docs: add v0.5.2 release notes

Add changelog entries for recent commits:
- Health package: Kubernetes health probes
- Logger package: runtime level control and fixes
- Database package: config file override support

CHANGELOG.md

@@ -1,3 +1,22 @@
+# Release Notes - v0.5.2
+
+## Health Package
+- **Kubernetes-native health probes** - Added dedicated handlers for liveness (`/healthz`), readiness (`/readyz`), and startup (`/startupz`) probes
+- **Flexible configuration options** - New `WithLivenessHandler`, `WithReadinessHandler`, `WithStartupHandler`, and `WithServiceName` options
+- **JSON response formats** - Structured probe responses with service identification
+- **Backward compatibility** - Maintains existing `/__health` and `/` endpoints
+
+## Logger Package
+- **Runtime log level control** - Independent level management for stderr and OTLP loggers via `SetLevel()` and `SetOTLPLevel()`
+- **Environment variable support** - Configure levels with `LOG_LEVEL` and `OTLP_LOG_LEVEL` env vars
+- **String parsing utility** - New `ParseLevel()` function for converting string levels to `slog.Level`
+- **Buffering exporter fix** - Resolved "unlock of unlocked mutex" panic in `bufferingExporter`
+- **Initialization redesign** - Eliminated race conditions in TLS/tracing setup retry logic
+
+## Database Package
+- **Configuration file override** - Added `DATABASE_CONFIG_FILE` environment variable to specify custom database configuration file paths
+- **Flexible path configuration** - Override default `["database.yaml", "/vault/secrets/database.yaml"]` search paths when needed
+
 # Release Notes - v0.5.1
 
 ## Observability Enhancements

database/config.go

@@ -1,6 +1,7 @@
 package database
 
 import (
+	"os"
 	"time"
 
 	"github.com/prometheus/client_golang/prometheus"
@@ -36,10 +37,20 @@ type ConfigOptions struct {
 	ConnMaxLifetime time.Duration
 }
 
+// getConfigFiles returns the list of config files to search for database configuration.
+// If DATABASE_CONFIG_FILE environment variable is set, it returns that single file.
+// Otherwise, it returns the default paths.
+func getConfigFiles() []string {
+	if configFile := os.Getenv("DATABASE_CONFIG_FILE"); configFile != "" {
+		return []string{configFile}
+	}
+	return []string{"database.yaml", "/vault/secrets/database.yaml"}
+}
+
 // DefaultConfigOptions returns the standard configuration options used by API package
 func DefaultConfigOptions() ConfigOptions {
 	return ConfigOptions{
-		ConfigFiles:          []string{"database.yaml", "/vault/secrets/database.yaml"},
+		ConfigFiles:          getConfigFiles(),
 		EnablePoolMonitoring: true,
 		PrometheusRegisterer: prometheus.DefaultRegisterer,
 		MaxOpenConns:         25,
@@ -51,7 +62,7 @@ func DefaultConfigOptions() ConfigOptions {
 // MonitorConfigOptions returns configuration options optimized for Monitor package
 func MonitorConfigOptions() ConfigOptions {
 	return ConfigOptions{
-		ConfigFiles:          []string{"database.yaml", "/vault/secrets/database.yaml"},
+		ConfigFiles:          getConfigFiles(),
 		EnablePoolMonitoring: false, // Monitor doesn't need metrics
 		PrometheusRegisterer: nil,   // No Prometheus dependency
 		MaxOpenConns:         10,

