodes.Items {
        capCPU[n.Name] = n.Status.Capacity.Cpu().MilliValue()
        capMem[n.Name] = n.Status.Capacity.Memory().Value()
    }

    // Pods per node
    pods, err := clientset.CoreV1().Pods("").List(ctx, metav1.ListOptions{})
    if err != nil {
        return err
    }
    podCount := map[string]int{}
    for _, p := range pods.Items {
        if p.Spec.NodeName != "" {
            podCount[p.Spec.NodeName]++
        }
    }

    // 4️⃣ Update pool in place
    for i := range pool.Nodes {
        n := &pool.Nodes[i]

        if kn, ok := clusterMap[n.Name]; ok {
            // Update CPU and Memory %
            if cpuCap, ok1 := capCPU[n.Name]; ok1 {
                if memCap, ok2 := capMem[n.Name]; ok2 {
                    if cpuUse, ok3 := usageCPU[n.Name]; ok3 {
                        if memUse, ok4 := usageMem[n.Name]; ok4 {
                            if cpuCap > 0 && memCap > 0 {
                                n.CPU = int((cpuUse * 100) / cpuCap)
                                n.Memory = int((memUse * 100) / memCap)
                            }
                        }
                    }
                }
            }

            // Update pod count
            n.Pods = podCount[n.Name]

            // Update temperature via SSH
            n.Temperature = getNodeTemp(kn, sshUser, sshPass)

            // Update status/role/cluster/last_active
            n.Status = "online"
            n.Cluster = os.Getenv("CLUSTER_NAME")
            n.LastActive = time.Now().Format(time.RFC3339)
            if isControlPlane(kn) {
                n.Role = "microk8s-controlplane"
            } else {
                n.Role = "worker"
            }
        }
        // else: node not in cluster → leave all fields untouched
    }

    // 5️⃣ Save pool back to disk
    return savePool(poolFile, pool)
}
// --- Temperature ---
func getNodeTemp(node *v1.Node, sshUser, sshPass string) float64 {
	ip := nodeIP(node)
	out, err := runSSH(ip, sshUser, sshPass, "cat /proc/device-tree/model")
	if err != nil {
                log.Printf("getNodeTemp: failed to read model from %s: %v", ip, err)
		return 0
	}
	hw := strings.ToUpper(strings.TrimSpace(out))

	var cmd string
	switch {
	case strings.Contains(hw, "RASPBERRY"):
		cmd = "vcgencmd measure_temp | egrep -o '[0-9]+\\.[0-9]+'"
	case strings.Contains(hw, "ODROID"):
		cmd = "awk '{printf \"%3.1f\", $1/1000}' /sys/class/thermal/thermal_zone0/temp"
	default:
		return 0
	}

	tempStr, err := runSSH(ip, sshUser, sshPass, cmd)
	if err != nil {
                log.Printf("getNodeTemp: failed to read temperature from %s: %v", ip, err)
		return 0
	}

	t, err := strconv.ParseFloat(strings.TrimSpace(tempStr), 64)
	if err != nil {
		return 0
	}
	return t
}

// --- Control-plane management ---
func ensureControlPlanes(cs *kubernetes.Clientset, poolFile, sshUser, sshPass, clusterName string, desired int) {
    // Load the current node pool from disk
    pool, err := loadPool(poolFile)
    if err != nil {
        log.Println("Cannot load pool in ensureControlPlanes:", err)
        return
    }

    ctx := context.Background()

    // List nodes currently in this Kubernetes cluster
    nodes, err := cs.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
    if err != nil {
        log.Println("failed to list nodes:", err)
        return
    }

    // Count current control-plane nodes
    cpCount := 0
    var seedIP string
    for _, n := range nodes.Items {
        if isControlPlane(&n) {
            cpCount++
            if seedIP == "" {
                seedIP = nodeIP(&n)
            }
        }
    }

