Merge pull request #80 from fosrl/dev

1.4.0

README.md

@@ -138,7 +138,7 @@ make
 
 ### Binary
 
-Make sure to have Go 1.23.1 installed.
+Make sure to have Go 1.26 installed.
 
 ```bash
 make local

docs/observability.md

@@ -83,6 +83,7 @@ type OTelConfig struct {
     Endpoint       string        // default: "localhost:4317"
     Insecure       bool          // default: true
     ExportInterval time.Duration // default: 60s
+    Timeout        time.Duration // default: 10s
 }
 ```
 
@@ -97,6 +98,7 @@ type OTelConfig struct {
 | `OTEL_METRICS_ENDPOINT` | `localhost:4317` | OTLP collector address |
 | `OTEL_METRICS_INSECURE` | `true` | Disable TLS for OTLP |
 | `OTEL_METRICS_EXPORT_INTERVAL` | `60s` | Push interval (e.g. `10s`, `1m`) |
+| `OTEL_METRICS_TIMEOUT` | `10s` | Timeout for OTLP exporter connection setup |
 | `DEPLOYMENT_ENVIRONMENT` | _(unset)_ | OTel deployment.environment attribute |
 
 ### CLI flags
@@ -108,7 +110,8 @@ type OTelConfig struct {
 --otel-metrics-protocol      string  (default: grpc)
 --otel-metrics-endpoint      string  (default: localhost:4317)
 --otel-metrics-insecure      bool    (default: true)
---otel-metrics-export-interval  duration  (default: 1m0s)
+--otel-metrics-export-interval  duration  (default: 60s)
+--otel-metrics-timeout          duration  (default: 10s)
 ```
 
 ---
@@ -164,6 +167,7 @@ export OTEL_METRICS_PROTOCOL=grpc
 export OTEL_METRICS_ENDPOINT=otel-collector:4317
 export OTEL_METRICS_INSECURE=true
 export OTEL_METRICS_EXPORT_INTERVAL=10s
+export OTEL_METRICS_TIMEOUT=10s
 export DEPLOYMENT_ENVIRONMENT=production
 ```
 
@@ -176,6 +180,7 @@ export DEPLOYMENT_ENVIRONMENT=production
          --otel-metrics-endpoint=otel-collector:4317 \
          --otel-metrics-insecure \
          --otel-metrics-export-interval=10s \
+         --otel-metrics-timeout=10s \
          --config=/etc/gerbil/config.json
 ```
 
@@ -225,7 +230,6 @@ All metrics use the prefix `gerbil_<component>_<name>`.
 | Metric | Type | Labels |
 |--------|------|--------|
 | `gerbil_proxy_mapping_active` | UpDownCounter | `ifname` |
-| `gerbil_session_active` | UpDownCounter | `ifname` |
 | `gerbil_active_sessions` | UpDownCounter | `ifname` |
 | `gerbil_udp_packets_total` | Counter | `ifname`, `type`, `direction` |
 | `gerbil_hole_punch_events_total` | Counter | `ifname`, `result` |
@@ -256,7 +260,7 @@ The `docker-compose.metrics.yml` provides a complete observability stack.
 **Prometheus mode:**
 
 ```bash
-METRICS_BACKEND=prometheus docker-compose -f docker-compose.metrics.yml up -d
+METRICS_BACKEND=prometheus docker-compose -f docker compose.metrics.yml up -d
 # Scrape at http://localhost:3003/metrics
 # Grafana at http://localhost:3000 (admin/admin)
 ```
@@ -265,5 +269,5 @@ METRICS_BACKEND=prometheus docker-compose -f docker-compose.metrics.yml up -d
 