health/health_server.go

@@ -1,13 +1,71 @@
 // Package health provides a standalone HTTP server for health checks.
 //
-// This package implements a simple health check server that can be used
-// to expose health status endpoints for monitoring and load balancing.
-// It supports custom health check handlers and provides structured logging
-// with graceful shutdown capabilities.
+// This package implements a flexible health check server that supports
+// different handlers for Kubernetes probe types (liveness, readiness, startup).
+// It provides structured logging, graceful shutdown, and standard HTTP endpoints
+// for monitoring and load balancing.
+//
+// # Kubernetes Probe Types
+//
+// Liveness Probe: Detects when a container is "dead" and needs restarting.
+// Should be a lightweight check that verifies the process is still running
+// and not in an unrecoverable state.
+//
+// Readiness Probe: Determines when a container is ready to accept traffic.
+// Controls which Pods are used as backends for Services. Should verify
+// the application can handle requests properly.
+//
+// Startup Probe: Verifies when a container application has successfully started.
+// Delays liveness and readiness probes until startup succeeds. Useful for
+// slow-starting applications.
+//
+// # Usage Examples
+//
+// Basic usage with a single handler for all probes:
+//
+//	srv := health.NewServer(myHealthHandler)
+//	srv.Listen(ctx, 9091)
+//
+// Advanced usage with separate handlers for each probe type:
+//
+//	srv := health.NewServer(nil,
+//		health.WithLivenessHandler(func(w http.ResponseWriter, r *http.Request) {
+//			// Simple alive check
+//			w.WriteHeader(http.StatusOK)
+//		}),
+//		health.WithReadinessHandler(func(w http.ResponseWriter, r *http.Request) {
+//			// Check if ready to serve traffic
+//			if err := checkDatabase(); err != nil {
+//				w.WriteHeader(http.StatusServiceUnavailable)
+//				return
+//			}
+//			w.WriteHeader(http.StatusOK)
+//		}),
+//		health.WithStartupHandler(func(w http.ResponseWriter, r *http.Request) {
+//			// Check if startup is complete
+//			if !applicationReady() {
+//				w.WriteHeader(http.StatusServiceUnavailable)
+//				return
+//			}
+//			w.WriteHeader(http.StatusOK)
+//		}),
+//		health.WithServiceName("my-service"),
+//	)
+//	srv.Listen(ctx, 9091)
+//
+// # Standard Endpoints
+//
+// The server exposes these endpoints:
+//   - /healthz - liveness probe (or general health if no specific handler)
+//   - /readyz - readiness probe (or general health if no specific handler)
+//   - /startupz - startup probe (or general health if no specific handler)
+//   - /__health - general health endpoint (backward compatibility)
+//   - / - general health endpoint (root path)
 package health
 
 import (
 	"context"
+	"encoding/json"
 	"log/slog"
 	"net/http"
 	"strconv"
@@ -21,23 +79,74 @@ import (
 // It runs separately from the main application server to ensure health
 // checks remain available even if the main server is experiencing issues.
 //
-// The server includes built-in timeouts, graceful shutdown, and structured
-// logging for monitoring and debugging health check behavior.
+// The server supports separate handlers for different Kubernetes probe types
+// (liveness, readiness, startup) and includes built-in timeouts, graceful
+// shutdown, and structured logging.
 type Server struct {
 	log              *slog.Logger
-	healthFn http.HandlerFunc
+	livenessHandler  http.HandlerFunc
+	readinessHandler http.HandlerFunc
+	startupHandler   http.HandlerFunc
+	generalHandler   http.HandlerFunc // fallback for /__health and / paths
+	serviceName      string
 }
 
-// NewServer creates a new health check server with the specified health handler.
-// If healthFn is nil, a default handler that returns HTTP 200 "ok" is used.
-func NewServer(healthFn http.HandlerFunc) *Server {
+// Option represents a configuration option for the health server.
+type Option func(*Server)
+
+// WithLivenessHandler sets a specific handler for the /healthz endpoint.
+// Liveness probes determine if a container should be restarted.
+func WithLivenessHandler(handler http.HandlerFunc) Option {
+	return func(s *Server) {
+		s.livenessHandler = handler
+	}
+}
+
+// WithReadinessHandler sets a specific handler for the /readyz endpoint.
+// Readiness probes determine if a container can receive traffic.
+func WithReadinessHandler(handler http.HandlerFunc) Option {
+	return func(s *Server) {
+		s.readinessHandler = handler
+	}
+}
+
+// WithStartupHandler sets a specific handler for the /startupz endpoint.
+// Startup probes determine if a container has finished initializing.
+func WithStartupHandler(handler http.HandlerFunc) Option {
+	return func(s *Server) {
+		s.startupHandler = handler
+	}
+}
+
+// WithServiceName sets the service name for JSON responses and logging.
+func WithServiceName(serviceName string) Option {
+	return func(s *Server) {
+		s.serviceName = serviceName
+	}
+}
+
+
+// NewServer creates a new health check server with optional probe-specific handlers.
+//
+// If healthFn is provided, it will be used as a fallback for any probe endpoints
+// that don't have specific handlers configured. If healthFn is nil, a default
+// handler that returns HTTP 200 "ok" is used as the fallback.
+//
+// Use the With* option functions to configure specific handlers for different
+// probe types (liveness, readiness, startup).
+func NewServer(healthFn http.HandlerFunc, opts ...Option) *Server {
 	if healthFn == nil {
 		healthFn = basicHealth
 	}
 	srv := &Server{
 		log:            logger.Setup(),
-		healthFn: healthFn,
+		generalHandler: healthFn,
 	}
+
+	for _, opt := range opts {
+		opt(srv)
+	}
+
 	return srv
 }
 