    // Nothing to do if enough control-planes
    if cpCount >= desired {
        return
    }

    if seedIP == "" {
        log.Println("no available control-plane node found")
        return
    }

    // Find an offline node in the pool to activate
    for i := range pool.Nodes {
        n := &pool.Nodes[i]
        if n.Status != "offline" {
            continue
        }

        log.Printf("Attempting to activate node %s as control-plane", n.Name)

        // Generate join command from seed control-plane
        out, err := runSSH(seedIP, sshUser, sshPass, "microk8s add-node")
        if err != nil {
            log.Println("add-node failed:", err)
            return
        }

        var joinCmd string
        for _, line := range strings.Split(out, "\n") {
            if strings.HasPrefix(strings.TrimSpace(line), "microk8s join") {
                joinCmd = strings.TrimSpace(line)
                break
            }
        }

        if joinCmd == "" {
            log.Println("no join command found")
            return
        }

        // Run join on offline node
        _, err = runSSH(n.IP, sshUser, sshPass, joinCmd)
        if err != nil {
            log.Println("join failed:", err)
            return
        }

        // Update node in pool
        n.Status = "online"
        n.Cluster = clusterName
        n.Role = "microk8s-controlplane"
        n.LastActive = time.Now().Format(time.RFC3339)
        log.Printf("Control-plane %s activated", n.Name)

        // Save back updated pool to disk
        if err := savePool(poolFile, pool); err != nil {
            log.Println("Failed to save pool after activating control-plane:", err)
        }

        return
    }
}
// --- Worker management ---
func activateOneWorker(cs *kubernetes.Clientset, pool *NodePool, poolFile, sshUser, sshPass, clusterName string) {
	ctx := context.Background()
	var workerNode *Node
	for i := range pool.Nodes {
		n := &pool.Nodes[i]
		if n.Status == "offline" {
			workerNode = &pool.Nodes[i]
			break
		}
	}

	if workerNode == nil {
		log.Println("No offline nodes available — skipping activation")
		return
	}

	log.Printf("Attempting to activate %s node as worker node", workerNode.Name)
	nodes, err := cs.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
	if err != nil {
		log.Println("Failed to list nodes:", err)
		return
	}

	var seedIP string
	for _, n := range nodes.Items {
		if isControlPlane(&n) {
			seedIP = nodeIP(&n)
			break
		}
	}

	if seedIP == "" {
		log.Println("No control-plane node available to generate join command")
		return
	}

	out, err := runSSH(seedIP, sshUser, sshPass, "microk8s add-node")
	if err != nil {
		log.Println("add-node failed:", err)
		return
	}

	var joinCmd string
	for _, line := range strings.Split(out, "\n") {
		if strings.HasPrefix(strings.TrimSpace(line), "microk8s join") {
			joinCmd = strings.TrimSpace(line)
			break
		}
	}

	if joinCmd == "" {
		log.Println("No join command found from seed control-plane")
		return
	}

	joinCmd += " --worker"
	_, err = runSSH(workerNode.IP, sshUser, sshPass, joinCmd)
	if err != nil {
		log.Println("Worker join failed:", err)
		return
	}

	workerNode.Status = "online"
	workerNode.Cluster = clusterName
	workerNode.Role = "worker"
	workerNode.LastActive = time.Now().Format(time.RFC3339)
	log.Printf("Node %s successfully activated as worker node", workerNode.Name)
}

func deactivateOneWorkerSafe(cs *kubernetes.Clientset, pool *NodePool, poolFile, sshUser, sshPass, clusterName string, waitSec int) {
	ctx := context.Background()
	nodes, err := cs.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
	if err != nil {
		log.Println("failed to list nodes:", err)
		return
	}

	var workerNode *v1.Node
	for _, n := range nodes.Items {
		if !isControlPlane(&n) && !n.Spec.Unschedulable {
			workerNode = &n
			break
		}
	}

	if workerNode == nil {
		log.Println("No worker nodes available for deactivation")
		return
	}

