collect top process

agent/cpu.go

@@ -2,14 +2,19 @@ package agent
 
 import (
 	"math"
+	"path/filepath"
 	"runtime"
+	"time"
 
 	"github.com/henrygd/beszel/internal/entities/system"
 	"github.com/shirou/gopsutil/v4/cpu"
+	"github.com/shirou/gopsutil/v4/process"
 )
 
 var lastCpuTimes = make(map[uint16]cpu.TimesStat)
 var lastPerCoreCpuTimes = make(map[uint16][]cpu.TimesStat)
+var lastProcessCpuTimes = make(map[uint16]map[int32]float64)
+var lastProcessCpuSampleTime = make(map[uint16]time.Time)
 
 // init initializes the CPU monitoring by storing the initial CPU times
 // for the default 60-second cache interval.
@@ -20,6 +25,16 @@ func init() {
 	if perCoreTimes, err := cpu.Times(true); err == nil {
 		lastPerCoreCpuTimes[60000] = perCoreTimes
 	}
+	if processes, err := process.Processes(); err == nil {
+		snapshot := make(map[int32]float64, len(processes))
+		for _, proc := range processes {
+			if times, err := proc.Times(); err == nil {
+				snapshot[proc.Pid] = times.Total()
+			}
+		}
+		lastProcessCpuTimes[60000] = snapshot
+		lastProcessCpuSampleTime[60000] = time.Now()
+	}
 }
 
 // CpuMetrics contains detailed CPU usage breakdown
@@ -105,6 +120,110 @@ func getPerCoreCpuUsage(cacheTimeMs uint16) (system.Uint8Slice, error) {
 	return usage, nil
 }
 
+// getTopCpuProcess returns the process with the highest CPU usage since the last run
+// for the given cache interval. It returns nil if insufficient data is available.
+func getTopCpuProcess(cacheTimeMs uint16) (*system.TopCpuProcess, error) {
+	processes, err := process.Processes()
+	if err != nil {
+		return nil, err
+	}
+
+	now := time.Now()
+
+	lastTimes, ok := lastProcessCpuTimes[cacheTimeMs]
+	if !ok {
+		if fallback := lastProcessCpuTimes[60000]; fallback != nil {
+			copied := make(map[int32]float64, len(fallback))
+			for pid, total := range fallback {
+				copied[pid] = total
+			}
+			lastTimes = copied
+			lastProcessCpuTimes[cacheTimeMs] = copied
+		} else {
+			lastTimes = make(map[int32]float64)
+			lastProcessCpuTimes[cacheTimeMs] = lastTimes
+		}
+	}
+
+	lastSample := lastProcessCpuSampleTime[cacheTimeMs]
+	if lastSample.IsZero() {
+		if fallback := lastProcessCpuSampleTime[60000]; !fallback.IsZero() {
+			lastSample = fallback
+			lastProcessCpuSampleTime[cacheTimeMs] = fallback
+		}
+	}
+
+	elapsed := now.Sub(lastSample).Seconds()
+	if lastSample.IsZero() || elapsed <= 0 {
+		snapshot := make(map[int32]float64, len(processes))
+		for _, proc := range processes {
+			if times, err := proc.Times(); err == nil {
+				snapshot[proc.Pid] = times.Total()
+			}
+		}
+		lastProcessCpuTimes[cacheTimeMs] = snapshot
+		lastProcessCpuSampleTime[cacheTimeMs] = now
+		return nil, nil
+	}
+
+	cpuCount := float64(runtime.NumCPU())
+	if cpuCount <= 0 {
+		cpuCount = 1
+	}
+
+	snapshot := make(map[int32]float64, len(processes))
+	var topName string
+	var topPercent float64
+
+	for _, proc := range processes {
+		times, err := proc.Times()
+		if err != nil {
+			continue
+		}
+
+		total := times.Total()
+		pid := proc.Pid
+		snapshot[pid] = total
+
+		lastTotal, ok := lastTimes[pid]
+		if !ok || total <= lastTotal {
+			continue
+		}
+
+		percent := clampPercent((total - lastTotal) / (elapsed * cpuCount) * 100)
+		if percent <= 0 {
+			continue
+		}
+
+		name, err := proc.Name()
+		if err != nil || name == "" {
+			if exe, exeErr := proc.Exe(); exeErr == nil && exe != "" {
+				name = filepath.Base(exe)
+			}
+		}
+		if name == "" {
+			continue
+		}
+
+		if percent > topPercent {
+			topPercent = percent
+			topName = name
+		}
+	}
+
+	lastProcessCpuTimes[cacheTimeMs] = snapshot
+	lastProcessCpuSampleTime[cacheTimeMs] = now
+
+	if topName == "" {
+		return nil, nil
+	}
+
+	return &system.TopCpuProcess{
+		Name:    topName,
+		Percent: topPercent,
+	}, nil
+}
+
 // calculateBusy calculates the CPU busy percentage between two time points.
 // It computes the ratio of busy time to total time elapsed between t1 and t2,
 // returning a percentage clamped between 0 and 100.

