Files
telemetry/doc.go
Rene Nochebuena b583af1973 feat(telemetry): initial implementation — OpenTelemetry traces, metrics, logs (v1.0.0)
Introduces code.nochebuena.dev/einherjar/telemetry — the observability bootstrap
starter for the Einherjar framework. Absorbs the telemetry package from micro-lib,
migrating from OpenCensus to OpenTelemetry SDK v1.42.

Bootstrap functions (not lifecycle.Component — telemetry must be initialized before
the launcher starts, and its shutdown must run after all components stop):
- New(ctx, cfg) (func(context.Context) error, error) — production mode; exports
  traces, metrics, and logs via OTLP over gRPC to the configured endpoint;
  returns a shutdown function to be deferred in main()
- NewConsole(ctx, logger, cfg) (func(context.Context) error, error) — development
  mode; writes structured telemetry to the provided logging.Logger; no network
  dependency; suitable for local development and CI

Config (EINHERJAR_OTEL_* env vars):
  ServiceName(required), ServiceVersion(unknown), Environment(development),
  OTLPEndpoint(required for New), OTLPInsecure(false)

ConsoleConfig (EINHERJAR_OTEL_* env vars):
  ServiceName(required), ServiceVersion(unknown), Environment(development)

- identifiable.go: package-level Module variable (observability.Identifiable) for version
  identification — telemetry bootstraps before the launcher; not registered as a lifecycle component
2026-05-29 16:09:54 +00:00

81 lines
2.8 KiB
Go

// Package telemetry bootstraps the OpenTelemetry SDK for Einherjar applications.
//
// # Overview
//
// There are two bootstrap functions — one for production (OTLP over gRPC) and one
// for local development (structured log output). Both set the three OTel global
// providers so that all starters using otel.Tracer / otel.Meter / global.Logger
// auto-instrument without any code changes.
//
// This package is app-only: import it only from main packages. Never import it
// from a starter or library — starters use only the OTel API, which is a zero-cost
// no-op until a real SDK is wired up here.
//
// # Production: OTLP over gRPC
//
// [New] connects to a Grafana Alloy (or any OTLP-compatible) collector and
// exports traces → Tempo, metrics → Mimir, and logs → Loki.
//
// func main() {
// ctx := context.Background()
//
// shutdown, err := telemetry.New(ctx, telemetry.Config{
// ServiceName: "order-service",
// ServiceVersion: "1.4.2",
// Environment: "production",
// OTLPEndpoint: "alloy:4317",
// OTLPInsecure: false,
// })
// if err != nil {
// log.Fatalf("telemetry: %v", err)
// }
// defer shutdown(ctx)
//
// // Wire into the launcher so shutdown runs before process exit:
// // launcher.BeforeStop(func() error { return shutdown(ctx) })
// }
//
// # Local Development: console mode
//
// [NewConsole] routes all three signals through a [logging.Logger] as structured
// log lines. No collector is required — spans, metrics, and OTel log records appear
// inline with your application logs.
//
// func main() {
// ctx := context.Background()
// logger := logz.New(logz.Config{})
//
// shutdown, err := telemetry.NewConsole(ctx, logger, telemetry.ConsoleConfig{
// ServiceName: "order-service",
// })
// if err != nil {
// log.Fatalf("telemetry: %v", err)
// }
// defer shutdown(ctx)
// }
//
// # Avoiding the slog feedback loop
//
// logz is backed by slog. The OTel ecosystem provides a slog bridge
// (go.opentelemetry.io/contrib/bridges/otelslog) that forwards slog records into
// the OTel log API. Do NOT use that bridge together with [NewConsole].
//
// The loop is:
//
// slog.Info("msg")
// → OTel log API (via slog bridge)
// → logLogExporter.Export()
// → logger.Info("otel: log", ...) ← this is slog again
// → OTel log API (via slog bridge)
// → ... ∞
//
// The slog bridge is safe with [New] because the OTLP exporter sends records over
// the network — it never calls back into slog. The loop only occurs with [NewConsole]
// because its log exporter writes back to the same logger that feeds it.
//
// Rule of thumb:
// - [New] + slog bridge: safe ✓
// - [NewConsole] + slog bridge: feedback loop ✗
// - [NewConsole] without slog bridge: safe ✓
package telemetry