 ```bash
 METRICS_BACKEND=otel OTEL_METRICS_ENDPOINT=otel-collector:4317 \
-  docker-compose -f docker-compose.metrics.yml up -d
+  docker compose -f docker-compose.metrics.yml up -d
 ```

examples/otel-collector-config.yaml

@@ -1,3 +1,4 @@
+file_format: '1.0'
 receivers:
   otlp:
     protocols:

go.mod

@@ -6,12 +6,12 @@ require (
 	github.com/patrickmn/go-cache v2.1.0+incompatible
 	github.com/prometheus/client_golang v1.20.5
 	github.com/vishvananda/netlink v1.3.1
-	go.opentelemetry.io/otel v1.42.0
+	go.opentelemetry.io/otel v1.43.0
-	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.42.0
+	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0
-	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.42.0
+	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.43.0
-	go.opentelemetry.io/otel/metric v1.42.0
+	go.opentelemetry.io/otel/metric v1.43.0
-	go.opentelemetry.io/otel/sdk v1.42.0
+	go.opentelemetry.io/otel/sdk v1.43.0
-	go.opentelemetry.io/otel/sdk/metric v1.42.0
+	go.opentelemetry.io/otel/sdk/metric v1.43.0
 	golang.org/x/crypto v0.49.0
 	golang.org/x/sync v0.20.0
 	golang.zx2c4.com/wireguard/wgctrl v0.0.0-20230429144221-925a1e7659e6
@@ -37,14 +37,14 @@ require (
 	github.com/prometheus/procfs v0.15.1 // indirect
 	github.com/vishvananda/netns v0.0.5 // indirect
 	go.opentelemetry.io/auto/sdk v1.2.1 // indirect
-	go.opentelemetry.io/otel/trace v1.42.0 // indirect
+	go.opentelemetry.io/otel/trace v1.43.0 // indirect
-	go.opentelemetry.io/proto/otlp v1.9.0 // indirect
+	go.opentelemetry.io/proto/otlp v1.10.0 // indirect
-	golang.org/x/net v0.51.0 // indirect
+	golang.org/x/net v0.52.0 // indirect
 	golang.org/x/sys v0.42.0 // indirect
 	golang.org/x/text v0.35.0 // indirect
 	golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b // indirect
-	google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57 // indirect
+	google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9 // indirect
-	google.golang.org/genproto/googleapis/rpc v0.0.0-20260209200024-4cfbd4190f57 // indirect
+	google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9 // indirect
-	google.golang.org/grpc v1.79.3 // indirect
+	google.golang.org/grpc v1.80.0 // indirect
 	google.golang.org/protobuf v1.36.11 // indirect
 )

go.sum

@@ -55,26 +55,26 @@ github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zd
 github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
 go.opentelemetry.io/auto/sdk v1.2.1 h1:jXsnJ4Lmnqd11kwkBV2LgLoFMZKizbCi5fNZ/ipaZ64=
 go.opentelemetry.io/auto/sdk v1.2.1/go.mod h1:KRTj+aOaElaLi+wW1kO/DZRXwkF4C5xPbEe3ZiIhN7Y=
-go.opentelemetry.io/otel v1.42.0 h1:lSQGzTgVR3+sgJDAU/7/ZMjN9Z+vUip7leaqBKy4sho=
+go.opentelemetry.io/otel v1.43.0 h1:mYIM03dnh5zfN7HautFE4ieIig9amkNANT+xcVxAj9I=
-go.opentelemetry.io/otel v1.42.0/go.mod h1:lJNsdRMxCUIWuMlVJWzecSMuNjE7dOYyWlqOXWkdqCc=
