=== 现象 ===
上一提交 b49677f 部署后平台直接 CrashLoopBackOff(5 次重启):
Initializing database...
panic: duplicate metrics collector registration attempted
lokistack/pkg/metrics.Register.func1() metrics.go:125
=== 根因 ===
prometheus/client_golang 在 init() 里已经把 GoCollector 与 ProcessCollector
注册进 **默认 registry**(prometheus.DefaultRegisterer)。我又对同一默认 registry
调用 MustRegister 注册了同一批 collector → 重复注册 → 运行时 panic。
致命点:go build / go vet / 普通单测**全都发现不了** ——
MustRegister 的重复检查只在运行期执行,只有真的调用 Register() 才会炸。
所以整个 P0 打磨在测试阶段是"全绿"的,一上线就挂。
=== 修法 ===
不再碰默认 registry,改用自有实例:
var registry = prometheus.NewRegistry() // 独立,不受 init() 影响
registry.MustRegister(...) // 注册全部指标
promhttp.HandlerFor(registry, ...) // 抓取读自有 registry
好处:完全掌控注册内容,未来再加 collector 不会与库的默认行为冲突。
=== 回归测试(这才是关键收获)===
新增 backend/pkg/metrics/metrics_test.go:
- TestRegisterIsIdempotent : 连续调用 Register 两次,必须不 panic
- TestHandlerExposesCustomMetrics : 断言抓取输出含 lokistack_* 与 go_goroutines
- TestRegisterConcurrent : 20 个 goroutine 并发调用不 panic
这三条在修复前必然失败,修复后全绿。以后再改指标定义会被它们拦住。
镜像:lghost/lokistack:latest + v0.1 = 2f8f3c85643e
185 lines
5.6 KiB
Go
185 lines
5.6 KiB
Go
// Package metrics 定义 LokiStack 自身的 Prometheus 指标体系。
|
||
//
|
||
// 为什么单独成包:这些指标要同时被 repository(DB 落盘耗时)和
|
||
// services(业务计数)以及 middleware(HTTP)使用,
|
||
// 而 services 依赖 repository —— 指标定义必须放在最底层,否则循环依赖。
|
||
package metrics
|
||
|
||
import (
|
||
"net/http"
|
||
"runtime"
|
||
"strconv"
|
||
"sync"
|
||
|
||
"github.com/prometheus/client_golang/prometheus"
|
||
"github.com/prometheus/client_golang/prometheus/collectors"
|
||
"github.com/prometheus/client_golang/prometheus/promhttp"
|
||
)
|
||
|
||
var (
|
||
// HTTP 层
|
||
HTTPRequestsTotal = prometheus.NewCounterVec(
|
||
prometheus.CounterOpts{
|
||
Name: "lokistack_http_requests_total",
|
||
Help: "HTTP 请求总数(按方法/路由模板/状态码)",
|
||
},
|
||
[]string{"method", "path", "status"},
|
||
)
|
||
HTTPRequestDuration = prometheus.NewHistogramVec(
|
||
prometheus.HistogramOpts{
|
||
Name: "lokistack_http_request_duration_seconds",
|
||
Help: "HTTP 请求耗时分布",
|
||
Buckets: []float64{0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 30},
|
||
},
|
||
[]string{"method", "path"},
|
||
)
|
||
HTTPInFlight = prometheus.NewGauge(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_http_in_flight_requests",
|
||
Help: "当前正在处理的请求数",
|
||
},
|
||
)
|
||
|
||
// 存储层(JSON DB 是平台最大风险点:全量读内存、全量写盘)
|
||
DBSizeBytes = prometheus.NewGauge(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_db_size_bytes",
|
||
Help: "JSON 数据库文件大小",
|
||
},
|
||
)
|
||
DBSaveDuration = prometheus.NewHistogram(
|
||
prometheus.HistogramOpts{
|
||
Name: "lokistack_db_save_duration_seconds",
|
||
Help: "数据库落盘耗时",
|
||
Buckets: []float64{0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1, 3, 5, 10},
|
||
},
|
||
)
|
||
DBSaveTotal = prometheus.NewCounterVec(
|
||
prometheus.CounterOpts{
|
||
Name: "lokistack_db_save_total",
|
||
Help: "数据库保存次数(按结果)",
|
||
},
|
||
[]string{"result"},
|
||
)
|
||
DBRecordsTotal = prometheus.NewGaugeVec(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_db_records",
|
||
Help: "各集合的记录数",
|
||
},
|
||
[]string{"collection"},
|
||
)
|
||
|
||
// 业务层
|
||
DeploymentsTotal = prometheus.NewGaugeVec(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_deployments",
|
||
Help: "Loki 部署数(按状态/运行模式)",
|
||
},
|
||
[]string{"status", "mode"},
|
||
)
|
||