@@ -47,13 +156,27 @@ func (srv *Server) SetLogger(log *slog.Logger) {
 }
 
 // Listen starts the health server on the specified port and blocks until ctx is cancelled.
-// The server exposes the health handler at "/__health" with graceful shutdown support.
+// The server exposes health check endpoints with graceful shutdown support.
+//
+// Standard endpoints exposed:
+//   - /healthz - liveness probe (uses livenessHandler or falls back to generalHandler)
+//   - /readyz - readiness probe (uses readinessHandler or falls back to generalHandler)
+//   - /startupz - startup probe (uses startupHandler or falls back to generalHandler)
+//   - /__health - general health endpoint (uses generalHandler)
+//   - / - root health endpoint (uses generalHandler)
 func (srv *Server) Listen(ctx context.Context, port int) error {
 	srv.log.Info("starting health listener", "port", port)
 
 	serveMux := http.NewServeMux()
 
-	serveMux.HandleFunc("/__health", srv.healthFn)
+	// Register probe-specific handlers
+	serveMux.HandleFunc("/healthz", srv.createProbeHandler("liveness"))
+	serveMux.HandleFunc("/readyz", srv.createProbeHandler("readiness"))
+	serveMux.HandleFunc("/startupz", srv.createProbeHandler("startup"))
+
+	// Register general health endpoints for backward compatibility
+	serveMux.HandleFunc("/__health", srv.createGeneralHandler())
+	serveMux.HandleFunc("/", srv.createGeneralHandler())
 
 	hsrv := &http.Server{
 		Addr:         ":" + strconv.Itoa(port),
@@ -89,6 +212,121 @@ func (srv *Server) Listen(ctx context.Context, port int) error {
 	return g.Wait()
 }
 
+// createProbeHandler creates a handler for a specific probe type that provides
+// appropriate JSON responses and falls back to the general handler if no specific
+// handler is configured.
+func (srv *Server) createProbeHandler(probeType string) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		var handler http.HandlerFunc
+
+		// Select the appropriate handler
+		switch probeType {
+		case "liveness":
+			handler = srv.livenessHandler
+		case "readiness":
+			handler = srv.readinessHandler
+		case "startup":
+			handler = srv.startupHandler
+		}
+
+		// Fall back to general handler if no specific handler is configured
+		if handler == nil {
+			handler = srv.generalHandler
+		}
+
+		// Create a response recorder to capture the handler's status code
+		recorder := &statusRecorder{ResponseWriter: w, statusCode: 200}
+		handler(recorder, r)
+
+		// If the handler already wrote a response, we're done
+		if recorder.written {
+			return
+		}
+
+		// Otherwise, provide a standard JSON response based on the status code
+		w.Header().Set("Content-Type", "application/json")
+
+		if recorder.statusCode >= 400 {
+			// Handler indicated unhealthy
+			switch probeType {
+			case "liveness":
+				json.NewEncoder(w).Encode(map[string]string{"status": "unhealthy"})
+			case "readiness":
+				json.NewEncoder(w).Encode(map[string]bool{"ready": false})
+			case "startup":
+				json.NewEncoder(w).Encode(map[string]bool{"started": false})
+			}
+		} else {
+			// Handler indicated healthy
+			switch probeType {
+			case "liveness":
+				json.NewEncoder(w).Encode(map[string]string{"status": "alive"})
+			case "readiness":
+				json.NewEncoder(w).Encode(map[string]bool{"ready": true})
+			case "startup":
+				json.NewEncoder(w).Encode(map[string]bool{"started": true})
+			}
+		}
+	}
+}
+
+// createGeneralHandler creates a handler for general health endpoints that provides
+// comprehensive health information.
+func (srv *Server) createGeneralHandler() http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		// Create a response recorder to capture the handler's status code
+		// Use a buffer to prevent the handler from writing to the actual response
+		recorder := &statusRecorder{ResponseWriter: &discardWriter{}, statusCode: 200}
+		srv.generalHandler(recorder, r)
+
+		// Always provide a comprehensive JSON response for general endpoints
+		w.Header().Set("Content-Type", "application/json")
+		w.WriteHeader(recorder.statusCode)
+
+		response := map[string]interface{}{
+			"status": map[bool]string{true: "healthy", false: "unhealthy"}[recorder.statusCode < 400],
+		}
+
+		if srv.serviceName != "" {
+			response["service"] = srv.serviceName
+		}
+
+		json.NewEncoder(w).Encode(response)
+	}
+}
+
+// statusRecorder captures the response status code from handlers while allowing
+// them to write their own response content if needed.
+type statusRecorder struct {
+	http.ResponseWriter
+	statusCode int
+	written    bool
+}
+
+func (r *statusRecorder) WriteHeader(code int) {
+	r.statusCode = code
+	r.ResponseWriter.WriteHeader(code)
+}
+
+func (r *statusRecorder) Write(data []byte) (int, error) {
+	r.written = true
+	return r.ResponseWriter.Write(data)
+}
+
+// discardWriter implements http.ResponseWriter but discards all writes.
+// Used to capture status codes without writing response content.
+type discardWriter struct{}
+
+func (d *discardWriter) Header() http.Header {
+	return make(http.Header)
+}
+
+func (d *discardWriter) Write([]byte) (int, error) {
+	return 0, nil
+}
+
+func (d *discardWriter) WriteHeader(int) {}
+
 // HealthCheckListener runs a simple HTTP server on the specified port for health check probes.
 func HealthCheckListener(ctx context.Context, port int, log *slog.Logger) error {
 	srv := NewServer(nil)

health/health_test.go

@@ -1,13 +1,14 @@
 package health
 
 import (
+	"fmt"
 	"io"
 	"net/http"
 	"net/http/httptest"
 	"testing"
 )
 
