Probable fix for Jetson gpu issue (#895)

adding the fallback to the GITHUB_TOKEN allows execution of goreleaser in a fork without additional configuration

.github/workflows/release.yml

@@ -39,4 +39,4 @@ jobs:
           version: latest
           args: release --clean
         env:
-          GITHUB_TOKEN: ${{ secrets.TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.TOKEN || secrets.GITHUB_TOKEN }}

beszel/internal/agent/gpu.go

@@ -243,21 +243,26 @@ func (gm *GPUManager) GetCurrentData() map[string]system.GPUData {
 	// copy / reset the data
 	gpuData := make(map[string]system.GPUData, len(gm.GpuDataMap))
 	for id, gpu := range gm.GpuDataMap {
-		// sum the data
-		gpu.Temperature = twoDecimals(gpu.Temperature)
-		gpu.MemoryUsed = twoDecimals(gpu.MemoryUsed)
-		gpu.MemoryTotal = twoDecimals(gpu.MemoryTotal)
-		gpu.Usage = twoDecimals(gpu.Usage / gpu.Count)
-		gpu.Power = twoDecimals(gpu.Power / gpu.Count)
-		// reset the count
-		gpu.Count = 1
-		// dereference to avoid overwriting anything else
-		gpuCopy := *gpu
+		var gpuAvg system.GPUData
+
+		gpuAvg.Temperature = twoDecimals(gpu.Temperature)
+		gpuAvg.MemoryUsed = twoDecimals(gpu.MemoryUsed)
+		gpuAvg.MemoryTotal = twoDecimals(gpu.MemoryTotal)
+
+		// avoid division by zero
+		if gpu.Count > 0 {
+			gpuAvg.Usage = twoDecimals(gpu.Usage / gpu.Count)
+			gpuAvg.Power = twoDecimals(gpu.Power / gpu.Count)
+		}
+
+		// reset accumulators in the original
+		gpu.Usage, gpu.Power, gpu.Count = 0, 0, 0
+
 		// append id to the name if there are multiple GPUs with the same name
 		if nameCounts[gpu.Name] > 1 {
-			gpuCopy.Name = fmt.Sprintf("%s %s", gpu.Name, id)
+			gpuAvg.Name = fmt.Sprintf("%s %s", gpu.Name, id)
 		}
-		gpuData[id] = gpuCopy
+		gpuData[id] = gpuAvg
 	}
 	slog.Debug("GPU", "data", gpuData)
 	return gpuData

beszel/internal/agent/gpu_test.go

@@ -279,6 +279,19 @@ func TestParseJetsonData(t *testing.T) {
 				Count:       1,
 			},
 		},
+		{
+			name:  "orin nano",
+			input: "06-18-2025 11:25:24 RAM 3452/7620MB (lfb 25x4MB) SWAP 1518/16384MB (cached 174MB) CPU [1%@1420,2%@1420,0%@1420,2%@1420,2%@729,1%@729] GR3D_FREQ 0% cpu@50.031C soc2@49.031C soc0@50C gpu@49.031C tj@50.25C soc1@50.25C VDD_IN 4824mW/4824mW VDD_CPU_GPU_CV 518mW/518mW VDD_SOC 1475mW/1475mW",
+			wantMetrics: &system.GPUData{
+				Name:        "GPU",
+				MemoryUsed:  3452.0,
+				MemoryTotal: 7620.0,
+				Usage:       0.0,
+				Temperature: 50.25,
+				Power:       0.518,
+				Count:       1,
+			},
+		},
 		{
 			name:  "missing temperature",
 			input: "11-14-2024 22:54:33 RAM 4300/30698MB GR3D_FREQ 45% VDD_GPU_SOC 2171mW",
@@ -318,44 +331,75 @@ func TestParseJetsonData(t *testing.T) {
 }
 