CollectorsTotal = prometheus.NewGaugeVec(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_collectors",
|
||
Help: "采集器数(按状态/类型/托管方式)",
|
||
},
|
||
[]string{"status", "type", "run_mode"},
|
||
)
|
||
|
||
// 后台任务层
|
||
AlertEvaluationsTotal = prometheus.NewCounterVec(
|
||
prometheus.CounterOpts{
|
||
Name: "lokistack_alert_evaluations_total",
|
||
Help: "告警规则评估次数(按结果)",
|
||
},
|
||
[]string{"result"},
|
||
)
|
||
AlertEvaluationDuration = prometheus.NewHistogram(
|
||
prometheus.HistogramOpts{
|
||
Name: "lokistack_alert_evaluation_duration_seconds",
|
||
Help: "单轮告警评估耗时",
|
||
Buckets: []float64{0.01, 0.05, 0.1, 0.5, 1, 3, 5, 10, 30},
|
||
},
|
||
)
|
||
LokiQueriesTotal = prometheus.NewCounterVec(
|
||
prometheus.CounterOpts{
|
||
Name: "lokistack_loki_queries_total",
|
||
Help: "对后端 Loki 的查询次数(按结果)",
|
||
},
|
||
[]string{"result"},
|
||
)
|
||
|
||
BuildInfo = prometheus.NewGaugeVec(
|
||
prometheus.GaugeOpts{
|
||
Name: "lokistack_build_info",
|
||
Help: "构建信息(恒为 1,标签里带版本与 Go 版本)",
|
||
},
|
||
[]string{"version", "go_version"},
|
||
)
|
||
|
||
// registry 用独立实例,不用 prometheus.DefaultRegisterer。
|
||
// 原因:client_golang 在 init() 里已经把 GoCollector / ProcessCollector
|
||
// 注册进默认 registry,再用 MustRegister 注册同一批 collector 会
|
||
// **运行时 panic: duplicate metrics collector registration attempted**
|
||
// (编译期和单测都发现不了,只在进程启动时炸)。
|
||
registry = prometheus.NewRegistry()
|
||
|
||
once sync.Once
|
||
)
|
||
|
||
// Register 注册全部指标(幂等)。version 一般传 main.version。
|
||
func Register(version string) {
|
||
once.Do(func() {
|
||
registry.MustRegister(
|
||
HTTPRequestsTotal, HTTPRequestDuration, HTTPInFlight,
|
||
DBSizeBytes, DBSaveDuration, DBSaveTotal, DBRecordsTotal,
|
||
DeploymentsTotal, CollectorsTotal,
|
||
AlertEvaluationsTotal, AlertEvaluationDuration, LokiQueriesTotal,
|
||
BuildInfo,
|
||
collectors.NewGoCollector(),
|
||
collectors.NewProcessCollector(collectors.ProcessCollectorOpts{}),
|
||
)
|
||
BuildInfo.WithLabelValues(version, runtime.Version()).Set(1)
|
||
})
|
||
}
|
||
|
||
// Handler 返回 Prometheus 抓取端点(读自有 registry,不含默认 registry 的重复项)。
|
||
func Handler() http.Handler {
|
||
return promhttp.HandlerFor(registry, promhttp.HandlerOpts{})
|
||
}
|
||
|
||
// ── 埋点辅助(各层直接调) ──
|
||
|
||
// ObserveHTTP 记录一次 HTTP 请求。pathLabel 必须是路由模板,不能用原始 URL。
|
||
func ObserveHTTP(method, pathLabel string, status int, seconds float64) {
|
||
HTTPRequestsTotal.WithLabelValues(method, pathLabel, strconv.Itoa(status)).Inc()
|
||
HTTPRequestDuration.WithLabelValues(method, pathLabel).Observe(seconds)
|
||
}
|
||
|
||
// IncInFlight / DecInFlight 在途请求计数。
|
||
func IncInFlight() { HTTPInFlight.Inc() }
|
||
func DecInFlight() { HTTPInFlight.Dec() }
|
||
|
||
// ObserveDBSave 记录一次落盘结果。
|
||
func ObserveDBSave(seconds float64, ok bool) {
|
||
DBSaveDuration.Observe(seconds)
|
||
DBSaveTotal.WithLabelValues(resultLabel(ok)).Inc()
|
||
}
|
||
|
||
// ObserveAlertEvaluation 记录一轮告警评估。
|
||
func ObserveAlertEvaluation(seconds float64, ok bool) {
|
||
AlertEvaluationDuration.Observe(seconds)
|
||
AlertEvaluationsTotal.WithLabelValues(resultLabel(ok)).Inc()
|
||
}
|
||
|
||
// ObserveLokiQuery 记录一次后端 Loki 查询。
|
||
func ObserveLokiQuery(ok bool) {
|
||
LokiQueriesTotal.WithLabelValues(resultLabel(ok)).Inc()
|
||
}
|
||
|
||
func resultLabel(ok bool) string {
|
||
if ok {
|
||
return "success"
|
||
}
|
||
return "error"
|
||
}
|