	log.Printf("Deactivating node: %s", workerNode.Name)
	ip := nodeIP(workerNode)

	patch := []byte(`{"spec":{"unschedulable":true}}`)
	_, err = cs.CoreV1().Nodes().Patch(ctx, workerNode.Name, types.StrategicMergePatchType, patch, metav1.PatchOptions{})
	if err != nil {
		log.Println("Failed to cordon node:", err)
		return
	}

	err = drainNode(cs, workerNode.Name)
	if err != nil {
		log.Println("Drain failed, uncordoning node")
		patch = []byte(`{"spec":{"unschedulable":false}}`)
		_, _ = cs.CoreV1().Nodes().Patch(ctx, workerNode.Name, types.StrategicMergePatchType, patch, metav1.PatchOptions{})
		return
	}

	_, err = runSSH(ip, sshUser, sshPass, "microk8s leave")
	if err != nil {
		log.Println("microk8s leave failed, keeping node object in cluster")
		return
	}

	_ = cs.CoreV1().Nodes().Delete(ctx, workerNode.Name, metav1.DeleteOptions{})

	for i := range pool.Nodes {
		if pool.Nodes[i].Name == workerNode.Name {
			pool.Nodes[i].Status = "offline"
			pool.Nodes[i].Cluster = "none"
			pool.Nodes[i].Role = "none"
			pool.Nodes[i].CPU = 0
			pool.Nodes[i].Memory = 0
			pool.Nodes[i].LastActive = time.Now().Format(time.RFC3339)
		}
	}
}

// --- Drain node ---
func drainNode(cs *kubernetes.Clientset, nodeName string) error {
	ctx := context.Background()
	pods, err := cs.CoreV1().Pods("").List(ctx, metav1.ListOptions{
		FieldSelector: fmt.Sprintf("spec.nodeName=%s", nodeName),
	})
	if err != nil {
		return err
	}

	for _, pod := range pods.Items {
		if _, ok := pod.ObjectMeta.Annotations["kubernetes.io/config.mirror"]; ok {
			continue
		}
		if pod.Namespace == "kube-system" {
			continue
		}
		grace := int64(60)
		_ = cs.CoreV1().Pods(pod.Namespace).Delete(ctx, pod.Name, metav1.DeleteOptions{
			GracePeriodSeconds: &grace,
		})
	}

	for {
		remaining, _ := cs.CoreV1().Pods("").List(ctx, metav1.ListOptions{
			FieldSelector: fmt.Sprintf("spec.nodeName=%s", nodeName),
		})
		active := 0
		for _, p := range remaining.Items {
			if p.Namespace != "kube-system" && p.DeletionTimestamp == nil {
				active++
			}
		}
		if active == 0 {
			break
		}
		time.Sleep(5 * time.Second)
	}

	return nil
}

// --- Cluster utilization ---
func clusterUtilization(cs *kubernetes.Clientset, ms *metrics.Clientset) (cpuPct int, memPct int, err error) {
	ctx := context.Background()
	nodes, err := cs.CoreV1().Nodes().List(ctx, metav1.ListOptions{})
	if err != nil {
		return 0, 0, err
	}

	metricsList, err := ms.MetricsV1beta1().NodeMetricses().List(ctx, metav1.ListOptions{})
	if err != nil {
		return 0, 0, err
	}

	var totalCPUCap, totalMemCap, totalCPUUse, totalMemUse int64
	for _, n := range nodes.Items {
		totalCPUCap += n.Status.Capacity.Cpu().MilliValue()
		totalMemCap += n.Status.Capacity.Memory().Value()
	}

	for _, m := range metricsList.Items {
		totalCPUUse += m.Usage.Cpu().MilliValue()
		totalMemUse += m.Usage.Memory().Value()
	}

	cpuPct = int((totalCPUUse * 100) / totalCPUCap)
	memPct = int((totalMemUse * 100) / totalMemCap)
	return cpuPct, memPct, nil
}

