nezha/service/singleton/servicesentinel.go
2023-04-15 19:04:38 +08:00

520 lines
19 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 singleton
import (
"fmt"
"log"
"sort"
"strings"
"sync"
"time"
"github.com/naiba/nezha/model"
pb "github.com/naiba/nezha/proto"
"github.com/nicksnyder/go-i18n/v2/i18n"
)
const (
_CurrentStatusSize = 30 // 统计 15 分钟内的数据为当前状态
)
var ServiceSentinelShared *ServiceSentinel
type ReportData struct {
Data *pb.TaskResult
Reporter uint64
}
// _TodayStatsOfMonitor 今日监控记录
type _TodayStatsOfMonitor struct {
Up int // 今日在线计数
Down int // 今日离线计数
Delay float32 // 今日平均延迟
}
// NewServiceSentinel 创建服务监控器
func NewServiceSentinel(serviceSentinelDispatchBus chan<- model.Monitor) {
ServiceSentinelShared = &ServiceSentinel{
serviceReportChannel: make(chan ReportData, 200),
serviceStatusToday: make(map[uint64]*_TodayStatsOfMonitor),
serviceCurrentStatusIndex: make(map[uint64]int),
serviceCurrentStatusData: make(map[uint64][]*pb.TaskResult),
lastStatus: make(map[uint64]int),
serviceResponseDataStoreCurrentUp: make(map[uint64]uint64),
serviceResponseDataStoreCurrentDown: make(map[uint64]uint64),
serviceResponseDataStoreCurrentAvgDelay: make(map[uint64]float32),
monitors: make(map[uint64]*model.Monitor),
sslCertCache: make(map[uint64]string),
// 30天数据缓存
monthlyStatus: make(map[uint64]*model.ServiceItemResponse),
dispatchBus: serviceSentinelDispatchBus,
}
// 加载历史记录
ServiceSentinelShared.loadMonitorHistory()
year, month, day := time.Now().Date()
today := time.Date(year, month, day, 0, 0, 0, 0, Loc)
var mhs []model.MonitorHistory
// 加载当日记录
DB.Where("created_at >= ?", today).Find(&mhs)
totalDelay := make(map[uint64]float32)
totalDelayCount := make(map[uint64]float32)
for i := 0; i < len(mhs); i++ {
totalDelay[mhs[i].MonitorID] += mhs[i].AvgDelay
totalDelayCount[mhs[i].MonitorID]++
ServiceSentinelShared.serviceStatusToday[mhs[i].MonitorID].Up += int(mhs[i].Up)
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].TotalUp += mhs[i].Up
ServiceSentinelShared.serviceStatusToday[mhs[i].MonitorID].Down += int(mhs[i].Down)
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].TotalDown += mhs[i].Down
}
for id, delay := range totalDelay {
ServiceSentinelShared.serviceStatusToday[id].Delay = delay / float32(totalDelayCount[id])
}
// 启动服务监控器
go ServiceSentinelShared.worker()
// 每日将游标往后推一天
_, err := Cron.AddFunc("0 0 0 * * *", ServiceSentinelShared.refreshMonthlyServiceStatus)
if err != nil {
panic(err)
}
}
/*
使用缓存 channel处理上报的 Service 请求结果,然后判断是否需要报警
需要记录上一次的状态信息
加锁顺序serviceResponseDataStoreLock > monthlyStatusLock > monitorsLock
*/
type ServiceSentinel struct {
// 服务监控任务上报通道
serviceReportChannel chan ReportData // 服务状态汇报管道
// 服务监控任务调度通道
dispatchBus chan<- model.Monitor
serviceResponseDataStoreLock sync.RWMutex
serviceStatusToday map[uint64]*_TodayStatsOfMonitor // [monitor_id] -> _TodayStatsOfMonitor
serviceCurrentStatusIndex map[uint64]int // [monitor_id] -> 该监控ID对应的 serviceCurrentStatusData 的最新索引下标
serviceCurrentStatusData map[uint64][]*pb.TaskResult // [monitor_id] -> []model.MonitorHistory
serviceResponseDataStoreCurrentUp map[uint64]uint64 // [monitor_id] -> 当前服务在线计数
serviceResponseDataStoreCurrentDown map[uint64]uint64 // [monitor_id] -> 当前服务离线计数
serviceResponseDataStoreCurrentAvgDelay map[uint64]float32 // [monitor_id] -> 当前服务离线计数
lastStatus map[uint64]int
sslCertCache map[uint64]string
monitorsLock sync.RWMutex
