436 lines
12 KiB
Go
436 lines
12 KiB
Go
package main
|
|
|
|
import (
|
|
"context"
|
|
"encoding/json"
|
|
"fmt"
|
|
"log"
|
|
"net/http"
|
|
"os"
|
|
"os/exec"
|
|
"os/signal"
|
|
"strings"
|
|
"sync"
|
|
"syscall"
|
|
"time"
|
|
|
|
"github.com/NVIDIA/go-nvml/pkg/nvml"
|
|
"github.com/shirou/gopsutil/v3/cpu"
|
|
"github.com/shirou/gopsutil/v3/disk"
|
|
"github.com/shirou/gopsutil/v3/net"
|
|
"github.com/shirou/gopsutil/v3/process"
|
|
)
|
|
|
|
const (
|
|
checkInterval = 10 * time.Second
|
|
monitoringPeriod = 5 * time.Minute
|
|
resumeGracePeriod = 5 * time.Minute // Time to wait after resume before allowing sleep again
|
|
cpuThreshold = 20.0 // percentage
|
|
gpuThreshold = 20.0 // percentage
|
|
diskThreshold = 5 * 1024 * 1024 // 5 MB/s
|
|
networkThreshold = 1 * 1024 * 1024 // 1 MB/s
|
|
httpPort = 8081
|
|
)
|
|
|
|
type ResourceUsage struct {
|
|
Timestamp time.Time `json:"timestamp"`
|
|
CpuUsage float64 `json:"cpu_usage"`
|
|
GpuUsage float64 `json:"gpu_usage"`
|
|
GpuAvailable bool `json:"gpu_available"`
|
|
DiskIO uint64 `json:"disk_io"`
|
|
NetworkIO uint64 `json:"network_io"`
|
|
SshConnections int `json:"ssh_connections"`
|
|
ActiveUsers int `json:"active_users"`
|
|
}
|
|
|
|
type SystemStatus struct {
|
|
CurrentUsage ResourceUsage `json:"current_usage"`
|
|
Blockers []string `json:"sleep_blockers"`
|
|
InGracePeriod bool `json:"in_grace_period,omitempty"`
|
|
GraceTimeLeft string `json:"grace_time_left,omitempty"`
|
|
}
|
|
|
|
var (
|
|
currentStatus SystemStatus
|
|
statusMutex sync.RWMutex
|
|
nvmlAvailable bool
|
|
lastResumeTime time.Time // Track when the system last resumed from sleep
|
|
lastTickTime time.Time // Track when we last processed a tick
|
|
)
|
|
|
|
func main() {
|
|
// Check if running as root
|
|
if os.Geteuid() != 0 {
|
|
log.Fatal("This program must be run as root")
|
|
}
|
|
|
|
// Initialize NVML for GPU monitoring
|
|
ret := nvml.Init()
|
|
if ret != nvml.SUCCESS {
|
|
log.Printf("Warning: Could not initialize NVML: %v", ret)
|
|
nvmlAvailable = false
|
|
} else {
|
|
nvmlAvailable = true
|
|
defer nvml.Shutdown()
|
|
}
|
|
|
|
// Create a context that we'll use to shut down the application
|
|
ctx, cancel := context.WithCancel(context.Background())
|
|
defer cancel()
|
|
|
|
// Set up signal handling
|
|
sigChan := make(chan os.Signal, 1)
|
|
signal.Notify(sigChan, syscall.SIGINT, syscall.SIGTERM)
|
|
|
|
// Start HTTP server with context
|
|
srv := startHTTPServer(ctx)
|
|
|
|
usageHistory := make([]ResourceUsage, 0)
|
|
ticker := time.NewTicker(checkInterval)
|
|
defer ticker.Stop()
|
|
|
|
log.Printf("Starting idle monitoring. System will suspend when:\n")
|
|
log.Printf("- Average CPU usage across all cores < %.1f%%\n", cpuThreshold)
|