// --- Web GUI ---
// startWebGUI starts a simple HTTP server showing the node pool
func startWebGUI(poolFile string) {
    // API endpoint for JSON status

    http.HandleFunc("/status", func(w http.ResponseWriter, r *http.Request) {
        pool, err := loadPool(poolFile)  // always fresh from disk
        if err != nil {
            log.Println("Cannot load node pool:", err)
            http.Error(w, "cannot load node pool", http.StatusInternalServerError)
            return
        }
        w.Header().Set("Content-Type", "application/json")
        json.NewEncoder(w).Encode(pool)
    })

    // Serve index.html and static assets (single file)
    http.Handle("/", http.FileServer(http.Dir("/app/web")))

    go func() {
        log.Println("Web GUI running at :8080")
        if err := http.ListenAndServe(":8080", nil); err != nil {
            log.Fatal(err)
        }
    }()
}
// --- Main ---
func main() {
    poolFile := mustEnv("NODE_POOL_FILE", "node-pool.yaml")
    sshUser := mustEnv("SSH_USER", "ubuntu")
    sshPass := mustEnv("SSH_PASS", "")
    // read thresholds from environment
    minCPU := mustIntEnv("MIN_CPU", 40)
    minMem := mustIntEnv("MIN_MEM", 50)
    maxCPU := mustIntEnv("MAX_CPU", 90)
    maxMem := mustIntEnv("MAX_MEM", 90)    
    clusterName := mustEnv("CLUSTER_NAME", "cluster1")
    desiredCP := mustIntEnv("DESIRED_CONTROLPLANES", 1)
    // Kubernetes client
    config, err := rest.InClusterConfig()
    if err != nil {
        log.Fatal("Cannot create in-cluster config:", err)
    }

    clientset, err := kubernetes.NewForConfig(config)
    if err != nil {
        log.Fatal("Cannot create clientset:", err)
    }

    metricsClient, err := metrics.NewForConfig(config)
    if err != nil {
        log.Fatal("Cannot create metrics client:", err)
    }

    // Web GUI
    startWebGUI(poolFile)

    // Main control loop
    for {
        // Always load pool fresh from disk
        pool, err := loadPool(poolFile)
        if err != nil {
            log.Println("Cannot load node pool:", err)
            time.Sleep(10 * time.Second)
            continue
        }

        // Update ONLY nodes that exist in THIS cluster
        err = updateNodePool(
            clientset,
            metricsClient,
            poolFile,
            sshUser,
            sshPass,
        )
        if err != nil {
            log.Println("updateNodePool error:", err)
        }
        // Ensure control-plane count
        ensureControlPlanes(
            clientset,
            poolFile,
            sshUser,
            sshPass,
            clusterName,
            desiredCP,
        )
        // inside the main loop, after computing cluster utilization
        cpuPct, memPct, err := clusterUtilization(clientset, metricsClient)
        if err == nil {
            log.Printf("Cluster utilization: CPU %d%%, MEM %d%%", cpuPct, memPct)

            // decide worker activation/deactivation based on thresholds
            if cpuPct > maxCPU || memPct > maxMem {
                activateOneWorker(clientset, pool, poolFile, sshUser, sshPass, clusterName)
              } else if cpuPct < minCPU && memPct < minMem {
                deactivateOneWorkerSafe(clientset, pool, poolFile, sshUser, sshPass, clusterName, 60)
            }
        }

        // Cluster-wide utilization (read-only)
        cpuPct, memPct, err = clusterUtilization(clientset, metricsClient)
        if err == nil {
            log.Printf(
                "Cluster utilization: CPU %d%%, MEM %d%%",
                cpuPct,
                memPct,
            )
        }

        // 💾 Persist pool back to disk (ALL nodes preserved)
        if err := savePool(poolFile, pool); err != nil {
            log.Println("Failed to save node pool:", err)
        }

        time.Sleep(10 * time.Second)
    }
}