+go.opentelemetry.io/otel v1.43.0/go.mod h1:JuG+u74mvjvcm8vj8pI5XiHy1zDeoCS2LB1spIq7Ay0=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.42.0 h1:MdKucPl/HbzckWWEisiNqMPhRrAOQX8r4jTuGr636gk=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0 h1:8UQVDcZxOJLtX6gxtDt3vY2WTgvZqMQRzjsqiIHQdkc=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.42.0/go.mod h1:RolT8tWtfHcjajEH5wFIZ4Dgh5jpPdFXYV9pTAk/qjc=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.43.0/go.mod h1:2lmweYCiHYpEjQ/lSJBYhj9jP1zvCvQW4BqL9dnT7FQ=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.42.0 h1:H7O6RlGOMTizyl3R08Kn5pdM06bnH8oscSj7o11tmLA=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.43.0 h1:w1K+pCJoPpQifuVpsKamUdn9U0zM3xUziVOqsGksUrY=
-go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.42.0/go.mod h1:mBFWu/WOVDkWWsR7Tx7h6EpQB8wsv7P0Yrh0Pb7othc=
+go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp v1.43.0/go.mod h1:HBy4BjzgVE8139ieRI75oXm3EcDN+6GhD88JT1Kjvxg=
-go.opentelemetry.io/otel/metric v1.42.0 h1:2jXG+3oZLNXEPfNmnpxKDeZsFI5o4J+nz6xUlaFdF/4=
+go.opentelemetry.io/otel/metric v1.43.0 h1:d7638QeInOnuwOONPp4JAOGfbCEpYb+K6DVWvdxGzgM=
-go.opentelemetry.io/otel/metric v1.42.0/go.mod h1:RlUN/7vTU7Ao/diDkEpQpnz3/92J9ko05BIwxYa2SSI=
+go.opentelemetry.io/otel/metric v1.43.0/go.mod h1:RDnPtIxvqlgO8GRW18W6Z/4P462ldprJtfxHxyKd2PY=
-go.opentelemetry.io/otel/sdk v1.42.0 h1:LyC8+jqk6UJwdrI/8VydAq/hvkFKNHZVIWuslJXYsDo=
+go.opentelemetry.io/otel/sdk v1.43.0 h1:pi5mE86i5rTeLXqoF/hhiBtUNcrAGHLKQdhg4h4V9Dg=
-go.opentelemetry.io/otel/sdk v1.42.0/go.mod h1:rGHCAxd9DAph0joO4W6OPwxjNTYWghRWmkHuGbayMts=
+go.opentelemetry.io/otel/sdk v1.43.0/go.mod h1:P+IkVU3iWukmiit/Yf9AWvpyRDlUeBaRg6Y+C58QHzg=
-go.opentelemetry.io/otel/sdk/metric v1.42.0 h1:D/1QR46Clz6ajyZ3G8SgNlTJKBdGp84q9RKCAZ3YGuA=
+go.opentelemetry.io/otel/sdk/metric v1.43.0 h1:S88dyqXjJkuBNLeMcVPRFXpRw2fuwdvfCGLEo89fDkw=
-go.opentelemetry.io/otel/sdk/metric v1.42.0/go.mod h1:Ua6AAlDKdZ7tdvaQKfSmnFTdHx37+J4ba8MwVCYM5hc=
+go.opentelemetry.io/otel/sdk/metric v1.43.0/go.mod h1:C/RJtwSEJ5hzTiUz5pXF1kILHStzb9zFlIEe85bhj6A=
-go.opentelemetry.io/otel/trace v1.42.0 h1:OUCgIPt+mzOnaUTpOQcBiM/PLQ/Op7oq6g4LenLmOYY=
+go.opentelemetry.io/otel/trace v1.43.0 h1:BkNrHpup+4k4w+ZZ86CZoHHEkohws8AY+WTX09nk+3A=
-go.opentelemetry.io/otel/trace v1.42.0/go.mod h1:f3K9S+IFqnumBkKhRJMeaZeNk9epyhnCmQh/EysQCdc=
+go.opentelemetry.io/otel/trace v1.43.0/go.mod h1:/QJhyVBUUswCphDVxq+8mld+AvhXZLhe+8WVFxiFff0=
-go.opentelemetry.io/proto/otlp v1.9.0 h1:l706jCMITVouPOqEnii2fIAuO3IVGBRPV5ICjceRb/A=
+go.opentelemetry.io/proto/otlp v1.10.0 h1:IQRWgT5srOCYfiWnpqUYz9CVmbO8bFmKcwYxpuCSL2g=
-go.opentelemetry.io/proto/otlp v1.9.0/go.mod h1:xE+Cx5E/eEHw+ISFkwPLwCZefwVjY+pqKg1qcK03+/4=
+go.opentelemetry.io/proto/otlp v1.10.0/go.mod h1:/CV4QoCR/S9yaPj8utp3lvQPoqMtxXdzn7ozvvozVqk=
 golang.org/x/crypto v0.49.0 h1:+Ng2ULVvLHnJ/ZFEq4KdcDd/cfjrrjjNSXNzxg0Y4U4=
 golang.org/x/crypto v0.49.0/go.mod h1:ErX4dUh2UM+CFYiXZRTcMpEcN8b/1gxEuv3nODoYtCA=
-golang.org/x/net v0.51.0 h1:94R/GTO7mt3/4wIKpcR5gkGmRLOuE/2hNGeWq/GBIFo=