monitors map[uint64]*model.Monitor // [monitor_id] -> model.Monitor
// 30天数据缓存
monthlyStatusLock sync.Mutex
monthlyStatus map[uint64]*model.ServiceItemResponse // [monitor_id] -> model.ServiceItemResponse
}
func (ss *ServiceSentinel) refreshMonthlyServiceStatus() {
// 刷新数据防止无人访问
ss.LoadStats()
// 将数据往前刷一天
ss.serviceResponseDataStoreLock.Lock()
defer ss.serviceResponseDataStoreLock.Unlock()
ss.monthlyStatusLock.Lock()
defer ss.monthlyStatusLock.Unlock()
for k, v := range ss.monthlyStatus {
for i := 0; i < len(v.Up)-1; i++ {
if i == 0 {
// 30 天在线率减去已经出30天之外的数据
v.TotalDown -= uint64(v.Down[i])
v.TotalUp -= uint64(v.Up[i])
}
v.Up[i], v.Down[i], v.Delay[i] = v.Up[i+1], v.Down[i+1], v.Delay[i+1]
}
v.Up[29] = 0
v.Down[29] = 0
v.Delay[29] = 0
// 清理前一天数据
ss.serviceResponseDataStoreCurrentUp[k] = 0
ss.serviceResponseDataStoreCurrentDown[k] = 0
ss.serviceResponseDataStoreCurrentAvgDelay[k] = 0
ss.serviceStatusToday[k].Delay = 0
ss.serviceStatusToday[k].Up = 0
ss.serviceStatusToday[k].Down = 0
}
}
// Dispatch 将传入的 ReportData 传给 服务状态汇报管道
func (ss *ServiceSentinel) Dispatch(r ReportData) {
ss.serviceReportChannel <- r
}
func (ss *ServiceSentinel) Monitors() []*model.Monitor {
ss.monitorsLock.RLock()
defer ss.monitorsLock.RUnlock()
var monitors []*model.Monitor
for _, v := range ss.monitors {
monitors = append(monitors, v)
}
sort.SliceStable(monitors, func(i, j int) bool {
return monitors[i].ID < monitors[j].ID
})
return monitors
}
// LoadStats 加载服务监控器的历史状态信息
func (ss *ServiceSentinel) loadMonitorHistory() {
var monitors []*model.Monitor
err := DB.Find(&monitors).Error
if err != nil {
panic(err)
}
ss.serviceResponseDataStoreLock.Lock()
defer ss.serviceResponseDataStoreLock.Unlock()
ss.monthlyStatusLock.Lock()
defer ss.monthlyStatusLock.Unlock()
ss.monitorsLock.Lock()
defer ss.monitorsLock.Unlock()
for i := 0; i < len(monitors); i++ {
// 旧版本可能不存在通知组 为其设置默认组
if monitors[i].NotificationTag == "" {
monitors[i].NotificationTag = "default"
DB.Save(monitors[i])
}
task := *monitors[i]
// 通过cron定时将服务监控任务传递给任务调度管道
monitors[i].CronJobID, err = Cron.AddFunc(task.CronSpec(), func() {
ss.dispatchBus <- task
})
if err != nil {
panic(err)
}
ss.monitors[monitors[i].ID] = monitors[i]
ss.serviceCurrentStatusData[monitors[i].ID] = make([]*pb.TaskResult, _CurrentStatusSize)
ss.serviceStatusToday[monitors[i].ID] = &_TodayStatsOfMonitor{}
}
year, month, day := time.Now().Date()
today := time.Date(year, month, day, 0, 0, 0, 0, Loc)
for i := 0; i < len(monitors); i++ {
ServiceSentinelShared.monthlyStatus[monitors[i].ID] = &model.ServiceItemResponse{
Monitor: monitors[i],
Delay: &[30]float32{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
Up: &[30]int{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
Down: &[30]int{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
}
}
// 加载服务监控历史记录
var mhs []model.MonitorHistory
DB.Where("created_at > ? AND created_at < ?", today.AddDate(0, 0, -29), today).Find(&mhs)
var delayCount = make(map[int]int)
for i := 0; i < len(mhs); i++ {
dayIndex := 28 - (int(today.Sub(mhs[i].CreatedAt).Hours()) / 24)
if dayIndex < 0 {
continue
}
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].Delay[dayIndex] = (ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].Delay[dayIndex]*float32(delayCount[dayIndex]) + mhs[i].AvgDelay) / float32(delayCount[dayIndex]+1)
delayCount[dayIndex]++
