Add homelab service and system health monitoring

system_health.go

@@ -0,0 +1,84 @@
+package main
+
+import (
+	"os/exec"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+)
+
+type SystemIssue struct {
+	Time     int64  `json:"time"`
+	Unit     string `json:"unit"`
+	Priority string `json:"priority"`
+	Message  string `json:"message"`
+}
+type SystemHealthMetrics struct {
+	Available    bool          `json:"available"`
+	FailedUnits  int           `json:"failedUnits"`
+	KernelErrors int           `json:"kernelErrors"`
+	SystemErrors int           `json:"systemErrors"`
+	Issues       []SystemIssue `json:"issues"`
+	CheckedAt    int64         `json:"checkedAt"`
+}
+type systemHealthCollector struct {
+	mu    sync.RWMutex
+	value SystemHealthMetrics
+}
+
+func newSystemHealthCollector() *systemHealthCollector {
+	c := &systemHealthCollector{}
+	go c.run()
+	return c
+}
+func (c *systemHealthCollector) collect() SystemHealthMetrics {
+	c.mu.RLock()
+	defer c.mu.RUnlock()
+	return c.value
+}
+func (c *systemHealthCollector) run() {
+	c.refresh()
+	t := time.NewTicker(time.Minute)
+	defer t.Stop()
+	for range t.C {
+		c.refresh()
+	}
+}
+func (c *systemHealthCollector) refresh() {
+	m := SystemHealthMetrics{CheckedAt: time.Now().Unix()}
+	if _, err := exec.LookPath("journalctl"); err != nil {
+		c.mu.Lock()
+		c.value = m
+		c.mu.Unlock()
+		return
+	}
+	m.Available = true
+	if out, err := exec.Command("systemctl", "--failed", "--no-legend", "--plain").Output(); err == nil {
+		for _, line := range strings.Split(strings.TrimSpace(string(out)), "\n") {
+			if strings.TrimSpace(line) != "" {
+				m.FailedUnits++
+			}
+		}
+	}
+	out, _ := exec.Command("journalctl", "--since", "24 hours ago", "-p", "0..3", "--no-pager", "-o", "short-unix", "-n", "100").Output()
+	for _, line := range strings.Split(strings.TrimSpace(string(out)), "\n") {
+		f := strings.Fields(line)
+		if len(f) < 2 {
+			continue
+		}
+		raw := strings.TrimSuffix(f[0], ":")
+		sec, _ := strconv.ParseFloat(raw, 64)
+		unit := strings.TrimSuffix(f[1], ":")
+		message := strings.Join(f[2:], " ")
+		issue := SystemIssue{Time: int64(sec), Unit: unit, Priority: "error", Message: message}
+		m.Issues = append(m.Issues, issue)
+		m.SystemErrors++
+		if strings.Contains(strings.ToLower(unit), "kernel") {
+			m.KernelErrors++
+		}
+	}
+	c.mu.Lock()
+	c.value = m
+	c.mu.Unlock()
+}