boc/aamos-admin-upgrade/backend/handlers/metrics.go

package handlers

import (
	"bufio"
	"encoding/json"
	"fmt"
	"net/http"
	"os"
	"strconv"
	"strings"
	"syscall"
	"time"
)

type MetricsResponse struct {
	Timestamp string      `json:"timestamp"`
	CPU       CPUMetrics  `json:"cpu"`
	RAM       RAMMetrics  `json:"ram"`
	Disk      DiskMetrics `json:"disk"`
	Processes ProcMetrics `json:"processes"`
}

type CPUMetrics struct {
	UsagePercent float64 `json:"usage_percent"`
}

type RAMMetrics struct {
	TotalBytes     uint64  `json:"total_bytes"`
	UsedBytes      uint64  `json:"used_bytes"`
	FreeBytes      uint64  `json:"free_bytes"`
	AvailableBytes uint64  `json:"available_bytes"`
	UsagePercent   float64 `json:"usage_percent"`
}

type DiskMetrics struct {
	Path         string  `json:"path"`
	TotalBytes   uint64  `json:"total_bytes"`
	UsedBytes    uint64  `json:"used_bytes"`
	FreeBytes    uint64  `json:"free_bytes"`
	UsagePercent float64 `json:"usage_percent"`
}

type ProcMetrics struct {
	Total   int `json:"total"`
	Running int `json:"running"`
}

func MetricsHandler(w http.ResponseWriter, r *http.Request) {
	if r.Method != http.MethodGet {
		http.Error(w, "method not allowed", http.StatusMethodNotAllowed)
		return
	}

	cpu, err := readCPU()
	if err != nil {
		http.Error(w, fmt.Sprintf("cpu read error: %v", err), http.StatusInternalServerError)
		return
	}

	ram, err := readRAM()
	if err != nil {
		http.Error(w, fmt.Sprintf("ram read error: %v", err), http.StatusInternalServerError)
		return
	}

	disk, err := readDisk("/")
	if err != nil {
		http.Error(w, fmt.Sprintf("disk read error: %v", err), http.StatusInternalServerError)
		return
	}

	procs, err := readProcesses()
	if err != nil {
		http.Error(w, fmt.Sprintf("proc read error: %v", err), http.StatusInternalServerError)
		return
	}

	resp := MetricsResponse{
		Timestamp: time.Now().UTC().Format(time.RFC3339),
		CPU:       cpu,
		RAM:       ram,
		Disk:      disk,
		Processes: procs,
	}

	w.Header().Set("Content-Type", "application/json")
	json.NewEncoder(w).Encode(resp)
}

// readCPU samples /proc/stat twice with a short sleep and returns usage %.
type cpuSample struct {
	user, nice, system, idleTime, iowait, irq, softirq, steal uint64
}

func parseCPUSample() (cpuSample, error) {
	f, err := os.Open("/proc/stat")
	if err != nil {
		return cpuSample{}, err
	}
	defer f.Close()

	scanner := bufio.NewScanner(f)
	for scanner.Scan() {
		line := scanner.Text()
		if !strings.HasPrefix(line, "cpu ") {
			continue
		}
		fields := strings.Fields(line)
		if len(fields) < 8 {
			return cpuSample{}, fmt.Errorf("unexpected cpu line: %s", line)
		}
		var s cpuSample
		vals := []*uint64{&s.user, &s.nice, &s.system, &s.idleTime, &s.iowait, &s.irq, &s.softirq, &s.steal}
		for i, vp := range vals {
			v, err := strconv.ParseUint(fields[i+1], 10, 64)
			if err != nil {
				return cpuSample{}, err
			}
			*vp = v
		}
		return s, nil
	}
	return cpuSample{}, fmt.Errorf("/proc/stat: cpu line not found")
}

func (s cpuSample) idleTotal() uint64 { return s.idleTime + s.iowait }
func (s cpuSample) total() uint64 {
	return s.user + s.nice + s.system + s.idleTime + s.iowait + s.irq + s.softirq + s.steal
}

func readCPU() (CPUMetrics, error) {
	s1, err := parseCPUSample()
	if err != nil {
		return CPUMetrics{}, err
	}
	time.Sleep(200 * time.Millisecond)
	s2, err := parseCPUSample()
	if err != nil {
		return CPUMetrics{}, err
	}

	totalDelta := s2.total() - s1.total()
	idleDelta := s2.idleTotal() - s1.idleTotal()

	if totalDelta == 0 {
		return CPUMetrics{UsagePercent: 0}, nil
	}
	usage := 100.0 * float64(totalDelta-idleDelta) / float64(totalDelta)
	return CPUMetrics{UsagePercent: roundTwo(usage)}, nil
}

func readRAM() (RAMMetrics, error) {
	f, err := os.Open("/proc/meminfo")
	if err != nil {
		return RAMMetrics{}, err
	}
	defer f.Close()

	info := make(map[string]uint64)
	scanner := bufio.NewScanner(f)
	for scanner.Scan() {
		fields := strings.Fields(scanner.Text())
		if len(fields) < 2 {
			continue
		}
		key := strings.TrimSuffix(fields[0], ":")
		val, err := strconv.ParseUint(fields[1], 10, 64)
		if err != nil {
			continue
		}
		info[key] = val * 1024 // kB → bytes
	}

	total := info["MemTotal"]
	free := info["MemFree"]
	available := info["MemAvailable"]
	used := total - free - info["Buffers"] - info["Cached"] - info["SReclaimable"] + info["Shmem"]

	var pct float64
	if total > 0 {
		pct = roundTwo(100.0 * float64(used) / float64(total))
	}

	return RAMMetrics{
		TotalBytes:     total,
		UsedBytes:      used,
		FreeBytes:      free,
		AvailableBytes: available,
		UsagePercent:   pct,
	}, nil
}

func readDisk(path string) (DiskMetrics, error) {
	var stat syscall.Statfs_t
	if err := syscall.Statfs(path, &stat); err != nil {
		return DiskMetrics{}, err
	}

	total := stat.Blocks * uint64(stat.Bsize)
	free := stat.Bfree * uint64(stat.Bsize)
	avail := stat.Bavail * uint64(stat.Bsize)
	used := total - free

	var pct float64
	if total > 0 {
		pct = roundTwo(100.0 * float64(used) / float64(total))
	}

	return DiskMetrics{
		Path:         path,
		TotalBytes:   total,
		UsedBytes:    used,
		FreeBytes:    avail,
		UsagePercent: pct,
	}, nil
}

// readProcesses counts entries in /proc that are numeric (one per process).
func readProcesses() (ProcMetrics, error) {
	entries, err := os.ReadDir("/proc")
	if err != nil {
		return ProcMetrics{}, err
	}

	total := 0
	running := 0
	for _, e := range entries {
		if !e.IsDir() {
			continue
		}
		if _, err := strconv.Atoi(e.Name()); err != nil {
			continue
		}
		total++

		statusPath := fmt.Sprintf("/proc/%s/status", e.Name())
		data, err := os.ReadFile(statusPath)
		if err != nil {
			continue
		}
		for _, line := range strings.Split(string(data), "\n") {
			if strings.HasPrefix(line, "State:") {
				// R = running, S = sleeping, D = disk sleep, Z = zombie, T = stopped
				if strings.Contains(line, "R (running)") {
					running++
				}
				break
			}
		}
	}

	return ProcMetrics{Total: total, Running: running}, nil
}

func roundTwo(v float64) float64 {
	return float64(int(v*100+0.5)) / 100
}