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].Up[dayIndex] += int(mhs[i].Up)
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].TotalUp += mhs[i].Up
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].Down[dayIndex] += int(mhs[i].Down)
ServiceSentinelShared.monthlyStatus[mhs[i].MonitorID].TotalDown += mhs[i].Down
}
}
func (ss *ServiceSentinel) OnMonitorUpdate(m model.Monitor) error {
ss.serviceResponseDataStoreLock.Lock()
defer ss.serviceResponseDataStoreLock.Unlock()
ss.monthlyStatusLock.Lock()
defer ss.monthlyStatusLock.Unlock()
ss.monitorsLock.Lock()
defer ss.monitorsLock.Unlock()
var err error
// 写入新任务
m.CronJobID, err = Cron.AddFunc(m.CronSpec(), func() {
ss.dispatchBus <- m
})
if err != nil {
return err
}
if ss.monitors[m.ID] != nil {
// 停掉旧任务
Cron.Remove(ss.monitors[m.ID].CronJobID)
} else {
// 新任务初始化数据
ss.monthlyStatus[m.ID] = &model.ServiceItemResponse{
Monitor: &m,
Delay: &[30]float32{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
Up: &[30]int{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
Down: &[30]int{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
}
ss.serviceCurrentStatusData[m.ID] = make([]*pb.TaskResult, _CurrentStatusSize)
ss.serviceStatusToday[m.ID] = &_TodayStatsOfMonitor{}
}
// 更新这个任务
ss.monitors[m.ID] = &m
return nil
}
func (ss *ServiceSentinel) OnMonitorDelete(id uint64) {
ss.serviceResponseDataStoreLock.Lock()
defer ss.serviceResponseDataStoreLock.Unlock()
ss.monthlyStatusLock.Lock()
defer ss.monthlyStatusLock.Unlock()
ss.monitorsLock.Lock()
defer ss.monitorsLock.Unlock()
delete(ss.serviceCurrentStatusIndex, id)
delete(ss.serviceCurrentStatusData, id)
delete(ss.lastStatus, id)
delete(ss.serviceResponseDataStoreCurrentUp, id)
delete(ss.serviceResponseDataStoreCurrentDown, id)
delete(ss.serviceResponseDataStoreCurrentAvgDelay, id)
delete(ss.sslCertCache, id)
delete(ss.serviceStatusToday, id)
// 停掉定时任务
Cron.Remove(ss.monitors[id].CronJobID)
delete(ss.monitors, id)
delete(ss.monthlyStatus, id)
}
func (ss *ServiceSentinel) LoadStats() map[uint64]*model.ServiceItemResponse {
ss.serviceResponseDataStoreLock.RLock()
defer ss.serviceResponseDataStoreLock.RUnlock()
ss.monthlyStatusLock.Lock()
defer ss.monthlyStatusLock.Unlock()
// 刷新最新一天的数据
for k := range ss.monitors {
ss.monthlyStatus[k].Monitor = ss.monitors[k]
v := ss.serviceStatusToday[k]
// 30 天在线率,
// |- 减去上次加的旧当天数据,防止出现重复计数
ss.monthlyStatus[k].TotalUp -= uint64(ss.monthlyStatus[k].Up[29])
ss.monthlyStatus[k].TotalDown -= uint64(ss.monthlyStatus[k].Down[29])
// |- 加上当日数据
ss.monthlyStatus[k].TotalUp += uint64(v.Up)
ss.monthlyStatus[k].TotalDown += uint64(v.Down)
ss.monthlyStatus[k].Up[29] = v.Up
ss.monthlyStatus[k].Down[29] = v.Down
ss.monthlyStatus[k].Delay[29] = v.Delay
}
// 最后 5 分钟的状态 与 monitor 对象填充
for k, v := range ss.serviceResponseDataStoreCurrentDown {
ss.monthlyStatus[k].CurrentDown = v
}
for k, v := range ss.serviceResponseDataStoreCurrentUp {
ss.monthlyStatus[k].CurrentUp = v
}
return ss.monthlyStatus
}
// worker 服务监控的实际工作流程
func (ss *ServiceSentinel) worker() {
// 从服务状态汇报管道获取汇报的服务数据
for r := range ss.serviceReportChannel {
if ss.monitors[r.Data.GetId()] == nil || ss.monitors[r.Data.GetId()].ID == 0 {
log.Printf("NEZAH>> 错误的服务监控上报 %+v", r)
continue
}
mh := r.Data
ss.serviceResponseDataStoreLock.Lock()
// 写入当天状态
if mh.Successful {
ss.serviceStatusToday[mh.GetId()].Delay = (ss.serviceStatusToday[mh.
