完善中央与边缘部署、远程写入与监控文档

- 增加中央与边缘完整配置和部署脚本 - 引入 VictoriaMetrics 数据源与 remote_write 故障排查说明 - 新增 edge-agent 配置脚本、ONVIF 自建 exporter 与 ping 监控示例 Co-authored-by: Cursor <cursoragent@cursor.com>
2026-02-25 04:24:40 -05:00
parent 9e37f79a36
commit 95a09fd9d8
52 changed files with 5978 additions and 0 deletions
--- a/edge-agent/onvif-exporter/main.go
+++ b/edge-agent/onvif-exporter/main.go
@@ -0,0 +1,122 @@
+package main
+
+import (
+	"encoding/json"
+	"log"
+	"net/http"
+	"os"
+	"strings"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+	"github.com/use-go/onvif"
+	"github.com/use-go/onvif/device"
+)
+
+const (
+	defaultTargetsFile = "/config/targets.json"
+	defaultPort        = "9600"
+)
+
+type targetGroup struct {
+	Targets []string          `json:"targets"`
+	Labels  map[string]string `json:"labels"`
+}
+
+func main() {
+	targetsFile := os.Getenv("TARGETS_FILE")
+	if targetsFile == "" {
+		targetsFile = defaultTargetsFile
+	}
+	port := os.Getenv("EXPORTER_PORT")
+	if port == "" {
+		port = defaultPort
+	}
+
+	up := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Name: "onvif_device_up",
+		Help: "1 if ONVIF device is reachable, 0 otherwise",
+	}, []string{"instance", "location", "model", "device_type"})
+	duration := prometheus.NewGaugeVec(prometheus.GaugeOpts{
+		Name: "onvif_probe_duration_seconds",
+		Help: "ONVIF probe duration in seconds",
+	}, []string{"instance", "location", "model", "device_type"})
+
+	reg := prometheus.NewRegistry()
+	reg.MustRegister(up, duration)
+
+	go func() {
+		ticker := time.NewTicker(60 * time.Second)
+		defer ticker.Stop()
+		for ; true; <-ticker.C {
+			probe(targetsFile, up, duration)
+		}
+	}()
+	// 启动时立即探测一次
+	probe(targetsFile, up, duration)
+
+	http.Handle("/metrics", promhttp.HandlerFor(reg, promhttp.HandlerOpts{EnableOpenMetrics: true}))
+	http.HandleFunc("/health", func(w http.ResponseWriter, _ *http.Request) { w.WriteHeader(http.StatusOK) })
+	log.Printf("ONVIF exporter listening on :%s", port)
+	if err := http.ListenAndServe(":"+port, nil); err != nil {
+		log.Fatal(err)
+	}
+}
+
+func probe(targetsFile string, up, duration *prometheus.GaugeVec) {
+	data, err := os.ReadFile(targetsFile)
+	if err != nil {
+		log.Printf("read targets file: %v", err)
+		return
+	}
+	var groups []targetGroup
+	if err := json.Unmarshal(data, &groups); err != nil {
+		log.Printf("parse targets: %v", err)
+		return
+	}
+	for _, g := range groups {
+		if len(g.Targets) == 0 {
+			continue
+		}
+		ip := strings.TrimSpace(g.Targets[0])
+		port := "80"
+		if p, ok := g.Labels["onvif_port"]; ok && p != "" {
+			port = strings.TrimSpace(p)
+		}
+		user := strings.TrimSpace(g.Labels["username"])
+		pass := strings.TrimSpace(g.Labels["password"])
+		location := strings.TrimSpace(g.Labels["location"])
+		model := strings.TrimSpace(g.Labels["model"])
+		deviceType := strings.TrimSpace(g.Labels["device_type"])
+		instance := ip + ":" + port
+		labels := prometheus.Labels{
+			"instance":    instance,
+			"location":    location,
+			"model":       model,
+			"device_type": deviceType,
+		}
+
+		start := time.Now()
+		err := probeONVIF(ip, port, user, pass)
+		elapsed := time.Since(start).Seconds()
+		duration.With(labels).Set(elapsed)
+		if err != nil {
+			log.Printf("onvif probe %s: %v", instance, err)
+			up.With(labels).Set(0)
+		} else {
+			up.With(labels).Set(1)
+		}
+	}
+}
+
+func probeONVIF(ip, port, username, password string) error {
+	xaddr := ip + ":" + port
+	params := onvif.DeviceParams{Xaddr: xaddr, Username: username, Password: password}
+	dev, err := onvif.NewDevice(params)
+	if err != nil {
+		return err
+	}
+	_, err = dev.CallMethod(device.GetDeviceInformation{})
+	return err
+}