-func TestHealthHandler(t *testing.T) {
+func TestBasicHealthHandler(t *testing.T) {
 	req := httptest.NewRequest(http.MethodGet, "/__health", nil)
 	w := httptest.NewRecorder()
 
@@ -24,3 +25,129 @@ func TestHealthHandler(t *testing.T) {
 		t.Errorf("expected ok got %q", string(data))
 	}
 }
+
+func TestProbeHandlers(t *testing.T) {
+	// Test with separate handlers for each probe type
+	srv := NewServer(nil,
+		WithLivenessHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusOK)
+		}),
+		WithReadinessHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusOK)
+		}),
+		WithStartupHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusOK)
+		}),
+		WithServiceName("test-service"),
+	)
+
+	tests := []struct {
+		handler        http.HandlerFunc
+		expectedStatus int
+		expectedBody   string
+	}{
+		{srv.createProbeHandler("liveness"), 200, `{"status":"alive"}`},
+		{srv.createProbeHandler("readiness"), 200, `{"ready":true}`},
+		{srv.createProbeHandler("startup"), 200, `{"started":true}`},
+		{srv.createGeneralHandler(), 200, `{"service":"test-service","status":"healthy"}`},
+	}
+
+	for i, tt := range tests {
+		t.Run(fmt.Sprintf("test_%d", i), func(t *testing.T) {
+			req := httptest.NewRequest(http.MethodGet, "/", nil)
+			w := httptest.NewRecorder()
+
+			tt.handler(w, req)
+
+			if w.Code != tt.expectedStatus {
+				t.Errorf("expected status %d, got %d", tt.expectedStatus, w.Code)
+			}
+
+			body := w.Body.String()
+			if body != tt.expectedBody+"\n" { // json.Encoder adds newline
+				t.Errorf("expected body %q, got %q", tt.expectedBody, body)
+			}
+		})
+	}
+}
+
+func TestProbeHandlerFallback(t *testing.T) {
+	// Test fallback to general handler when no specific handler is configured
+	generalHandler := func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusOK)
+	}
+
+	srv := NewServer(generalHandler, WithServiceName("test-service"))
+
+	tests := []struct {
+		handler        http.HandlerFunc
+		expectedStatus int
+		expectedBody   string
+	}{
+		{srv.createProbeHandler("liveness"), 200, `{"status":"alive"}`},
+		{srv.createProbeHandler("readiness"), 200, `{"ready":true}`},
+		{srv.createProbeHandler("startup"), 200, `{"started":true}`},
+	}
+
+	for i, tt := range tests {
+		t.Run(fmt.Sprintf("fallback_%d", i), func(t *testing.T) {
+			req := httptest.NewRequest(http.MethodGet, "/", nil)
+			w := httptest.NewRecorder()
+
+			tt.handler(w, req)
+
+			if w.Code != tt.expectedStatus {
+				t.Errorf("expected status %d, got %d", tt.expectedStatus, w.Code)
+			}
+
+			body := w.Body.String()
+			if body != tt.expectedBody+"\n" { // json.Encoder adds newline
+				t.Errorf("expected body %q, got %q", tt.expectedBody, body)
+			}
+		})
+	}
+}
+
+func TestUnhealthyProbeHandlers(t *testing.T) {
+	// Test with handlers that return unhealthy status
+	srv := NewServer(nil,
+		WithLivenessHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusServiceUnavailable)
+		}),
+		WithReadinessHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusServiceUnavailable)
+		}),
+		WithStartupHandler(func(w http.ResponseWriter, r *http.Request) {
+			w.WriteHeader(http.StatusServiceUnavailable)
+		}),
+		WithServiceName("test-service"),
+	)
+
+	tests := []struct {
+		handler        http.HandlerFunc
+		expectedStatus int
+		expectedBody   string
+	}{
+		{srv.createProbeHandler("liveness"), 503, `{"status":"unhealthy"}`},
+		{srv.createProbeHandler("readiness"), 503, `{"ready":false}`},
+		{srv.createProbeHandler("startup"), 503, `{"started":false}`},
+	}
+
+	for i, tt := range tests {
+		t.Run(fmt.Sprintf("unhealthy_%d", i), func(t *testing.T) {
+			req := httptest.NewRequest(http.MethodGet, "/", nil)
+			w := httptest.NewRecorder()
+
+			tt.handler(w, req)
+
+			if w.Code != tt.expectedStatus {
+				t.Errorf("expected status %d, got %d", tt.expectedStatus, w.Code)
+			}
+
+			body := w.Body.String()
+			if body != tt.expectedBody+"\n" { // json.Encoder adds newline
+				t.Errorf("expected body %q, got %q", tt.expectedBody, body)
+			}
+		})
+	}
+}