+golang.org/x/net v0.52.0 h1:He/TN1l0e4mmR3QqHMT2Xab3Aj3L9qjbhRm78/6jrW0=
-golang.org/x/net v0.51.0/go.mod h1:aamm+2QF5ogm02fjy5Bb7CQ0WMt1/WVM7FtyaTLlA9Y=
+golang.org/x/net v0.52.0/go.mod h1:R1MAz7uMZxVMualyPXb+VaqGSa3LIaUqk0eEt3w36Sw=
 golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
 golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
 golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
@@ -87,14 +87,14 @@ golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b h1:J1CaxgLerRR5lgx
 golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b/go.mod h1:tqur9LnfstdR9ep2LaJT4lFUl0EjlHtge+gAjmsHUG4=
 golang.zx2c4.com/wireguard/wgctrl v0.0.0-20230429144221-925a1e7659e6 h1:CawjfCvYQH2OU3/TnxLx97WDSUDRABfT18pCOYwc2GE=
 golang.zx2c4.com/wireguard/wgctrl v0.0.0-20230429144221-925a1e7659e6/go.mod h1:3rxYc4HtVcSG9gVaTs2GEBdehh+sYPOwKtyUWEOTb80=
-gonum.org/v1/gonum v0.16.0 h1:5+ul4Swaf3ESvrOnidPp4GZbzf0mxVQpDCYUQE7OJfk=
+gonum.org/v1/gonum v0.17.0 h1:VbpOemQlsSMrYmn7T2OUvQ4dqxQXU+ouZFQsZOx50z4=
-gonum.org/v1/gonum v0.16.0/go.mod h1:fef3am4MQ93R2HHpKnLk4/Tbh/s0+wqD5nfa6Pnwy4E=
+gonum.org/v1/gonum v0.17.0/go.mod h1:El3tOrEuMpv2UdMrbNlKEh9vd86bmQ6vqIcDwxEOc1E=
-google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57 h1:JLQynH/LBHfCTSbDWl+py8C+Rg/k1OVH3xfcaiANuF0=
+google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9 h1:VPWxll4HlMw1Vs/qXtN7BvhZqsS9cdAittCNvVENElA=
-google.golang.org/genproto/googleapis/api v0.0.0-20260209200024-4cfbd4190f57/go.mod h1:kSJwQxqmFXeo79zOmbrALdflXQeAYcUbgS7PbpMknCY=
+google.golang.org/genproto/googleapis/api v0.0.0-20260401024825-9d38bb4040a9/go.mod h1:7QBABkRtR8z+TEnmXTqIqwJLlzrZKVfAUm7tY3yGv0M=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260209200024-4cfbd4190f57 h1:mWPCjDEyshlQYzBpMNHaEof6UX1PmHcaUODUywQ0uac=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9 h1:m8qni9SQFH0tJc1X0vmnpw/0t+AImlSvp30sEupozUg=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20260209200024-4cfbd4190f57/go.mod h1:j9x/tPzZkyxcgEFkiKEEGxfvyumM01BEtsW8xzOahRQ=
+google.golang.org/genproto/googleapis/rpc v0.0.0-20260401024825-9d38bb4040a9/go.mod h1:4Hqkh8ycfw05ld/3BWL7rJOSfebL2Q+DVDeRgYgxUU8=
-google.golang.org/grpc v1.79.3 h1:sybAEdRIEtvcD68Gx7dmnwjZKlyfuc61Dyo9pGXXkKE=
+google.golang.org/grpc v1.80.0 h1:Xr6m2WmWZLETvUNvIUmeD5OAagMw3FiKmMlTdViWsHM=
-google.golang.org/grpc v1.79.3/go.mod h1:KmT0Kjez+0dde/v2j9vzwoAScgEPx/Bw1CYChhHLrHQ=
+google.golang.org/grpc v1.80.0/go.mod h1:ho/dLnxwi3EDJA4Zghp7k2Ec1+c2jqup0bFkw07bwF4=
 google.golang.org/protobuf v1.36.11 h1:fV6ZwhNocDyBLK0dj+fg8ektcVegBBuEolpbTQyBNVE=
 google.golang.org/protobuf v1.36.11/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
 gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=

internal/metrics/metrics.go

@@ -7,7 +7,9 @@ package metrics
 
 import (
 	"context"
+	"fmt"
 	"net/http"
+	"sync"
 
 	"github.com/fosrl/gerbil/internal/observability"
 )