 func TestGetCurrentData(t *testing.T) {
-	gm := &GPUManager{
-		GpuDataMap: map[string]*system.GPUData{
-			"0": {
-				Name:        "GPU1",
-				Temperature: 50,
-				MemoryUsed:  2048,
-				MemoryTotal: 4096,
-				Usage:       100, // 100 over 2 counts = 50 avg
-				Power:       200, // 200 over 2 counts = 100 avg
-				Count:       2,
+	t.Run("calculates averages and resets accumulators", func(t *testing.T) {
+		gm := &GPUManager{
+			GpuDataMap: map[string]*system.GPUData{
+				"0": {
+					Name:        "GPU1",
+					Temperature: 50,
+					MemoryUsed:  2048,
+					MemoryTotal: 4096,
+					Usage:       100, // 100 over 2 counts = 50 avg
+					Power:       200, // 200 over 2 counts = 100 avg
+					Count:       2,
+				},
+				"1": {
+					Name:        "GPU1",
+					Temperature: 60,
+					MemoryUsed:  3072,
+					MemoryTotal: 8192,
+					Usage:       30,
+					Power:       60,
+					Count:       1,
+				},
 			},
-			"1": {
-				Name:        "GPU1",
-				Temperature: 60,
-				MemoryUsed:  3072,
-				MemoryTotal: 8192,
-				Usage:       30,
-				Power:       60,
-				Count:       1,
+		}
+
+		result := gm.GetCurrentData()
+
+		// Verify name disambiguation
+		assert.Equal(t, "GPU1 0", result["0"].Name)
+		assert.Equal(t, "GPU1 1", result["1"].Name)
+
+		// Check averaged values in the result
+		assert.InDelta(t, 50.0, result["0"].Usage, 0.01)
+		assert.InDelta(t, 100.0, result["0"].Power, 0.01)
+		assert.InDelta(t, 30.0, result["1"].Usage, 0.01)
+		assert.InDelta(t, 60.0, result["1"].Power, 0.01)
+
+		// Verify that accumulators in the original map are reset
+		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Count, "GPU 0 Count should be reset")
+		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Usage, "GPU 0 Usage should be reset")
+		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Power, "GPU 0 Power should be reset")
+		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Count, "GPU 1 Count should be reset")
+		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Usage, "GPU 1 Usage should be reset")
+		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Power, "GPU 1 Power should be reset")
+	})
+
+	t.Run("handles zero count without panicking", func(t *testing.T) {
+		gm := &GPUManager{
+			GpuDataMap: map[string]*system.GPUData{
+				"0": {
+					Name:  "TestGPU",
+					Count: 0,
+					Usage: 0,
+					Power: 0,
+				},
 			},
-		},
-	}
+		}
 
-	result := gm.GetCurrentData()
+		var result map[string]system.GPUData
+		assert.NotPanics(t, func() {
+			result = gm.GetCurrentData()
+		})
 
-	// Verify name disambiguation
-	assert.Equal(t, "GPU1 0", result["0"].Name)
-	assert.Equal(t, "GPU1 1", result["1"].Name)
+		// Check that usage and power are 0
+		assert.Equal(t, 0.0, result["0"].Usage)
+		assert.Equal(t, 0.0, result["0"].Power)
 
-	// Check averaged values
-	assert.InDelta(t, 50.0, result["0"].Usage, 0.01)
-	assert.InDelta(t, 100.0, result["0"].Power, 0.01)
-	assert.InDelta(t, 30.0, result["1"].Usage, 0.01)
-	assert.InDelta(t, 60.0, result["1"].Power, 0.01)
-
-	// Verify reset counts
-	assert.Equal(t, float64(1), gm.GpuDataMap["0"].Count)
-	assert.Equal(t, float64(1), gm.GpuDataMap["1"].Count)
+		// Verify reset count
+		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Count)
+	})
 }
 
 func TestDetectGPUs(t *testing.T) {
@@ -722,6 +766,18 @@ func TestAccumulation(t *testing.T) {
 				assert.InDelta(t, expected.avgUsage, gpu.Usage, 0.01, "Average usage in GetCurrentData should match")
 				assert.InDelta(t, expected.avgPower, gpu.Power, 0.01, "Average power in GetCurrentData should match")
 			}
+
+			// Verify that accumulators in the original map are reset
+			for id := range tt.expectedValues {
+				gpu, exists := gm.GpuDataMap[id]
+				assert.True(t, exists, "GPU with ID %s should still exist after GetCurrentData", id)
+				if !exists {
+					continue
+				}
+				assert.Equal(t, float64(0), gpu.Count, "Count should be reset for GPU ID %s", id)
+				assert.Equal(t, float64(0), gpu.Usage, "Usage should be reset for GPU ID %s", id)
+				assert.Equal(t, float64(0), gpu.Power, "Power should be reset for GPU ID %s", id)
+			}
 		})
 	}
 }