logger/buffering_exporter.go

@@ -23,8 +23,7 @@ type bufferingExporter struct {
 	// Real exporter (created when tracing is configured)
 	exporter otellog.Exporter
 
-	// Thread-safe initialization
-	initOnce sync.Once
+	// Thread-safe initialization state (managed only by checkReadiness)
 	initErr error
 
 	// Background checker
@@ -48,13 +47,7 @@ func newBufferingExporter() *bufferingExporter {
 
 // Export implements otellog.Exporter
 func (e *bufferingExporter) Export(ctx context.Context, records []otellog.Record) error {
-	// Try initialization once
-	e.initOnce.Do(func() {
-		e.initErr = e.initialize()
-	})
-
-	// If initialization succeeded, use the exporter
-	if e.initErr == nil {
+	// Check if exporter is ready (initialization handled by checkReadiness goroutine)
 	e.mu.RLock()
 	exporter := e.exporter
 	e.mu.RUnlock()
@@ -62,7 +55,6 @@ func (e *bufferingExporter) Export(ctx context.Context, records []otellog.Record
 	if exporter != nil {
 		return exporter.Export(ctx, records)
 	}
-	}
 
 	// Not ready yet, buffer the records
 	return e.bufferRecords(records)
@@ -117,24 +109,31 @@ func (e *bufferingExporter) bufferRecords(records []otellog.Record) error {
 	return nil
 }
 
-// checkReadiness periodically checks if tracing is configured
+// checkReadiness periodically attempts initialization until successful
 func (e *bufferingExporter) checkReadiness() {
 	defer close(e.checkerDone)
 
-	ticker := time.NewTicker(1 * time.Second) // Reduced frequency since OTLP handles retries
+	ticker := time.NewTicker(1 * time.Second)
 	defer ticker.Stop()
 
 	for {
 		select {
 		case <-ticker.C:
-			// If initialization failed, reset sync.Once to allow retry
-			// The OTLP exporter will handle its own retry logic
-			if e.initErr != nil {
-				e.initOnce = sync.Once{}
-			} else if e.exporter != nil {
+			// Check if we already have a working exporter
+			e.mu.RLock()
+			hasExporter := e.exporter != nil
+			e.mu.RUnlock()
+
+			if hasExporter {
 				return // Exporter ready, checker no longer needed
 			}
 
+			// Try to initialize
+			err := e.initialize()
+			e.mu.Lock()
+			e.initErr = err
+			e.mu.Unlock()
+
 		case <-e.stopChecker:
 			return
 		}
@@ -180,14 +179,21 @@ func (e *bufferingExporter) Shutdown(ctx context.Context) error {
 	// Stop the readiness checker from continuing
 	close(e.stopChecker)
 
-	// Give one final chance for TLS/tracing to become ready before fully shutting down
-	e.initOnce.Do(func() {
-		e.initErr = e.initialize()
-	})
-
 	// Wait for readiness checker goroutine to complete
 	<-e.checkerDone
 
+	// Give one final chance for TLS/tracing to become ready for buffer flushing
+	e.mu.RLock()
+	hasExporter := e.exporter != nil
+	e.mu.RUnlock()
+
+	if !hasExporter {
+		err := e.initialize()
+		e.mu.Lock()
+		e.initErr = err
+		e.mu.Unlock()
+	}
+
 	e.mu.Lock()
 	defer e.mu.Unlock()
 

