Files
go-sms-V2/internal/pkg/metrics/prometheus.go
T
Tory 01c8874831
ci / test (push) Canceled after 0s
ci / web-test (push) Canceled after 0s
ci / build (push) Canceled after 0s
ci / deploy-dev (push) Canceled after 0s
S4-D3: 明细按天分区热5天+ClickHouse冷热切换——S4泳道收官(冷查16ms<1s验收✅): ①分区=迁移v22 PARTITION BY RANGE(id)(雪花时间位当天零点边界,主键即分区键零唯一键改造;热5天+未来3天缓冲+pmax兜底;MySQL守卫sqlite跳过)+archive/partition.go EnsurePartitions幂等滚动(REORGANIZE pmax拆分;过期分区归档校验后才DROP)+真机9分区存量无损;②归档=archive域(CH表toDate分区/ArchiveJob 1h:CH当日分区重写幂等→分批2000迁出→count校验相等才DROP MySQL分区或DELETE兜底=零丢失方向;consumer装配启动即首轮)+metrics archive_rows/errors;③冷查=GET /record/history(CH查询/task_no MySQL转译回填/D26契约直连)真机16ms+mobile筛选✓,导出挂账D29下载中心;④settle-check热窗感知=月中对账DetailPartial=true明细↔汇总退化两方基线(日汇总↔结算单,reconcile头注预声明落地)+settle.Service时钟注入nowFn+测试相对时间/period动态;⑤真机5连坑全修存档:MySQL8.0 information_schema非分区表也返回NULL行(判据加PARTITION_NAME IS NOT NULL)/parseTime下DATE() Scan []string带RFC3339尾巴(DISTINCT时刻Scan []time.Time+Go去重双方言)/CH24.3无DROP PARTITION IF EXISTS(先查system.parts)/CH count()返回UInt64(*int64不支持+count错误不可吞)/CH绑定time.Time UTC歧义(预格式化墙钟串)——techContext存档;⑥门禁=build/vet/lint 0/-race 41包全绿/迁移v16..v22两遍幂等(v15断言随窗口右移移除);archive单测5件+settle热窗外基线用例;验收单s4-d3-partition-archive.md+CHANGELOG+memory-bank(S4收官快照/S5下一步/挂账清单)
2026-11-16 12:11:31 +08:00

251 lines
10 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 metrics Prometheus 指标中心(Day 13):
// 独立 Registry(不污染全局)+ HTTP 服务指标三件套 + 业务指标预留。
// 暴露方式:app 装配 metrics.Handler() 至 GET /metrics。
package metrics
import (
"net/http"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/collectors"
"github.com/prometheus/client_golang/prometheus/promhttp"
)
// durationBuckets 耗时直方图分桶(对齐 P99<50ms 目标精化低区间)。
var durationBuckets = []float64{
0.001, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10,
}
// HTTP 指标标签维度。
const (
LabelMethod = "method"
LabelRoute = "route"
LabelStatus = "status"
LabelTopic = "topic"
LabelKind = "kind"
)
var (
// HTTPRequestsTotal HTTP 请求总数(method/route/status)。
HTTPRequestsTotal *prometheus.CounterVec
// HTTPRequestDuration HTTP 请求耗时分布(秒,method/route)。
HTTPRequestDuration *prometheus.HistogramVec
// HTTPRequestsInFlight 当前在途请求数。
HTTPRequestsInFlight prometheus.Gauge
// SMSSubmitTotal 短信提交总数(阶段 4 发送链路埋点:result)。
SMSSubmitTotal *prometheus.CounterVec
// SMSReportTotal 状态回执处理总数(阶段 5:status)。
SMSReportTotal *prometheus.CounterVec
// KafkaProducerMessages Kafka 生产结果(topic/kind=ok|retry|dlq)。
KafkaProducerMessages *prometheus.CounterVec
// ---- 资金域(Day54 R1,前缀 go_sms_balance_ 与 alert_rules.yml/评审纪要三方锁定) ----
// BalanceReconcileDiff 对账差额(Redis 余额 vs 流水重算;覆写前采样,{account})。
BalanceReconcileDiff *prometheus.GaugeVec
// BalanceMode 当前预扣模式(1=redis Lua/0=mysql 行锁;flag 切换即置)。
BalanceMode prometheus.Gauge
// BalanceFallbackTotal Redis 基础设施错回退 MySQL 次数(kill Redis 验收观测点)。
BalanceFallbackTotal prometheus.Counter
// BalanceFlowlogPending 异步流水队列水位(批写链路健康度)。
BalanceFlowlogPending prometheus.Gauge
// ---- 消费域(Day55,挂账清偿) ----
// KafkaLag submit 消费组 lag 合计({topic};背压 L1 告警源,Prometheus 抓取)。
KafkaLag *prometheus.GaugeVec
// ---- 长短信聚合域(Day 69,百万级压测观测;Day 71 挂起制语义更新) ----
// LongSmsMergedTotal 入向长短信聚合成功数(组到齐合并单次受理)。