agent/disk.go

@@ -31,6 +31,7 @@ func (a *Agent) initializeDiskInfo() {
 	filesystem, _ := GetEnv("FILESYSTEM")
 	efPath := "/extra-filesystems"
 	hasRoot := false
+	isWindows := runtime.GOOS == "windows"
 
 	partitions, err := disk.Partitions(false)
 	if err != nil {
@@ -38,6 +39,13 @@ func (a *Agent) initializeDiskInfo() {
 	}
 	slog.Debug("Disk", "partitions", partitions)
 
+	// trim trailing backslash for Windows devices (#1361)
+	if isWindows {
+		for i, p := range partitions {
+			partitions[i].Device = strings.TrimSuffix(p.Device, "\\")
+		}
+	}
+
 	// ioContext := context.WithValue(a.sensorsContext,
 	// 	common.EnvKey, common.EnvMap{common.HostProcEnvKey: "/tmp/testproc"},
 	// )
@@ -52,7 +60,7 @@ func (a *Agent) initializeDiskInfo() {
 	// Helper function to add a filesystem to fsStats if it doesn't exist
 	addFsStat := func(device, mountpoint string, root bool, customName ...string) {
 		var key string
-		if runtime.GOOS == "windows" {
+		if isWindows {
 			key = device
 		} else {
 			key = filepath.Base(device)

agent/system.go

@@ -98,6 +98,15 @@ func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
 		slog.Error("Error getting cpu metrics", "err", err)
 	}
 
+	if topProcess, err := getTopCpuProcess(cacheTimeMs); err == nil {
+		if topProcess != nil {
+			topProcess.Percent = twoDecimals(topProcess.Percent)
+			systemStats.TopCpuProcess = topProcess
+		}
+	} else {
+		slog.Error("Error getting top cpu process", "err", err)
+	}
+
 	// per-core cpu usage
 	if perCoreUsage, err := getPerCoreCpuUsage(cacheTimeMs); err == nil {
 		systemStats.CpuCoresUsage = perCoreUsage

internal/dockerfile_agent_nvidia

@@ -2,7 +2,6 @@ FROM --platform=$BUILDPLATFORM golang:alpine AS builder
 
 WORKDIR /app
 
-
 COPY ../go.mod ../go.sum ./
 RUN go mod download
 
@@ -13,7 +12,24 @@ COPY . ./
 ARG TARGETOS TARGETARCH
 RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./internal/cmd/agent
 
-RUN rm -rf /tmp/*
+# --------------------------
+# Smartmontools builder stage
+# --------------------------
+FROM nvidia/cuda:12.2.2-base-ubuntu22.04 AS smartmontools-builder
+
+RUN apt-get update && apt-get install -y \
+  wget \
+  build-essential \
+  && wget https://downloads.sourceforge.net/project/smartmontools/smartmontools/7.5/smartmontools-7.5.tar.gz \
+  && tar zxvf smartmontools-7.5.tar.gz \
+  && cd smartmontools-7.5 \
+  && ./configure --prefix=/usr --sysconfdir=/etc \
+  && make \
+  && make install \
+  && rm -rf /smartmontools-7.5* \
+  && apt-get remove -y wget build-essential \
+  && apt-get autoremove -y \
+  && rm -rf /var/lib/apt/lists/*
 
 # --------------------------
 # Final image: GPU-enabled agent with nvidia-smi
@@ -21,10 +37,8 @@ RUN rm -rf /tmp/*
 FROM nvidia/cuda:12.2.2-base-ubuntu22.04
 COPY --from=builder /agent /agent
 
-# this is so we don't need to create the /tmp directory in the scratch container
-COPY --from=builder /tmp /tmp
-
-RUN apt-get update && apt-get install -y smartmontools && rm -rf /var/lib/apt/lists/*
+# Copy smartmontools binaries and config files
+COPY --from=smartmontools-builder /usr/sbin/smartctl /usr/sbin/smartctl
 
 # Ensure data persistence across container recreations
 VOLUME ["/var/lib/beszel-agent"]

internal/entities/system/system.go

@@ -47,6 +47,7 @@ type Stats struct {
 	MaxDiskIO         [2]uint64            `json:"diom,omitzero" cbor:"-"`                      // [max read bytes, max write bytes]
 	CpuBreakdown      []float64            `json:"cpub,omitempty" cbor:"33,keyasint,omitempty"` // [user, system, iowait, steal, idle]
 	CpuCoresUsage     Uint8Slice           `json:"cpus,omitempty" cbor:"34,keyasint,omitempty"` // per-core busy usage [CPU0..]
+	TopCpuProcess     *TopCpuProcess       `json:"tcp,omitempty" cbor:"35,keyasint,omitempty"`
 }
 
 // Uint8Slice wraps []uint8 to customize JSON encoding while keeping CBOR efficient.
@@ -153,3 +154,8 @@ type CombinedData struct {
 	Info       Info               `json:"info" cbor:"1,keyasint"`
 	Containers []*container.Stats `json:"container" cbor:"2,keyasint"`
 }
+
+type TopCpuProcess struct {
+	Name    string  `json:"n" cbor:"0,keyasint"`
+	Percent float64 `json:"p" cbor:"1,keyasint"`
+}