logger/level_test.go

@@ -0,0 +1,235 @@
+package logger
+
+import (
+	"context"
+	"log/slog"
+	"os"
+	"testing"
+	"time"
+)
+
+func TestParseLevel(t *testing.T) {
+	tests := []struct {
+		name        string
+		input       string
+		expected    slog.Level
+		expectError bool
+	}{
+		{"empty string", "", slog.LevelInfo, false},
+		{"DEBUG upper", "DEBUG", slog.LevelDebug, false},
+		{"debug lower", "debug", slog.LevelDebug, false},
+		{"INFO upper", "INFO", slog.LevelInfo, false},
+		{"info lower", "info", slog.LevelInfo, false},
+		{"WARN upper", "WARN", slog.LevelWarn, false},
+		{"warn lower", "warn", slog.LevelWarn, false},
+		{"ERROR upper", "ERROR", slog.LevelError, false},
+		{"error lower", "error", slog.LevelError, false},
+		{"invalid level", "invalid", slog.LevelInfo, true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			level, err := ParseLevel(tt.input)
+			if tt.expectError {
+				if err == nil {
+					t.Errorf("expected error for input %q, got nil", tt.input)
+				}
+			} else {
+				if err != nil {
+					t.Errorf("unexpected error for input %q: %v", tt.input, err)
+				}
+				if level != tt.expected {
+					t.Errorf("expected level %v for input %q, got %v", tt.expected, tt.input, level)
+				}
+			}
+		})
+	}
+}
+
+func TestSetLevel(t *testing.T) {
+	// Store original level to restore later
+	originalLevel := Level.Level()
+	defer Level.Set(originalLevel)
+
+	SetLevel(slog.LevelDebug)
+	if Level.Level() != slog.LevelDebug {
+		t.Errorf("expected Level to be Debug, got %v", Level.Level())
+	}
+
+	SetLevel(slog.LevelError)
+	if Level.Level() != slog.LevelError {
+		t.Errorf("expected Level to be Error, got %v", Level.Level())
+	}
+}
+
+func TestSetOTLPLevel(t *testing.T) {
+	// Store original level to restore later
+	originalLevel := OTLPLevel.Level()
+	defer OTLPLevel.Set(originalLevel)
+
+	SetOTLPLevel(slog.LevelWarn)
+	if OTLPLevel.Level() != slog.LevelWarn {
+		t.Errorf("expected OTLPLevel to be Warn, got %v", OTLPLevel.Level())
+	}
+
+	SetOTLPLevel(slog.LevelDebug)
+	if OTLPLevel.Level() != slog.LevelDebug {
+		t.Errorf("expected OTLPLevel to be Debug, got %v", OTLPLevel.Level())
+	}
+}
+
+func TestOTLPLevelHandler(t *testing.T) {
+	// Create a mock handler that counts calls
+	callCount := 0
+	mockHandler := &mockHandler{
+		handleFunc: func(ctx context.Context, r slog.Record) error {
+			callCount++
+			return nil
+		},
+	}
+
+	// Set OTLP level to Warn
+	originalLevel := OTLPLevel.Level()
+	defer OTLPLevel.Set(originalLevel)
+	OTLPLevel.Set(slog.LevelWarn)
+
+	// Create OTLP level handler
+	handler := newOTLPLevelHandler(mockHandler)
+
+	ctx := context.Background()
+
+	// Test that Debug and Info are filtered out
+	if handler.Enabled(ctx, slog.LevelDebug) {
+		t.Error("Debug level should be disabled when OTLP level is Warn")
+	}
+	if handler.Enabled(ctx, slog.LevelInfo) {
+		t.Error("Info level should be disabled when OTLP level is Warn")
+	}
+
+	// Test that Warn and Error are enabled
+	if !handler.Enabled(ctx, slog.LevelWarn) {
+		t.Error("Warn level should be enabled when OTLP level is Warn")
+	}
+	if !handler.Enabled(ctx, slog.LevelError) {
+		t.Error("Error level should be enabled when OTLP level is Warn")
+	}
+
+	// Test that Handle respects level filtering
+	now := time.Now()
+	debugRecord := slog.NewRecord(now, slog.LevelDebug, "debug message", 0)
+	warnRecord := slog.NewRecord(now, slog.LevelWarn, "warn message", 0)
+
+	handler.Handle(ctx, debugRecord)
+	if callCount != 0 {
+		t.Error("Debug record should not be passed to underlying handler")
+	}
+
+	handler.Handle(ctx, warnRecord)
+	if callCount != 1 {
+		t.Error("Warn record should be passed to underlying handler")
+	}
+}
+
+func TestEnvironmentVariables(t *testing.T) {
+	tests := []struct {
+		name         string
+		envVar       string
+		envValue     string
+		configPrefix string
+		testFunc     func(t *testing.T)
+	}{
+		{
+			name:     "LOG_LEVEL sets stderr level",
+			envVar:   "LOG_LEVEL",
+			envValue: "ERROR",
+			testFunc: func(t *testing.T) {
+				// Reset the setup state
+				resetLoggerSetup()
+
+				// Call setupStdErrHandler which should read the env var
+				handler := setupStdErrHandler()
+				if handler == nil {
+					t.Fatal("setupStdErrHandler returned nil")
+				}
+
+				if Level.Level() != slog.LevelError {
+					t.Errorf("expected Level to be Error after setting LOG_LEVEL=ERROR, got %v", Level.Level())
+				}
+			},
+		},
+		{
+			name:         "Prefixed LOG_LEVEL",
+			envVar:       "TEST_LOG_LEVEL",
+			envValue:     "DEBUG",
+			configPrefix: "TEST",
+			testFunc: func(t *testing.T) {
+				ConfigPrefix = "TEST"
+				defer func() { ConfigPrefix = "" }()
+
+				resetLoggerSetup()
+				handler := setupStdErrHandler()
+				if handler == nil {
+					t.Fatal("setupStdErrHandler returned nil")
+				}
+
+				if Level.Level() != slog.LevelDebug {
+					t.Errorf("expected Level to be Debug after setting TEST_LOG_LEVEL=DEBUG, got %v", Level.Level())
+				}
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			// Store original env value and level
+			originalEnv := os.Getenv(tt.envVar)
+			originalLevel := Level.Level()
+			defer func() {
+				os.Setenv(tt.envVar, originalEnv)
+				Level.Set(originalLevel)
+			}()
+
+			// Set test environment variable
+			os.Setenv(tt.envVar, tt.envValue)
+
+			// Run the test
+			tt.testFunc(t)
+		})
+	}
+}
+
+// mockHandler is a simple mock implementation of slog.Handler for testing
+type mockHandler struct {
+	handleFunc func(ctx context.Context, r slog.Record) error
+}
+
+func (m *mockHandler) Enabled(ctx context.Context, level slog.Level) bool {
+	return true
+}
+
+func (m *mockHandler) Handle(ctx context.Context, r slog.Record) error {
+	if m.handleFunc != nil {
+		return m.handleFunc(ctx, r)
+	}
+	return nil
+}
+
+func (m *mockHandler) WithAttrs(attrs []slog.Attr) slog.Handler {
+	return m
+}
+
+func (m *mockHandler) WithGroup(name string) slog.Handler {
+	return m
+}
+
+// resetLoggerSetup resets the sync.Once instances for testing
+func resetLoggerSetup() {
+	// Reset package-level variables
+	textLogger = nil
+	otlpLogger = nil
+	multiLogger = nil
+
+	// Note: We can't easily reset sync.Once instances in tests,
+	// but for the specific test we're doing (environment variable parsing)
+	// we can test the setupStdErrHandler function directly
+}