LongSmsMergedTotal prometheus.Counter
// LongSmsDegradeTotal 退化直发片数({reason}: overflow=并发组超上限新组拒绝/
// ref_conflict=同组 total 不一致协议异常片直发。Day 71 起 timeout 语义废除,
// 挂起组到期走 LongSmsExpireFailTotal)。
LongSmsDegradeTotal *prometheus.CounterVec
// LongSmsExpireFailTotal Day 71:挂起组 72h 到期失败终态数(不齐不发,
// 建 LP task+失败 detail+UNDELIV 回执给下家)。
LongSmsExpireFailTotal prometheus.Counter
// LongSmsDupFragTotal 同 idx 重复片幂等覆盖数(客户端重发观测)。
LongSmsDupFragTotal prometheus.Counter
// ReportRequeueTotal unpaired 回执短重投次数(秒回竞态窗口观测)。
ReportRequeueTotal prometheus.Counter
// ---- 回执聚合域(Day 69 二轮,百万级压测观测) ----
// ReportPendingTotal 多段聚合未到齐挂起事件数。
ReportPendingTotal prometheus.Counter
// ReportRecheckTotal pending 后新事务复查结果({result}: finalized=补刀
// 终态/still_pending=真未到齐;双盲竞态健康度=finalized 占比)。
ReportRecheckTotal *prometheus.CounterVec
// SettleAggregateRows 结算域消耗日汇总落账行数(scope: account=billing_
// consume_log / channel=stats_channel_daily;幂等 UPSERT 重刷亦计入,
// 单调累计供观测 job 活性,S5 基线期再评估是否细化)。
SettleAggregateRows *prometheus.CounterVec
// SettleAggregateErrors 消耗日汇总执行失败次数(失败 Warn 不中断,
// 幂等下轮自愈;持续增长=真故障信号)。
SettleAggregateErrors prometheus.Counter
// ---- 回调推送域(S4·C1) ----
// CallbackPushTotal 推送结果计数({type}: report/uplink × {result}:
// sent=2xx 送达/retry=失败待重试/dead=死信)。
CallbackPushTotal *prometheus.CounterVec
// CallbackBlacklistTotal 上行自动拉黑新增数(S4·C2;幂等冲突不计)。
CallbackBlacklistTotal prometheus.Counter
// ArchiveRows 归档迁出 MySQL→CH 的明细行数(单调累计,S4·D3)。
ArchiveRows prometheus.Counter
// ArchiveErrors 归档链路失败计数(校验不匹配/CH/MySQL 故障;下轮自愈,
// 持续增长=真故障信号)。
ArchiveErrors prometheus.Counter
)
var registry = prometheus.NewRegistry()
func init() {
HTTPRequestsTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "http_requests_total",
Help: "HTTP 请求总数",
}, []string{LabelMethod, LabelRoute, LabelStatus})
HTTPRequestDuration = prometheus.NewHistogramVec(prometheus.HistogramOpts{
Name: "http_request_duration_seconds",
Help: "HTTP 请求耗时分布(秒)",
Buckets: durationBuckets,
}, []string{LabelMethod, LabelRoute})
HTTPRequestsInFlight = prometheus.NewGauge(prometheus.GaugeOpts{
Name: "http_requests_in_flight",
Help: "当前在途请求数",
})
SMSSubmitTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "sms_submit_total",
Help: "短信提交总数(阶段 4 发送链路)",
}, []string{"result"})
SMSReportTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "sms_report_total",
Help: "状态回执处理总数(阶段 5)",
}, []string{"status"})
KafkaProducerMessages = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "kafka_producer_messages_total",
Help: "Kafka 生产消息结果(ok/retry/dlq)",
}, []string{LabelTopic, LabelKind})
BalanceReconcileDiff = prometheus.NewGaugeVec(prometheus.GaugeOpts{
Name: "go_sms_balance_reconcile_diff",
Help: "R1 对账差额:Redis 预扣余额-流水重算值(覆写前采样;0=一致)",
}, []string{"account"})
BalanceMode = prometheus.NewGauge(prometheus.GaugeOpts{
Name: "go_sms_balance_mode",
Help: "预扣模式(1=redis Lua/0=mysql 行锁;GOSMS_BALANCE_MODE 切换)",
})
BalanceFallbackTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_balance_fallback_total",
Help: "Redis 预扣基础设施错单次回退 MySQL 行锁次数",
})
BalanceFlowlogPending = prometheus.NewGauge(prometheus.GaugeOpts{
Name: "go_sms_balance_flowlog_pending",
Help: "Redis 模式异步流水队列待写条数",
})
