add basic systemd service monitoring (#1153)

Co-authored-by: Shelby Tucker <shelby.tucker@gmail.com>

internal/entities/system/system.go

@@ -7,6 +7,7 @@ import (
 	"time"
 
 	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/systemd"
 )
 
 type Stats struct {
@@ -149,7 +150,8 @@ type Info struct {
 
 // Final data structure to return to the hub
 type CombinedData struct {
-	Stats      Stats              `json:"stats" cbor:"0,keyasint"`
-	Info       Info               `json:"info" cbor:"1,keyasint"`
-	Containers []*container.Stats `json:"container" cbor:"2,keyasint"`
+	Stats           Stats              `json:"stats" cbor:"0,keyasint"`
+	Info            Info               `json:"info" cbor:"1,keyasint"`
+	Containers      []*container.Stats `json:"container" cbor:"2,keyasint"`
+	SystemdServices []*systemd.Service `json:"systemd,omitempty" cbor:"3,keyasint,omitempty"`
 }

internal/entities/systemd/systemd.go

@@ -0,0 +1,127 @@
+package systemd
+
+import (
+	"math"
+	"runtime"
+	"time"
+)
+
+// ServiceState represents the status of a systemd service
+type ServiceState uint8
+
+const (
+	StatusActive ServiceState = iota
+	StatusInactive
+	StatusFailed
+	StatusActivating
+	StatusDeactivating
+	StatusReloading
+)
+
+// ServiceSubState represents the sub status of a systemd service
+type ServiceSubState uint8
+
+const (
+	SubStateDead ServiceSubState = iota
+	SubStateRunning
+	SubStateExited
+	SubStateFailed
+	SubStateUnknown
+)
+
+// ParseServiceStatus converts a string status to a ServiceStatus enum value
+func ParseServiceStatus(status string) ServiceState {
+	switch status {
+	case "active":
+		return StatusActive
+	case "inactive":
+		return StatusInactive
+	case "failed":
+		return StatusFailed
+	case "activating":
+		return StatusActivating
+	case "deactivating":
+		return StatusDeactivating
+	case "reloading":
+		return StatusReloading
+	default:
+		return StatusInactive
+	}
+}
+
+// ParseServiceSubState converts a string sub status to a ServiceSubState enum value
+func ParseServiceSubState(subState string) ServiceSubState {
+	switch subState {
+	case "dead":
+		return SubStateDead
+	case "running":
+		return SubStateRunning
+	case "exited":
+		return SubStateExited
+	case "failed":
+		return SubStateFailed
+	default:
+		return SubStateUnknown
+	}
+}
+
+// Service represents a single systemd service with its stats.
+type Service struct {
+	Name         string          `json:"n" cbor:"0,keyasint"`
+	State        ServiceState    `json:"s" cbor:"1,keyasint"`
+	Cpu          float64         `json:"c" cbor:"2,keyasint"`
+	Mem          uint64          `json:"m" cbor:"3,keyasint"`
+	MemPeak      uint64          `json:"mp" cbor:"4,keyasint"`
+	Sub          ServiceSubState `json:"ss" cbor:"5,keyasint"`
+	CpuPeak      float64         `json:"cp" cbor:"6,keyasint"`
+	PrevCpuUsage uint64          `json:"-"`
+	PrevReadTime time.Time       `json:"-"`
+}
+
+// UpdateCPUPercent calculates the CPU usage percentage for the service.
+func (s *Service) UpdateCPUPercent(cpuUsage uint64) {
+	now := time.Now()
+
+	if s.PrevReadTime.IsZero() || cpuUsage < s.PrevCpuUsage {
+		s.Cpu = 0
+		s.PrevCpuUsage = cpuUsage
+		s.PrevReadTime = now
+		return
+	}
+
+	duration := now.Sub(s.PrevReadTime).Nanoseconds()
+	if duration <= 0 {
+		s.PrevCpuUsage = cpuUsage
+		s.PrevReadTime = now
+		return
+	}
+
+	coreCount := int64(runtime.NumCPU())
+	duration *= coreCount
+
+	usageDelta := cpuUsage - s.PrevCpuUsage
+	cpuPercent := float64(usageDelta) / float64(duration)
+	s.Cpu = twoDecimals(cpuPercent * 100)
+
+	if s.Cpu > s.CpuPeak {
+		s.CpuPeak = s.Cpu
+	}
+
+	s.PrevCpuUsage = cpuUsage
+	s.PrevReadTime = now
+}
+
+func twoDecimals(value float64) float64 {
+	return math.Round(value*100) / 100
+}
+
+// ServiceDependency represents a unit that the service depends on.
+type ServiceDependency struct {
+	Name        string `json:"name"`
+	Description string `json:"description,omitempty"`
+	ActiveState string `json:"activeState,omitempty"`
+	SubState    string `json:"subState,omitempty"`
+}
+
+// ServiceDetails contains extended information about a systemd service.
+type ServiceDetails map[string]any