logger/logger.go

@@ -18,12 +18,15 @@
 //   - Context propagation for request-scoped logging
 //
 // Environment variables:
-//   - DEBUG: Enable debug level logging (configurable prefix via ConfigPrefix)
+//   - LOG_LEVEL: Set stderr log level (DEBUG, INFO, WARN, ERROR) (configurable prefix via ConfigPrefix)
+//   - OTLP_LOG_LEVEL: Set OTLP log level independently (configurable prefix via ConfigPrefix)
+//   - DEBUG: Enable debug level logging for backward compatibility (configurable prefix via ConfigPrefix)
 //   - INVOCATION_ID: Systemd detection for timestamp handling
 package logger
 
 import (
 	"context"
+	"fmt"
 	"log"
 	"log/slog"
 	"os"
@@ -43,6 +46,16 @@ import (
 // This enables multiple services to have independent logging configuration.
 var ConfigPrefix = ""
 
+var (
+	// Level controls the log level for the default stderr logger.
+	// Can be changed at runtime to adjust logging verbosity.
+	Level = new(slog.LevelVar) // Info by default
+
+	// OTLPLevel controls the log level for OTLP output.
+	// Can be changed independently from the stderr logger level.
+	OTLPLevel = new(slog.LevelVar) // Info by default
+)
+
 var (
 	textLogger  *slog.Logger
 	otlpLogger  *slog.Logger
@@ -56,21 +69,64 @@ var (
 	mu         sync.Mutex
 )
 
+// SetLevel sets the log level for the default stderr logger.
+// This affects the primary application logger returned by Setup().
+func SetLevel(level slog.Level) {
+	Level.Set(level)
+}
+
+// SetOTLPLevel sets the log level for OTLP output.
+// This affects the logger returned by SetupOLTP() and the OTLP portion of SetupMultiLogger().
+func SetOTLPLevel(level slog.Level) {
+	OTLPLevel.Set(level)
+}
+
+// ParseLevel converts a string log level to slog.Level.
+// Supported levels: "DEBUG", "INFO", "WARN", "ERROR" (case insensitive).
+// Returns an error for unrecognized level strings.
+func ParseLevel(level string) (slog.Level, error) {
+	switch {
+	case level == "":
+		return slog.LevelInfo, nil
+	case level == "DEBUG" || level == "debug":
+		return slog.LevelDebug, nil
+	case level == "INFO" || level == "info":
+		return slog.LevelInfo, nil
+	case level == "WARN" || level == "warn":
+		return slog.LevelWarn, nil
+	case level == "ERROR" || level == "error":
+		return slog.LevelError, nil
+	default:
+		return slog.LevelInfo, fmt.Errorf("unknown log level: %s", level)
+	}
+}
+
 func setupStdErrHandler() slog.Handler {
-	programLevel := new(slog.LevelVar) // Info by default
-
-	envVar := "DEBUG"
+	// Parse LOG_LEVEL environment variable
+	logLevelVar := "LOG_LEVEL"
 	if len(ConfigPrefix) > 0 {
-		envVar = ConfigPrefix + "_" + envVar
+		logLevelVar = ConfigPrefix + "_" + logLevelVar
 	}
 