KafkaLag = prometheus.NewGaugeVec(prometheus.GaugeOpts{
Name: "go_sms_kafka_lag",
Help: "Kafka 消费组 lag 合计(高水位-已提交;背压 L1 告警源,Day53 挂账 Day55 清偿)",
}, []string{LabelTopic})
LongSmsMergedTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_longsms_merged_total",
Help: "入向长短信聚合成功数(UDH 组到齐合并单次受理,Day69)",
})
LongSmsDegradeTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "go_sms_longsms_degrade_total",
Help: "长短信退化直发片数(reason: overflow=并发组超上限/ref_conflict=协议异常;Day71 起 timeout 废除)",
}, []string{LabelKind})
LongSmsExpireFailTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_longsms_expire_fail_total",
Help: "长短信挂起组 72h 到期失败终态数(不齐不发,LP task+UNDELIV 回执,Day71)",
})
LongSmsDupFragTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_longsms_dup_frag_total",
Help: "长短信同 idx 重复片幂等覆盖数(客户端重发观测)",
})
ReportRequeueTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_report_requeue_total",
Help: "unpaired 回执短重投次数(seg 落库竞态窗口,Day69 三轮)",
})
ReportPendingTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_report_pending_total",
Help: "多段回执聚合未到齐挂起事件数(Day69 二轮)",
})
ReportRecheckTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "go_sms_report_recheck_total",
Help: "pending 后新事务复查结果(result: finalized=补刀终态/still_pending=真未到齐)",
}, []string{LabelKind})
SettleAggregateRows = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "go_sms_settle_aggregate_rows_total",
Help: "结算域消耗日汇总落账行数(scope: account/channel;UPSERT 重刷亦计入,S2)",
}, []string{LabelKind})
SettleAggregateErrors = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_settle_aggregate_errors_total",
Help: "消耗日汇总执行失败次数(幂等下轮自愈,持续增长=真故障,S2)",
})
CallbackPushTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "go_sms_callback_push_total",
Help: "回调推送结果(type: report/uplink × result: sent/retry/dead,S4·C1)",
}, []string{"type", "result"})
CallbackBlacklistTotal = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_callback_blacklist_total",
Help: "上行自动拉黑新增数(幂等冲突不计,S4·C2)",
})
ArchiveRows = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_archive_rows_total",
Help: "明细归档迁出 MySQL→ClickHouse 行数(按日校验通过后计,S4·D3)",
})
ArchiveErrors = prometheus.NewCounter(prometheus.CounterOpts{
Name: "go_sms_archive_errors_total",
Help: "归档链路失败计数(下轮自愈,持续增长=真故障,S4·D3)",
})
registry.MustRegister(
HTTPRequestsTotal, HTTPRequestDuration, HTTPRequestsInFlight,
SMSSubmitTotal, SMSReportTotal, KafkaProducerMessages,
BalanceReconcileDiff, BalanceMode, BalanceFallbackTotal, BalanceFlowlogPending,
KafkaLag,
LongSmsMergedTotal, LongSmsDegradeTotal, LongSmsDupFragTotal,
LongSmsExpireFailTotal,
ReportPendingTotal, ReportRecheckTotal, ReportRequeueTotal,
SettleAggregateRows, SettleAggregateErrors,
CallbackPushTotal, CallbackBlacklistTotal,
ArchiveRows, ArchiveErrors,
)
// Go 运行时与进程指标
registry.MustRegister(
collectors.NewGoCollector(),
collectors.NewProcessCollector(collectors.ProcessCollectorOpts{}),
)
}
// Registry 返回指标注册中心(测试或自托管采集用)。
func Registry() *prometheus.Registry { return registry }
// Handler /metrics 输出 handler(app 装配:gin.WrapH(metrics.Handler()))。
func Handler() http.Handler {
return promhttp.HandlerFor(registry, promhttp.HandlerOpts{EnableOpenMetrics: true})
}