diff --git a/backend/pkg/metrics/metrics.go b/backend/pkg/metrics/metrics.go index 0c0f7f7..c2cd44a 100644 --- a/backend/pkg/metrics/metrics.go +++ b/backend/pkg/metrics/metrics.go @@ -116,13 +116,20 @@ var ( []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() { - prometheus.MustRegister( + registry.MustRegister( HTTPRequestsTotal, HTTPRequestDuration, HTTPInFlight, DBSizeBytes, DBSaveDuration, DBSaveTotal, DBRecordsTotal, DeploymentsTotal, CollectorsTotal, @@ -135,8 +142,10 @@ func Register(version string) { }) } -// Handler 返回 Prometheus 抓取端点。 -func Handler() http.Handler { return promhttp.Handler() } +// 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) ── diff --git a/backend/pkg/metrics/metrics_test.go b/backend/pkg/metrics/metrics_test.go new file mode 100644 index 0000000..7a76689 --- /dev/null +++ b/backend/pkg/metrics/metrics_test.go @@ -0,0 +1,67 @@ +package metrics + +import ( + "io" + "net/http/httptest" + "strings" + "sync" + "testing" +) + +// 回归:Register 必须幂等。 +// 线上事故:client_golang 在 init() 里已把 GoCollector / ProcessCollector +// 注册进默认 registry,我们又 MustRegister 到默认 registry → 进程启动即 +// panic: duplicate metrics collector registration attempted。 +// 这个 bug 编译期和普通单测都发现不了(只有真的调用 Register 才炸), +// 所以这里必须实际调用两次。 +func TestRegisterIsIdempotent(t *testing.T) { + defer func() { + if r := recover(); r != nil { + t.Fatalf("Register 触发 panic(重复注册):%v", r) + } + }() + Register("test") + Register("test") // 第二次必须安全返回 +} + +// 抓取端点必须能输出我们自己定义的指标 +func TestHandlerExposesCustomMetrics(t *testing.T) { + Register("test") + ObserveHTTP("GET", "/api/v1/deployments", 200, 0.01) + ObserveDBSave(0.002, true) + ObserveAlertEvaluation(0.3, true) + ObserveLokiQuery(true) + DBRecordsTotal.WithLabelValues("deployments").Set(3) + + rec := httptest.NewRecorder() + req := httptest.NewRequest("GET", "/metrics", nil) + Handler().ServeHTTP(rec, req) + + body, _ := io.ReadAll(rec.Body) + out := string(body) + for _, want := range []string{ + "lokistack_http_requests_total", + "lokistack_db_save_duration_seconds", + "lokistack_alert_evaluations_total", + "lokistack_loki_queries_total", + "lokistack_db_records", + "go_goroutines", // GoCollector 生效 + } { + if !strings.Contains(out, want) { + t.Fatalf("抓取结果缺少指标 %q", want) + } + } +} + +// 并发调用不能 panic(sync.Once 保护) +func TestRegisterConcurrent(t *testing.T) { + var wg sync.WaitGroup + for i := 0; i < 20; i++ { + wg.Add(1) + go func() { + defer wg.Done() + Register("test") + }() + } + wg.Wait() +}