-	if opt := os.Getenv(envVar); len(opt) > 0 {
+	if levelStr := os.Getenv(logLevelVar); levelStr != "" {
+		if level, err := ParseLevel(levelStr); err == nil {
+			Level.Set(level)
+		}
+	}
+
+	// Maintain backward compatibility with DEBUG environment variable
+	debugVar := "DEBUG"
+	if len(ConfigPrefix) > 0 {
+		debugVar = ConfigPrefix + "_" + debugVar
+	}
+
+	if opt := os.Getenv(debugVar); len(opt) > 0 {
 		if debug, _ := strconv.ParseBool(opt); debug {
-			programLevel.Set(slog.LevelDebug)
+			Level.Set(slog.LevelDebug)
 		}
 	}
 
-	logOptions := &slog.HandlerOptions{Level: programLevel}
+	logOptions := &slog.HandlerOptions{Level: Level}
 
 	if len(os.Getenv("INVOCATION_ID")) > 0 {
 		// don't add timestamps when running under systemd
@@ -88,6 +144,18 @@ func setupStdErrHandler() slog.Handler {
 
 func setupOtlpLogger() *slog.Logger {
 	setupOtlp.Do(func() {
+		// Parse OTLP_LOG_LEVEL environment variable
+		otlpLevelVar := "OTLP_LOG_LEVEL"
+		if len(ConfigPrefix) > 0 {
+			otlpLevelVar = ConfigPrefix + "_" + otlpLevelVar
+		}
+
+		if levelStr := os.Getenv(otlpLevelVar); levelStr != "" {
+			if level, err := ParseLevel(levelStr); err == nil {
+				OTLPLevel.Set(level)
+			}
+		}
+
 		// Create our buffering exporter
 		// It will buffer until tracing is configured
 		bufferingExp := newBufferingExporter()
@@ -107,8 +175,9 @@ func setupOtlpLogger() *slog.Logger {
 		// Set global provider
 		global.SetLoggerProvider(provider)
 
-		// Create slog handler
-		handler := newLogFmtHandler(otelslog.NewHandler("common"))
+		// Create slog handler with level control
+		baseHandler := newLogFmtHandler(otelslog.NewHandler("common"))
+		handler := newOTLPLevelHandler(baseHandler)
 		otlpLogger = slog.New(handler)
 	})
 	return otlpLogger

logger/otlp_handler.go

@@ -0,0 +1,48 @@
+package logger
+
+import (
+	"context"
+	"log/slog"
+)
+
+// otlpLevelHandler is a wrapper that enforces level checking for OTLP handlers.
+// This allows independent level control for OTLP output separate from stderr logging.
+type otlpLevelHandler struct {
+	next slog.Handler
+}
+
+// newOTLPLevelHandler creates a new OTLP level wrapper handler.
+func newOTLPLevelHandler(next slog.Handler) slog.Handler {
+	return &otlpLevelHandler{
+		next: next,
+	}
+}
+
+// Enabled checks if the log level should be processed by the OTLP handler.
+// It uses the OTLPLevel variable to determine if the record should be processed.
+func (h *otlpLevelHandler) Enabled(ctx context.Context, level slog.Level) bool {
+	return level >= OTLPLevel.Level()
+}
+
+// Handle processes the log record if the level is enabled.
+// If disabled by level checking, the record is silently dropped.
+func (h *otlpLevelHandler) Handle(ctx context.Context, r slog.Record) error {
+	if !h.Enabled(ctx, r.Level) {
+		return nil
+	}
+	return h.next.Handle(ctx, r)
+}
+
+// WithAttrs returns a new handler with the specified attributes added.
+func (h *otlpLevelHandler) WithAttrs(attrs []slog.Attr) slog.Handler {
+	return &otlpLevelHandler{
+		next: h.next.WithAttrs(attrs),
+	}
+}
+
+// WithGroup returns a new handler with the specified group name.
+func (h *otlpLevelHandler) WithGroup(name string) slog.Handler {
+	return &otlpLevelHandler{
+		next: h.next.WithGroup(name),
+	}
+}