|
if nvmlAvailable {
|
|
log.Printf("- Average GPU usage across all GPUs < %.1f%%\n", gpuThreshold)
|
|
} else {
|
|
log.Printf("- GPU monitoring disabled (NVML initialization failed)")
|
|
}
|
|
log.Printf("- Disk I/O < %.1f MB/s\n", float64(diskThreshold)/(1024*1024))
|
|
log.Printf("- Network I/O < %.1f MB/s\n", float64(networkThreshold)/(1024*1024))
|
|
log.Printf("- No active SSH connections\n")
|
|
log.Printf("- No active user sessions\n")
|
|
log.Printf("- Over the last %v\n", monitoringPeriod)
|
|
log.Printf("- System will not suspend for %v after resuming from sleep\n", resumeGracePeriod)
|
|
log.Printf("HTTP status endpoint available at http://localhost:%d/status\n", httpPort)
|
|
log.Printf("Press Ctrl+C to exit\n")
|
|
|
|
mainLoop:
|
|
for {
|
|
select {
|
|
case <-ctx.Done():
|
|
break mainLoop
|
|
case sig := <-sigChan:
|
|
log.Printf("Received signal %v, shutting down...", sig)
|
|
cancel()
|
|
break mainLoop
|
|
case <-ticker.C:
|
|
now := time.Now()
|
|
|
|
// Check if we just resumed from sleep
|
|
if !lastTickTime.IsZero() {
|
|
gap := now.Sub(lastTickTime)
|
|
// If there was a significant gap, probably resumed from sleep
|
|
if gap > (checkInterval*3) && gap < time.Hour {
|
|
log.Printf("Detected system resume after gap of %v", gap)
|
|
lastResumeTime = now
|
|
}
|
|
}
|
|
lastTickTime = now
|
|
|
|
usage := getCurrentUsage()
|
|
usageHistory = append(usageHistory, usage)
|
|
|
|
// Update current status
|
|
updateSystemStatus(usage, usageHistory)
|
|
|
|
// Remove entries older than monitoring period
|
|
cutoff := time.Now().Add(-monitoringPeriod)
|
|
for i, u := range usageHistory {
|
|
if u.Timestamp.After(cutoff) {
|
|
usageHistory = usageHistory[i:]
|
|
break
|
|
}
|
|
}
|
|
|
|
if len(usageHistory) > 0 && isSystemIdle(usageHistory) {
|
|
log.Printf("System status before suspend:\n")
|
|
log.Printf("- CPU: %.1f%%\n", usageHistory[len(usageHistory)-1].CpuUsage)
|
|
if nvmlAvailable {
|
|
log.Printf("- GPU: %.1f%%\n", usageHistory[len(usageHistory)-1].GpuUsage)
|
|
}
|
|
log.Printf("- SSH connections: %d\n", usageHistory[len(usageHistory)-1].SshConnections)
|
|
log.Printf("- Active users: %d\n", usageHistory[len(usageHistory)-1].ActiveUsers)
|
|
log.Println("System has been idle for the monitoring period. Suspending...")
|
|
if err := suspendSystem(); err != nil {
|
|
log.Printf("Failed to suspend system: %v", err)
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Graceful shutdown of HTTP server
|
|
shutdownCtx, shutdownCancel := context.WithTimeout(context.Background(), 5*time.Second)
|
|
defer shutdownCancel()
|
|
|
|
if err := srv.Shutdown(shutdownCtx); err != nil {
|
|
log.Printf("HTTP server shutdown error: %v", err)
|
|
}
|
|
|
|
log.Println("Goodbye!")
