add more cpu metrics (#1356)

- adds monitoring for cpu state time and per-core usage

Co-authored-by: Sven van Ginkel <svenvanginkel@icloud.com>

agent/battery/battery.go

@@ -56,7 +56,7 @@ func GetBatteryStats() (batteryPercent uint8, batteryState uint8, err error) {
 			// if there were some errors, like missing data, skip it
 			continue
 		}
-		if bat.Full == 0 {
+		if bat == nil || bat.Full == 0 {
 			// skip batteries with no capacity. Charge is unlikely to ever be zero, but
 			// we can't guarantee that, so don't skip based on charge.
 			continue

agent/cpu.go

@@ -4,10 +4,12 @@ import (
 	"math"
 	"runtime"
 
+	"github.com/henrygd/beszel/internal/entities/system"
 	"github.com/shirou/gopsutil/v4/cpu"
 )
 
 var lastCpuTimes = make(map[uint16]cpu.TimesStat)
+var lastPerCoreCpuTimes = make(map[uint16][]cpu.TimesStat)
 
 // init initializes the CPU monitoring by storing the initial CPU times
 // for the default 60-second cache interval.
@@ -15,23 +17,92 @@ func init() {
 	if times, err := cpu.Times(false); err == nil {
 		lastCpuTimes[60000] = times[0]
 	}
+	if perCoreTimes, err := cpu.Times(true); err == nil {
+		lastPerCoreCpuTimes[60000] = perCoreTimes
+	}
 }
 
-// getCpuPercent calculates the CPU usage percentage using cached previous measurements.
-// It uses the specified cache time interval to determine the time window for calculation.
-// Returns the CPU usage percentage (0-100) and any error encountered.
-func getCpuPercent(cacheTimeMs uint16) (float64, error) {
+// CpuMetrics contains detailed CPU usage breakdown
+type CpuMetrics struct {
+	Total  float64
+	User   float64
+	System float64
+	Iowait float64
+	Steal  float64
+	Idle   float64
+}
+
+// getCpuMetrics calculates detailed CPU usage metrics using cached previous measurements.
+// It returns percentages for total, user, system, iowait, and steal time.
+func getCpuMetrics(cacheTimeMs uint16) (CpuMetrics, error) {
 	times, err := cpu.Times(false)
 	if err != nil || len(times) == 0 {
-		return 0, err
+		return CpuMetrics{}, err
 	}
 	// if cacheTimeMs is not in lastCpuTimes, use 60000 as fallback lastCpuTime
 	if _, ok := lastCpuTimes[cacheTimeMs]; !ok {
 		lastCpuTimes[cacheTimeMs] = lastCpuTimes[60000]
 	}
-	delta := calculateBusy(lastCpuTimes[cacheTimeMs], times[0])
+
+	t1 := lastCpuTimes[cacheTimeMs]
+	t2 := times[0]
+
+	t1All, _ := getAllBusy(t1)
+	t2All, _ := getAllBusy(t2)
+
+	totalDelta := t2All - t1All
+	if totalDelta <= 0 {
+		return CpuMetrics{}, nil
+	}
+
+	metrics := CpuMetrics{
+		Total:  calculateBusy(t1, t2),
+		User:   clampPercent((t2.User - t1.User) / totalDelta * 100),
+		System: clampPercent((t2.System - t1.System) / totalDelta * 100),
+		Iowait: clampPercent((t2.Iowait - t1.Iowait) / totalDelta * 100),
+		Steal:  clampPercent((t2.Steal - t1.Steal) / totalDelta * 100),
+		Idle:   clampPercent((t2.Idle - t1.Idle) / totalDelta * 100),
+	}
+
 	lastCpuTimes[cacheTimeMs] = times[0]
-	return delta, nil
+	return metrics, nil
+}
+
+// clampPercent ensures the percentage is between 0 and 100
+func clampPercent(value float64) float64 {
+	return math.Min(100, math.Max(0, value))
+}
+
+// getPerCoreCpuUsage calculates per-core CPU busy usage as integer percentages (0-100).
+// It uses cached previous measurements for the provided cache interval.
+func getPerCoreCpuUsage(cacheTimeMs uint16) (system.Uint8Slice, error) {
+	perCoreTimes, err := cpu.Times(true)
+	if err != nil || len(perCoreTimes) == 0 {
+		return nil, err
+	}
+
+	// Initialize cache if needed
+	if _, ok := lastPerCoreCpuTimes[cacheTimeMs]; !ok {
+		lastPerCoreCpuTimes[cacheTimeMs] = lastPerCoreCpuTimes[60000]
+	}
+
+	lastTimes := lastPerCoreCpuTimes[cacheTimeMs]
+
+	// Limit to the number of cores available in both samples
+	length := len(perCoreTimes)
+	if len(lastTimes) < length {
+		length = len(lastTimes)
+	}
+
+	usage := make([]uint8, length)
+	for i := 0; i < length; i++ {
+		t1 := lastTimes[i]
+		t2 := perCoreTimes[i]
+		usage[i] = uint8(math.Round(calculateBusy(t1, t2)))
+	}
+
+	lastPerCoreCpuTimes[cacheTimeMs] = perCoreTimes
+	return usage, nil
 }
 
 // calculateBusy calculates the CPU busy percentage between two time points.
@@ -41,13 +112,10 @@ func calculateBusy(t1, t2 cpu.TimesStat) float64 {
 	t1All, t1Busy := getAllBusy(t1)
 	t2All, t2Busy := getAllBusy(t2)
 
-	if t2Busy <= t1Busy {
+	if t2All <= t1All || t2Busy <= t1Busy {
 		return 0
 	}
-	if t2All <= t1All {
-		return 100
-	}
-	return math.Min(100, math.Max(0, (t2Busy-t1Busy)/(t2All-t1All)*100))
+	return clampPercent((t2Busy - t1Busy) / (t2All - t1All) * 100)
 }
 
 // getAllBusy calculates the total CPU time and busy CPU time from CPU times statistics.

agent/system.go

@@ -83,12 +83,24 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
 		systemStats.Battery[1] = batteryState
 	}
 
-	// cpu percent
-	cpuPercent, err := getCpuPercent(cacheTimeMs)
+	// cpu metrics
+	cpuMetrics, err := getCpuMetrics(cacheTimeMs)
 	if err == nil {
-		systemStats.Cpu = twoDecimals(cpuPercent)
+		systemStats.Cpu = twoDecimals(cpuMetrics.Total)
+		systemStats.CpuBreakdown = []float64{
+			twoDecimals(cpuMetrics.User),
+			twoDecimals(cpuMetrics.System),
+			twoDecimals(cpuMetrics.Iowait),
+			twoDecimals(cpuMetrics.Steal),
+			twoDecimals(cpuMetrics.Idle),
+		}
 	} else {
-		slog.Error("Error getting cpu percent", "err", err)
+		slog.Error("Error getting cpu metrics", "err", err)
+	}
+
+	// per-core cpu usage
+	if perCoreUsage, err := getPerCoreCpuUsage(cacheTimeMs); err == nil {
+		systemStats.CpuCoresUsage = perCoreUsage
 	}
 
 	// load average