internal/entities/systemd/systemd_test.go

@@ -0,0 +1,113 @@
+//go:build testing
+
+package systemd_test
+
+import (
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/systemd"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestParseServiceStatus(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected systemd.ServiceState
+	}{
+		{"active", systemd.StatusActive},
+		{"inactive", systemd.StatusInactive},
+		{"failed", systemd.StatusFailed},
+		{"activating", systemd.StatusActivating},
+		{"deactivating", systemd.StatusDeactivating},
+		{"reloading", systemd.StatusReloading},
+		{"unknown", systemd.StatusInactive}, // default case
+		{"", systemd.StatusInactive},        // default case
+	}
+
+	for _, test := range tests {
+		t.Run(test.input, func(t *testing.T) {
+			result := systemd.ParseServiceStatus(test.input)
+			assert.Equal(t, test.expected, result)
+		})
+	}
+}
+
+func TestParseServiceSubState(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected systemd.ServiceSubState
+	}{
+		{"dead", systemd.SubStateDead},
+		{"running", systemd.SubStateRunning},
+		{"exited", systemd.SubStateExited},
+		{"failed", systemd.SubStateFailed},
+		{"unknown", systemd.SubStateUnknown},
+		{"other", systemd.SubStateUnknown}, // default case
+		{"", systemd.SubStateUnknown},      // default case
+	}
+
+	for _, test := range tests {
+		t.Run(test.input, func(t *testing.T) {
+			result := systemd.ParseServiceSubState(test.input)
+			assert.Equal(t, test.expected, result)
+		})
+	}
+}
+
+func TestServiceUpdateCPUPercent(t *testing.T) {
+	t.Run("initial call sets CPU to 0", func(t *testing.T) {
+		service := &systemd.Service{}
+		service.UpdateCPUPercent(1000)
+		assert.Equal(t, 0.0, service.Cpu)
+		assert.Equal(t, uint64(1000), service.PrevCpuUsage)
+		assert.False(t, service.PrevReadTime.IsZero())
+	})
+
+	t.Run("subsequent call calculates CPU percentage", func(t *testing.T) {
+		service := &systemd.Service{}
+		service.PrevCpuUsage = 1000
+		service.PrevReadTime = time.Now().Add(-time.Second)
+
+		service.UpdateCPUPercent(8000000000) // 8 seconds of CPU time
+
+		// CPU usage should be positive and reasonable
+		assert.Greater(t, service.Cpu, 0.0, "CPU usage should be positive")
+		assert.LessOrEqual(t, service.Cpu, 100.0, "CPU usage should not exceed 100%")
+		assert.Equal(t, uint64(8000000000), service.PrevCpuUsage)
+		assert.Greater(t, service.CpuPeak, 0.0, "CPU peak should be set")
+	})
+
+	t.Run("CPU peak updates only when higher", func(t *testing.T) {
+		service := &systemd.Service{}
+		service.PrevCpuUsage = 1000
+		service.PrevReadTime = time.Now().Add(-time.Second)
+		service.UpdateCPUPercent(8000000000) // Set initial peak to ~50%
+		initialPeak := service.CpuPeak
+
+		// Now try with much lower CPU usage - should not update peak
+		service.PrevReadTime = time.Now().Add(-time.Second)
+		service.UpdateCPUPercent(1000000) // Much lower usage
+		assert.Equal(t, initialPeak, service.CpuPeak, "Peak should not update for lower CPU usage")
+	})
+
+	t.Run("handles zero duration", func(t *testing.T) {
+		service := &systemd.Service{}
+		service.PrevCpuUsage = 1000
+		now := time.Now()
+		service.PrevReadTime = now
+		// Mock time.Now() to return the same time to ensure zero duration
+		// Since we can't mock time in Go easily, we'll check the logic manually
+		// The zero duration case happens when duration <= 0
+		assert.Equal(t, 0.0, service.Cpu, "CPU should start at 0")
+	})
+
+	t.Run("handles CPU usage wraparound", func(t *testing.T) {
+		service := &systemd.Service{}
+		// Simulate wraparound where new usage is less than previous
+		service.PrevCpuUsage = 1000
+		service.PrevReadTime = time.Now().Add(-time.Second)
+		service.UpdateCPUPercent(500) // Less than previous, should reset
+		assert.Equal(t, 0.0, service.Cpu)
+	})
+}