|
|
}
|
|
|
|
// Function to check if we're within the resume grace period
|
|
func isInsideResumeGracePeriod() bool {
|
|
return !lastResumeTime.IsZero() && time.Since(lastResumeTime) < resumeGracePeriod
|
|
}
|
|
|
|
// Function to calculate time left in grace period
|
|
func timeLeftInGracePeriod() time.Duration {
|
|
if !isInsideResumeGracePeriod() {
|
|
return 0
|
|
}
|
|
return resumeGracePeriod - time.Since(lastResumeTime)
|
|
}
|
|
|
|
func startHTTPServer(ctx context.Context) *http.Server {
|
|
srv := &http.Server{
|
|
Addr: fmt.Sprintf(":%d", httpPort),
|
|
}
|
|
|
|
http.HandleFunc("/status", handleStatus)
|
|
|
|
go func() {
|
|
if err := srv.ListenAndServe(); err != http.ErrServerClosed {
|
|
log.Printf("HTTP server error: %v", err)
|
|
}
|
|
}()
|
|
|
|
return srv
|
|
}
|
|
|
|
func handleStatus(w http.ResponseWriter, r *http.Request) {
|
|
statusMutex.RLock()
|
|
defer statusMutex.RUnlock()
|
|
|
|
w.Header().Set("Content-Type", "application/json")
|
|
json.NewEncoder(w).Encode(currentStatus)
|
|
}
|
|
|
|
func updateSystemStatus(current ResourceUsage, history []ResourceUsage) {
|
|
statusMutex.Lock()
|
|
defer statusMutex.Unlock()
|
|
|
|
currentStatus.CurrentUsage = current
|
|
currentStatus.Blockers = []string{}
|
|
|
|
// Add grace period info to status
|
|
if isInsideResumeGracePeriod() {
|
|
timeLeft := timeLeftInGracePeriod()
|
|
currentStatus.InGracePeriod = true
|
|
currentStatus.GraceTimeLeft = timeLeft.Round(time.Second).String()
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("Resume grace period: %v remaining", timeLeft.Round(time.Second)))
|
|
} else {
|
|
currentStatus.InGracePeriod = false
|
|
currentStatus.GraceTimeLeft = ""
|
|
}
|
|
|
|
if len(history) >= 2 {
|
|
// Calculate rates using last two samples
|
|
duration := history[len(history)-1].Timestamp.Sub(history[len(history)-2].Timestamp).Seconds()
|
|
diskIORate := float64(history[len(history)-1].DiskIO-history[len(history)-2].DiskIO) / duration
|
|
netIORate := float64(history[len(history)-1].NetworkIO-history[len(history)-2].NetworkIO) / duration
|
|
|
|
if current.CpuUsage >= cpuThreshold {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("CPU usage too high: %.1f%% >= %.1f%%", current.CpuUsage, cpuThreshold))
|
|
}
|
|
if nvmlAvailable && current.GpuUsage >= gpuThreshold {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("GPU usage too high: %.1f%% >= %.1f%%", current.GpuUsage, gpuThreshold))
|
|
}
|
|
if diskIORate >= float64(diskThreshold) {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("Disk I/O too high: %.1f MB/s >= %.1f MB/s",
|
|
diskIORate/(1024*1024), float64(diskThreshold)/(1024*1024)))
|
|
}
|
|
if netIORate >= float64(networkThreshold) {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("Network I/O too high: %.1f MB/s >= %.1f MB/s",
|
|
netIORate/(1024*1024), float64(networkThreshold)/(1024*1024)))
|
|
}
|
|
}
|
|
|
|
if current.SshConnections > 0 {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("Active SSH connections: %d", current.SshConnections))
|
|
}
|
|
if current.ActiveUsers > 0 {
|
|
currentStatus.Blockers = append(currentStatus.Blockers,
|
|
fmt.Sprintf("Active user sessions: %d", current.ActiveUsers))
|
|
}
|
|
|
|
if len(currentStatus.Blockers) == 0 {
|
|
currentStatus.Blockers = append(currentStatus.Blockers, "No blockers - system can sleep")
|
|
}
|
|
}
|
|
|
|
func getCurrentUsage() ResourceUsage {
|
|
usage := ResourceUsage{
|
|
Timestamp: time.Now(),
|
|
GpuAvailable: nvmlAvailable,
|
|
}
|
|
|
|
// Get CPU usage across all cores
|
|
if cpuPercent, err := cpu.Percent(0, true); err == nil && len(cpuPercent) > 0 {
|
|
// Calculate average CPU usage across all cores
|
|
var totalCPU float64
|
|
for _, percent := range cpuPercent {
|
|
totalCPU += percent
|
|
}
|
|
usage.CpuUsage = totalCPU / float64(len(cpuPercent))