GetId()].Delay*float32(ss.serviceStatusToday[mh.GetId()].Up) +
mh.Delay) / float32(ss.serviceStatusToday[mh.GetId()].Up+1)
ss.serviceStatusToday[mh.GetId()].Up++
} else {
ss.serviceStatusToday[mh.GetId()].Down++
}
// 写入当前数据
ss.serviceCurrentStatusData[mh.GetId()][ss.serviceCurrentStatusIndex[mh.GetId()]] = mh
ss.serviceCurrentStatusIndex[mh.GetId()]++
// 更新当前状态
ss.serviceResponseDataStoreCurrentUp[mh.GetId()] = 0
ss.serviceResponseDataStoreCurrentDown[mh.GetId()] = 0
ss.serviceResponseDataStoreCurrentAvgDelay[mh.GetId()] = 0
// 永远是最新的 30 个数据的状态 [01:00, 02:00, 03:00] -> [04:00, 02:00, 03: 00]
for i := 0; i < len(ss.serviceCurrentStatusData[mh.GetId()]); i++ {
if ss.serviceCurrentStatusData[mh.GetId()][i].GetId() > 0 {
if ss.serviceCurrentStatusData[mh.GetId()][i].Successful {
ss.serviceResponseDataStoreCurrentUp[mh.GetId()]++
ss.serviceResponseDataStoreCurrentAvgDelay[mh.GetId()] = (ss.serviceResponseDataStoreCurrentAvgDelay[mh.GetId()]*float32(ss.serviceResponseDataStoreCurrentUp[mh.GetId()]-1) + ss.serviceCurrentStatusData[mh.GetId()][i].Delay) / float32(ss.serviceResponseDataStoreCurrentUp[mh.GetId()])
} else {
ss.serviceResponseDataStoreCurrentDown[mh.GetId()]++
}
}
}
// 计算在线率,
var upPercent uint64 = 0
if ss.serviceResponseDataStoreCurrentDown[mh.GetId()]+ss.serviceResponseDataStoreCurrentUp[mh.GetId()] > 0 {
upPercent = ss.serviceResponseDataStoreCurrentUp[mh.GetId()] * 100 / (ss.serviceResponseDataStoreCurrentDown[mh.GetId()] + ss.serviceResponseDataStoreCurrentUp[mh.GetId()])
}
stateCode := GetStatusCode(upPercent)
// 数据持久化
if ss.serviceCurrentStatusIndex[mh.GetId()] == _CurrentStatusSize {
ss.serviceCurrentStatusIndex[mh.GetId()] = 0
if err := DB.Create(&model.MonitorHistory{
MonitorID: mh.GetId(),
AvgDelay: ss.serviceResponseDataStoreCurrentAvgDelay[mh.GetId()],
Data: mh.Data,
Up: ss.serviceResponseDataStoreCurrentUp[mh.GetId()],
Down: ss.serviceResponseDataStoreCurrentDown[mh.GetId()],
}).Error; err != nil {
log.Println("NEZHA>> 服务监控数据持久化失败:", err)
}
}
// 延迟报警
if mh.Delay > 0 {
ss.monitorsLock.RLock()
if ss.monitors[mh.GetId()].LatencyNotify {
if mh.Delay > ss.monitors[mh.GetId()].MaxLatency {
ServerLock.RLock()
reporterServer := ServerList[r.Reporter]
msg := fmt.Sprintf("[Latency] %s %2f > %2f, Reporter: %s", ss.monitors[mh.GetId()].Name, mh.Delay, ss.monitors[mh.GetId()].MaxLatency, reporterServer.Name)
muteLabel := NotificationMuteLabel.ServiceLatencyMin(mh.GetId())
go SendNotification(ss.monitors[mh.GetId()].NotificationTag, msg, muteLabel)
ServerLock.RUnlock()
}
if mh.Delay < ss.monitors[mh.GetId()].MinLatency {
ServerLock.RLock()
reporterServer := ServerList[r.Reporter]
go SendNotification(ss.monitors[mh.GetId()].NotificationTag, fmt.Sprintf("[Latency] %s %2f < %2f, Reporter: %s", ss.monitors[mh.GetId()].Name, mh.Delay, ss.monitors[mh.GetId()].MinLatency, reporterServer.Name), NotificationMuteLabel.ServiceLatencyMax(mh.GetId()))
ServerLock.RUnlock()
}
}
ss.monitorsLock.RUnlock()
}
// 状态变更报警
if stateCode == StatusDown || stateCode != ss.lastStatus[mh.GetId()] {
ss.monitorsLock.Lock()
lastStatus := ss.lastStatus[mh.GetId()]
// 存储新的状态值
ss.lastStatus[mh.GetId()] = stateCode
// 判断是否需要发送提醒
isNeedSendNotification := ss.monitors[mh.GetId()].Notify && (lastStatus != 0 || stateCode == StatusDown)
