Files
junhong_cmp_fiber/internal/service/polling/monitoring_service.go
huang 931e140e8e
All checks were successful
构建并部署到测试环境(无 SSH) / build-and-deploy (push) Successful in 6m35s
feat: 实现 IoT 卡轮询系统(支持千万级卡规模)
实现功能:
- 实名状态检查轮询(可配置间隔)
- 卡流量检查轮询(支持跨月流量追踪)
- 套餐检查与超额自动停机
- 分布式并发控制(Redis 信号量)
- 手动触发轮询(单卡/批量/条件筛选)
- 数据清理配置与执行
- 告警规则与历史记录
- 实时监控统计(队列/性能/并发)

性能优化:
- Redis 缓存卡信息,减少 DB 查询
- Pipeline 批量写入 Redis
- 异步流量记录写入
- 渐进式初始化(10万卡/批)

压测工具(scripts/benchmark/):
- Mock Gateway 模拟上游服务
- 测试卡生成器
- 配置初始化脚本
- 实时监控脚本

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-05 17:32:44 +08:00

284 lines
8.9 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
package polling
import (
"context"
"time"
"github.com/redis/go-redis/v9"
"github.com/break/junhong_cmp_fiber/pkg/constants"
"github.com/break/junhong_cmp_fiber/pkg/errors"
)
// MonitoringService 轮询监控服务
type MonitoringService struct {
redis *redis.Client
}
// NewMonitoringService 创建轮询监控服务实例
func NewMonitoringService(redis *redis.Client) *MonitoringService {
return &MonitoringService{redis: redis}
}
// OverviewStats 总览统计
type OverviewStats struct {
TotalCards int64 `json:"total_cards"` // 总卡数
InitializedCards int64 `json:"initialized_cards"` // 已初始化卡数
InitProgress float64 `json:"init_progress"` // 初始化进度 (0-100)
IsInitializing bool `json:"is_initializing"` // 是否正在初始化
RealnameQueueSize int64 `json:"realname_queue_size"` // 实名检查队列大小
CarddataQueueSize int64 `json:"carddata_queue_size"` // 流量检查队列大小
PackageQueueSize int64 `json:"package_queue_size"` // 套餐检查队列大小
}
// QueueStatus 队列状态
type QueueStatus struct {
TaskType string `json:"task_type"` // 任务类型
TaskTypeName string `json:"task_type_name"` // 任务类型名称
QueueSize int64 `json:"queue_size"` // 队列大小
ManualPending int64 `json:"manual_pending"` // 手动触发待处理数
DueCount int64 `json:"due_count"` // 到期待处理数
AvgWaitTime float64 `json:"avg_wait_time_s"` // 平均等待时间(秒)
}
// TaskStats 任务统计
type TaskStats struct {
TaskType string `json:"task_type"` // 任务类型
TaskTypeName string `json:"task_type_name"` // 任务类型名称
SuccessCount1h int64 `json:"success_count_1h"` // 1小时成功数
FailureCount1h int64 `json:"failure_count_1h"` // 1小时失败数
TotalCount1h int64 `json:"total_count_1h"` // 1小时总数
SuccessRate float64 `json:"success_rate"` // 成功率 (0-100)
AvgDurationMs float64 `json:"avg_duration_ms"` // 平均耗时(毫秒)
}
// InitProgress 初始化进度
type InitProgress struct {
TotalCards int64 `json:"total_cards"` // 总卡数
InitializedCards int64 `json:"initialized_cards"` // 已初始化卡数
Progress float64 `json:"progress"` // 进度百分比 (0-100)
IsComplete bool `json:"is_complete"` // 是否完成
StartedAt time.Time `json:"started_at"` // 开始时间
EstimatedETA string `json:"estimated_eta"` // 预计完成时间
}
// GetOverview 获取总览统计
func (s *MonitoringService) GetOverview(ctx context.Context) (*OverviewStats, error) {
stats := &OverviewStats{}
// 获取初始化进度
progressKey := constants.RedisPollingInitProgressKey()
progressData, err := s.redis.HGetAll(ctx, progressKey).Result()
if err != nil && err != redis.Nil {
return nil, errors.Wrap(errors.CodeRedisError, err, "获取初始化进度失败")
}
if total, ok := progressData["total"]; ok {
stats.TotalCards = parseInt64(total)
}
if initialized, ok := progressData["initialized"]; ok {
stats.InitializedCards = parseInt64(initialized)
}
if stats.TotalCards > 0 {
stats.InitProgress = float64(stats.InitializedCards) / float64(stats.TotalCards) * 100
}
stats.IsInitializing = stats.InitializedCards < stats.TotalCards && stats.TotalCards > 0
// 获取队列大小
stats.RealnameQueueSize, _ = s.redis.ZCard(ctx, constants.RedisPollingQueueRealnameKey()).Result()
stats.CarddataQueueSize, _ = s.redis.ZCard(ctx, constants.RedisPollingQueueCarddataKey()).Result()
stats.PackageQueueSize, _ = s.redis.ZCard(ctx, constants.RedisPollingQueuePackageKey()).Result()
return stats, nil
}
// GetQueueStatuses 获取所有队列状态
func (s *MonitoringService) GetQueueStatuses(ctx context.Context) ([]*QueueStatus, error) {
taskTypes := []string{
constants.TaskTypePollingRealname,
constants.TaskTypePollingCarddata,
constants.TaskTypePollingPackage,
}
result := make([]*QueueStatus, 0, len(taskTypes))
now := time.Now().Unix()
for _, taskType := range taskTypes {
status := &QueueStatus{
TaskType: taskType,
TaskTypeName: s.getTaskTypeName(taskType),
}
// 获取队列 key
var queueKey string
switch taskType {
case constants.TaskTypePollingRealname:
queueKey = constants.RedisPollingQueueRealnameKey()
case constants.TaskTypePollingCarddata:
queueKey = constants.RedisPollingQueueCarddataKey()
case constants.TaskTypePollingPackage:
queueKey = constants.RedisPollingQueuePackageKey()
}
// 获取队列大小
status.QueueSize, _ = s.redis.ZCard(ctx, queueKey).Result()
// 获取到期数量score <= now
status.DueCount, _ = s.redis.ZCount(ctx, queueKey, "-inf", formatInt64(now)).Result()
// 获取手动触发队列待处理数
manualKey := constants.RedisPollingManualQueueKey(taskType)
status.ManualPending, _ = s.redis.LLen(ctx, manualKey).Result()
// 计算平均等待时间取最早的10个任务的平均等待时间
earliest, err := s.redis.ZRangeWithScores(ctx, queueKey, 0, 9).Result()
if err == nil && len(earliest) > 0 {
var totalWait float64
for _, z := range earliest {
waitTime := float64(now) - z.Score
if waitTime > 0 {
totalWait += waitTime
}
}
status.AvgWaitTime = totalWait / float64(len(earliest))
}
result = append(result, status)
}
return result, nil
}
// GetTaskStatuses 获取所有任务统计
func (s *MonitoringService) GetTaskStatuses(ctx context.Context) ([]*TaskStats, error) {
taskTypes := []string{
constants.TaskTypePollingRealname,
constants.TaskTypePollingCarddata,
constants.TaskTypePollingPackage,
}
result := make([]*TaskStats, 0, len(taskTypes))
for _, taskType := range taskTypes {
stats := &TaskStats{
TaskType: taskType,
TaskTypeName: s.getTaskTypeName(taskType),
}
// 获取统计数据
statsKey := constants.RedisPollingStatsKey(taskType)
data, err := s.redis.HGetAll(ctx, statsKey).Result()
if err != nil && err != redis.Nil {
continue
}
if success, ok := data["success_count_1h"]; ok {
stats.SuccessCount1h = parseInt64(success)
}
if failure, ok := data["failure_count_1h"]; ok {
stats.FailureCount1h = parseInt64(failure)
}
stats.TotalCount1h = stats.SuccessCount1h + stats.FailureCount1h
if stats.TotalCount1h > 0 {
stats.SuccessRate = float64(stats.SuccessCount1h) / float64(stats.TotalCount1h) * 100
if duration, ok := data["total_duration_1h"]; ok {
totalDuration := parseInt64(duration)
stats.AvgDurationMs = float64(totalDuration) / float64(stats.TotalCount1h)
}
}
result = append(result, stats)
}
return result, nil
}
// GetInitProgress 获取初始化进度详情
func (s *MonitoringService) GetInitProgress(ctx context.Context) (*InitProgress, error) {
progressKey := constants.RedisPollingInitProgressKey()
data, err := s.redis.HGetAll(ctx, progressKey).Result()
if err != nil && err != redis.Nil {
return nil, errors.Wrap(errors.CodeRedisError, err, "获取初始化进度失败")
}
progress := &InitProgress{}
if total, ok := data["total"]; ok {
progress.TotalCards = parseInt64(total)
}
if initialized, ok := data["initialized"]; ok {
progress.InitializedCards = parseInt64(initialized)
}
if startedAt, ok := data["started_at"]; ok {
if t, err := time.Parse(time.RFC3339, startedAt); err == nil {
progress.StartedAt = t
}
}
if progress.TotalCards > 0 {
progress.Progress = float64(progress.InitializedCards) / float64(progress.TotalCards) * 100
}
progress.IsComplete = progress.InitializedCards >= progress.TotalCards && progress.TotalCards > 0
// 估算完成时间
if !progress.IsComplete && progress.InitializedCards > 0 && !progress.StartedAt.IsZero() {
elapsed := time.Since(progress.StartedAt)
remaining := progress.TotalCards - progress.InitializedCards
rate := float64(progress.InitializedCards) / elapsed.Seconds()
if rate > 0 {
etaSeconds := float64(remaining) / rate
eta := time.Now().Add(time.Duration(etaSeconds) * time.Second)
progress.EstimatedETA = eta.Format("15:04:05")
}
}
return progress, nil
}
// getTaskTypeName 获取任务类型的中文名称
func (s *MonitoringService) getTaskTypeName(taskType string) string {
switch taskType {
case constants.TaskTypePollingRealname:
return "实名检查"
case constants.TaskTypePollingCarddata:
return "流量检查"
case constants.TaskTypePollingPackage:
return "套餐检查"
default:
return taskType
}
}
// parseInt64 解析字符串为 int64
func parseInt64(s string) int64 {
var result int64
for _, c := range s {
if c >= '0' && c <= '9' {
result = result*10 + int64(c-'0')
}
}
return result
}
// formatInt64 格式化 int64 为字符串
func formatInt64(n int64) string {
if n == 0 {
return "0"
}
var result []byte
negative := n < 0
if negative {
n = -n
}
for n > 0 {
result = append([]byte{byte('0' + n%10)}, result...)
n /= 10
}
if negative {
result = append([]byte{'-'}, result...)
}
return string(result)
}