beszel/internal/agent/sensors.go

@@ -3,6 +3,7 @@ package agent
 import (
 	"beszel/internal/entities/system"
 	"context"
+	"fmt"
 	"log/slog"
 	"path"
 	"strconv"
@@ -30,6 +31,9 @@ func (a *Agent) newSensorConfig() *SensorConfig {
 	return a.newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal, skipCollection)
 }
 
+// Matches sensors.TemperaturesWithContext to allow for panic recovery (gopsutil/issues/1832)
+type getTempsFn func(ctx context.Context) ([]sensors.TemperatureStat, error)
+
 // newSensorConfigWithEnv creates a SensorConfig with the provided environment variables
 // sensorsSet indicates if the SENSORS environment variable was explicitly set (even to empty string)
 func (a *Agent) newSensorConfigWithEnv(primarySensor, sysSensors, sensorsEnvVal string, skipCollection bool) *SensorConfig {
@@ -78,8 +82,18 @@ func (a *Agent) updateTemperatures(systemStats *system.Stats) {
 	// reset high temp
 	a.systemInfo.DashboardTemp = 0
 
-	// get sensor data
-	temps, _ := sensors.TemperaturesWithContext(a.sensorConfig.context)
+	temps, err := a.getTempsWithPanicRecovery(sensors.TemperaturesWithContext)
+	if err != nil {
+		// retry once on panic (gopsutil/issues/1832)
+		temps, err = a.getTempsWithPanicRecovery(sensors.TemperaturesWithContext)
+		if err != nil {
+			slog.Warn("Error updating temperatures", "err", err)
+			if len(systemStats.Temperatures) > 0 {
+				systemStats.Temperatures = make(map[string]float64)
+			}
+			return
+		}
+	}
 	slog.Debug("Temperature", "sensors", temps)
 
 	// return if no sensors
@@ -107,15 +121,28 @@ func (a *Agent) updateTemperatures(systemStats *system.Stats) {
 			continue
 		}
 		// set dashboard temperature
-		if a.sensorConfig.primarySensor == "" {
+		switch a.sensorConfig.primarySensor {
+		case "":
 			a.systemInfo.DashboardTemp = max(a.systemInfo.DashboardTemp, sensor.Temperature)
-		} else if a.sensorConfig.primarySensor == sensorName {
+		case sensorName:
 			a.systemInfo.DashboardTemp = sensor.Temperature
 		}
 		systemStats.Temperatures[sensorName] = twoDecimals(sensor.Temperature)
 	}
 }
 
+// getTempsWithPanicRecovery wraps sensors.TemperaturesWithContext to recover from panics (gopsutil/issues/1832)
+func (a *Agent) getTempsWithPanicRecovery(getTemps getTempsFn) (temps []sensors.TemperatureStat, err error) {
+	defer func() {
+		if r := recover(); r != nil {
+			err = fmt.Errorf("panic: %v", r)
+		}
+	}()
+	// get sensor data (error ignored intentionally as it may be only with one sensor)
+	temps, _ = getTemps(a.sensorConfig.context)
+	return
+}
+
 // isValidSensor checks if a sensor is valid based on the sensor name and the sensor config
 func isValidSensor(sensorName string, config *SensorConfig) bool {
 	// if no sensors configured, everything is valid

beszel/internal/agent/sensors_test.go

@@ -4,11 +4,14 @@
 package agent
 
 import (
+	"beszel/internal/entities/system"
 	"context"
+	"fmt"
 	"os"
 	"testing"
 