if isNeedSendNotification {
ServerLock.RLock()
reporterServer := ServerList[r.Reporter]
go SendNotification(ss.monitors[mh.GetId()].NotificationTag, fmt.Sprintf("[%s] %s Reporter: %s, Error: %s", StatusCodeToString(stateCode), ss.monitors[mh.GetId()].Name, reporterServer.Name, mh.Data), NotificationMuteLabel.ServiceStateChanged(mh.GetId()))
ServerLock.RUnlock()
}
// 判断是否需要触发任务
if ss.monitors[mh.GetId()].EnableTriggerTask && lastStatus != 0 {
ServerLock.RLock()
reporterServer := ServerList[r.Reporter]
ServerLock.RUnlock()
if stateCode == StatusGood && lastStatus != stateCode {
// 当前状态正常 前序状态异常时 触发恢复任务
go SendTriggerTasks(ss.monitors[mh.GetId()].RecoverTriggerTasks, reporterServer.ID)
} else if lastStatus == StatusGood && lastStatus != stateCode {
// 前序状态正常 当前状态异常时 触发失败任务
go SendTriggerTasks(ss.monitors[mh.GetId()].FailTriggerTasks, reporterServer.ID)
}
}
// 当前状态正常 前序状态非正常时触发恢复任务
ss.monitorsLock.Unlock()
}
ss.serviceResponseDataStoreLock.Unlock()
// SSL 证书报警
var errMsg string
if strings.HasPrefix(mh.Data, "SSL证书错误") {
// 排除 i/o timeout、connection timeout、EOF 错误
if !strings.HasSuffix(mh.Data, "timeout") &&
!strings.HasSuffix(mh.Data, "EOF") &&
!strings.HasSuffix(mh.Data, "timed out") {
errMsg = mh.Data
}
} else {
var newCert = strings.Split(mh.Data, "|")
if len(newCert) > 1 {
if ss.sslCertCache[mh.GetId()] == "" {
ss.sslCertCache[mh.GetId()] = mh.Data
}
expiresNew, _ := time.Parse("2006-01-02 15:04:05 -0700 MST", newCert[1])
// 证书过期提醒
if expiresNew.Before(time.Now().AddDate(0, 0, 7)) {
errMsg = fmt.Sprintf(
"The SSL certificate will expire within seven days. Expiration time: %s",
expiresNew.Format("2006-01-02 15:04:05"))
}
// 证书变更提醒
var oldCert = strings.Split(ss.sslCertCache[mh.GetId()], "|")
var expiresOld time.Time
if len(oldCert) > 1 {
expiresOld, _ = time.Parse("2006-01-02 15:04:05 -0700 MST", oldCert[1])
}
if oldCert[0] != newCert[0] && !expiresNew.Equal(expiresOld) {
ss.sslCertCache[mh.GetId()] = mh.Data
errMsg = fmt.Sprintf(
"SSL certificate changed, old: %s, %s expired; new: %s, %s expired.",
oldCert[0], expiresOld.Format("2006-01-02 15:04:05"), newCert[0], expiresNew.Format("2006-01-02 15:04:05"))
}
}
}
if errMsg != "" {
ss.monitorsLock.RLock()
if ss.monitors[mh.GetId()].Notify {
go SendNotification(ss.monitors[mh.GetId()].NotificationTag, fmt.Sprintf("[SSL] %s %s", ss.monitors[mh.GetId()].Name, errMsg), NotificationMuteLabel.ServiceSSL(mh.GetId()))
}
ss.monitorsLock.RUnlock()
}
}
}
const (
_ = iota
StatusNoData
StatusGood
StatusLowAvailability
StatusDown
)
func GetStatusCode[T float32 | uint64](percent T) int {
if percent == 0 {
return StatusNoData
}
if percent > 95 {
return StatusGood
}
if percent > 80 {
return StatusLowAvailability
}
return StatusDown
}
func StatusCodeToString(statusCode int) string {
switch statusCode {
case StatusNoData:
return Localizer.MustLocalize(&i18n.LocalizeConfig{MessageID: "StatusNoData"})
case StatusGood:
return Localizer.MustLocalize(&i18n.LocalizeConfig{MessageID: "StatusGood"})
case StatusLowAvailability:
return Localizer.MustLocalize(&i18n.LocalizeConfig{MessageID: "StatusLowAvailability"})
case StatusDown:
return Localizer.MustLocalize(&i18n.LocalizeConfig{MessageID: "StatusDown"})
default:
return ""
}
}