|
|
}
|
|
|
|
// Get GPU usage across all GPUs if available
|
|
if nvmlAvailable {
|
|
count, ret := nvml.DeviceGetCount()
|
|
if ret == nvml.SUCCESS && count > 0 {
|
|
var totalGPU float64
|
|
var activeGPUs int
|
|
for i := 0; i < count; i++ {
|
|
device, ret := nvml.DeviceGetHandleByIndex(i)
|
|
if ret == nvml.SUCCESS {
|
|
utilization, ret := device.GetUtilizationRates()
|
|
if ret == nvml.SUCCESS {
|
|
totalGPU += float64(utilization.Gpu)
|
|
activeGPUs++
|
|
}
|
|
}
|
|
}
|
|
if activeGPUs > 0 {
|
|
usage.GpuUsage = totalGPU / float64(activeGPUs)
|
|
}
|
|
}
|
|
}
|
|
|
|
// Get disk I/O
|
|
if diskStats, err := disk.IOCounters(); err == nil {
|
|
var totalIO uint64
|
|
for _, stat := range diskStats {
|
|
totalIO += stat.ReadBytes + stat.WriteBytes
|
|
}
|
|
usage.DiskIO = totalIO
|
|
}
|
|
|
|
// Get network I/O
|
|
if netStats, err := net.IOCounters(false); err == nil && len(netStats) > 0 {
|
|
usage.NetworkIO = netStats[0].BytesSent + netStats[0].BytesRecv
|
|
}
|
|
|
|
// Count SSH connections
|
|
if sshCount, err := getSSHConnectionCount(); err == nil {
|
|
usage.SshConnections = sshCount
|
|
}
|
|
|
|
// Count active user sessions
|
|
if userCount, err := getActiveUserCount(); err == nil {
|
|
usage.ActiveUsers = userCount
|
|
}
|
|
|
|
return usage
|
|
}
|
|
|
|
func getSSHConnectionCount() (int, error) {
|
|
processes, err := process.Processes()
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
|
|
count := 0
|
|
for _, p := range processes {
|
|
name, err := p.Name()
|
|
if err != nil {
|
|
continue
|
|
}
|
|
if name == "sshd" {
|
|
cmdline, err := p.Cmdline()
|
|
if err != nil {
|
|
continue
|
|
}
|
|
// Only count sshd processes that are handling connections
|
|
// The main sshd process doesn't have "@" in its cmdline
|
|
if strings.Contains(cmdline, "@") {
|
|
count++
|
|
}
|
|
}
|
|
}
|
|
return count, nil
|
|
}
|
|
|
|
func getActiveUserCount() (int, error) {
|
|
out, err := exec.Command("who", "-s").Output()
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
|
|
// Count non-empty lines
|
|
lines := strings.Split(strings.TrimSpace(string(out)), "\n")
|
|
count := 0
|
|
for _, line := range lines {
|
|
if line != "" {
|
|
count++
|
|
}
|
|
}
|
|
return count, nil
|
|
}
|
|
|
|
func isSystemIdle(history []ResourceUsage) bool {
|
|
// Don't allow sleep during grace period after resume
|
|
if isInsideResumeGracePeriod() {
|
|
return false
|
|
}
|
|
|
|
if len(history) < 2 {
|
|
return false
|
|
}
|
|
|
|
var avgCPU, avgGPU float64
|
|
var maxSSHConnections, maxActiveUsers int
|
|
samples := len(history)
|
|
|
|
for _, usage := range history {
|
|
avgCPU += usage.CpuUsage
|
|
if nvmlAvailable {
|
|
avgGPU += usage.GpuUsage
|
|
}
|
|
if usage.SshConnections > maxSSHConnections {
|
|
maxSSHConnections = usage.SshConnections
|
|
}
|
|
if usage.ActiveUsers > maxActiveUsers {
|
|
maxActiveUsers = usage.ActiveUsers
|
|
}
|
|
}
|
|
|
|
// Calculate I/O rates using first and last samples
|
|
duration := history[samples-1].Timestamp.Sub(history[0].Timestamp).Seconds()
|
|
diskIORate := float64(history[samples-1].DiskIO-history[0].DiskIO) / duration
|
|
netIORate := float64(history[samples-1].NetworkIO-history[0].NetworkIO) / duration
|
|
|
|
avgCPU /= float64(samples)
|
|
if nvmlAvailable {
|
|
avgGPU /= float64(samples)
|
|
}
|
|
|
|
// Basic checks that always apply
|
|
if avgCPU >= cpuThreshold ||
|
|
diskIORate >= float64(diskThreshold) ||
|
|
netIORate >= float64(networkThreshold) ||
|
|
maxSSHConnections > 0 ||
|
|
maxActiveUsers > 0 {
|
|
return false
|
|
}
|
|
|
|
// GPU check only if NVML is available
|
|
if nvmlAvailable && avgGPU >= gpuThreshold {
|
|
return false
|
|
}
|
|
|
|
return true
|
|
}
|
|
|
|
func suspendSystem() error {
|
|
cmd := exec.Command("systemctl", "suspend")
|
|
return cmd.Run()
|
|
}
|