 	"github.com/shirou/gopsutil/v4/common"
+	"github.com/shirou/gopsutil/v4/sensors"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 )
@@ -454,3 +457,97 @@ func TestScaleTemperatureLogic(t *testing.T) {
 			result, expected)
 	})
 }
+
+func TestGetTempsWithPanicRecovery(t *testing.T) {
+	agent := &Agent{
+		systemInfo: system.Info{},
+		sensorConfig: &SensorConfig{
+			context: context.Background(),
+		},
+	}
+
+	tests := []struct {
+		name        string
+		getTempsFn  getTempsFn
+		expectError bool
+		errorMsg    string
+	}{
+		{
+			name: "successful_function_call",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				return []sensors.TemperatureStat{
+					{SensorKey: "test_sensor", Temperature: 45.0},
+				}, nil
+			},
+			expectError: false,
+		},
+		{
+			name: "function_returns_error",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				return []sensors.TemperatureStat{
+					{SensorKey: "test_sensor", Temperature: 45.0},
+				}, fmt.Errorf("sensor error")
+			},
+			expectError: false, // getTempsWithPanicRecovery ignores errors from the function
+		},
+		{
+			name: "function_panics_with_string",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				panic("test panic")
+			},
+			expectError: true,
+			errorMsg:    "panic: test panic",
+		},
+		{
+			name: "function_panics_with_error",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				panic(fmt.Errorf("panic error"))
+			},
+			expectError: true,
+			errorMsg:    "panic:",
+		},
+		{
+			name: "function_panics_with_index_out_of_bounds",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				slice := []int{1, 2, 3}
+				_ = slice[10] // out of bounds panic
+				return nil, nil
+			},
+			expectError: true,
+			errorMsg:    "panic:",
+		},
+		{
+			name: "function_panics_with_any_conversion",
+			getTempsFn: func(ctx context.Context) ([]sensors.TemperatureStat, error) {
+				var i any = "string"
+				_ = i.(int) // type assertion panic
+				return nil, nil
+			},
+			expectError: true,
+			errorMsg:    "panic:",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			var temps []sensors.TemperatureStat
+			var err error
+
+			// The function should not panic, regardless of what the injected function does
+			assert.NotPanics(t, func() {
+				temps, err = agent.getTempsWithPanicRecovery(tt.getTempsFn)
+			}, "getTempsWithPanicRecovery should not panic")
+
+			if tt.expectError {
+				assert.Error(t, err, "Expected an error to be returned")
+				if tt.errorMsg != "" {
+					assert.Contains(t, err.Error(), tt.errorMsg,
+						"Error message should contain expected text")
+				}
+				assert.Nil(t, temps, "Temps should be nil when panic occurs")
+			} else {
+				assert.NoError(t, err, "Should not return error for successful calls")
+			}
+		})
+	}
+}

supplemental/scripts/install-agent.ps1

@@ -182,7 +182,7 @@ function Install-BeszelAgentWithScoop {
     scoop bucket add beszel https://github.com/henrygd/beszel-scoops | Out-Null
     
     Write-Host "Installing / updating beszel-agent..."
-    scoop install beszel-agent
+    scoop install beszel-agent | Out-Null
     
     if (-not (Test-CommandExists "beszel-agent")) {
         throw "Failed to install beszel-agent"

supplemental/scripts/install-agent.sh

@@ -316,18 +316,27 @@ fi
 # Create a dedicated user for the service if it doesn't exist
 if is_alpine; then
   if ! id -u beszel >/dev/null 2>&1; then
+    echo "Creating a dedicated group for the Beszel Agent service..."
+    addgroup beszel
     echo "Creating a dedicated user for the Beszel Agent service..."
-    adduser -S -D -H -s /sbin/nologin beszel
+    adduser -S -D -H -s /sbin/nologin -G beszel beszel
   fi
-  # Add the user to the docker group to allow access to the Docker socket
-  addgroup beszel docker
+  # Add the user to the docker group to allow access to the Docker socket if group docker exists
+  if getent group docker; then
+	  echo "Adding besel to docker group"
+	  usermod -aG docker beszel
+  fi
+
 else
   if ! id -u beszel >/dev/null 2>&1; then
     echo "Creating a dedicated user for the Beszel Agent service..."
     useradd --system --home-dir /nonexistent --shell /bin/false beszel
   fi
-  # Add the user to the docker group to allow access to the Docker socket
-  usermod -aG docker beszel
+# Add the user to the docker group to allow access to the Docker socket if group docker exists
+  if getent group docker; then
+	  echo "Adding besel to docker group"
+	  usermod -aG docker beszel
+  fi
 fi
 
 # Create the directory for the Beszel Agent