@@ -24,6 +26,7 @@ type OTelConfig = observability.OTelConfig
 
 var (
 	backend observability.Backend
+	initMu  sync.Mutex
 
 	// Interface and peer metrics
 	wgInterfaceUp      observability.Int64Gauge
@@ -55,7 +58,6 @@ var (
 	udpPacketSizeBytes         observability.Histogram
 	holePunchEventsTotal       observability.Counter
 	proxyMappingActive         observability.UpDownCounter
-	sessionActive              observability.UpDownCounter
 	sessionRebuiltTotal        observability.Counter
 	commPatternActive          observability.UpDownCounter
 	proxyCleanupRemovedTotal   observability.Counter
@@ -107,6 +109,13 @@ func DefaultConfig() Config {
 // Initialize sets up the metrics system using the selected backend.
 // It returns the /metrics HTTP handler (non-nil only for Prometheus backend).
 func Initialize(cfg Config) (http.Handler, error) {
+	initMu.Lock()
+	defer initMu.Unlock()
+
+	if backend != nil {
+		return backend.HTTPHandler(), nil
+	}
+
 	b, err := observability.New(cfg)
 	if err != nil {
 		return nil, err
@@ -114,6 +123,7 @@ func Initialize(cfg Config) (http.Handler, error) {
 	backend = b
 
 	if err := createInstruments(); err != nil {
+		backend = nil
 		return nil, err
 	}
 
@@ -122,8 +132,13 @@ func Initialize(cfg Config) (http.Handler, error) {
 
 // Shutdown gracefully shuts down the metrics backend.
 func Shutdown(ctx context.Context) error {
-	if backend != nil {
+	initMu.Lock()
-		return backend.Shutdown(ctx)
+	b := backend
+	backend = nil
+	initMu.Unlock()
+
+	if b != nil {
+		return b.Shutdown(ctx)
 	}
 	return nil
 }
@@ -135,129 +150,346 @@ func createInstruments() error {
 
 	b := backend
 
-	wgInterfaceUp = b.NewInt64Gauge("gerbil_wg_interface_up",
+	newCounter := func(name, desc string, labelNames ...string) (observability.Counter, error) {
+		c, err := b.NewCounter(name, desc, labelNames...)
+		if err != nil {
+			return nil, fmt.Errorf("create counter %q: %w", name, err)
+		}
+		return c, nil
+	}
+
+	newUpDownCounter := func(name, desc string, labelNames ...string) (observability.UpDownCounter, error) {
+		c, err := b.NewUpDownCounter(name, desc, labelNames...)
+		if err != nil {
+			return nil, fmt.Errorf("create updown counter %q: %w", name, err)
+		}
+		return c, nil
+	}
+
+	newInt64Gauge := func(name, desc string, labelNames ...string) (observability.Int64Gauge, error) {
+		g, err := b.NewInt64Gauge(name, desc, labelNames...)
+		if err != nil {
+			return nil, fmt.Errorf("create int64 gauge %q: %w", name, err)
+		}
+		return g, nil
+	}
+
+	newFloat64Gauge := func(name, desc string, labelNames ...string) (observability.Float64Gauge, error) {
+		g, err := b.NewFloat64Gauge(name, desc, labelNames...)
+		if err != nil {
+			return nil, fmt.Errorf("create float64 gauge %q: %w", name, err)
+		}
+		return g, nil
+	}
+
+	newHistogram := func(name, desc string, buckets []float64, labelNames ...string) (observability.Histogram, error) {
+		h, err := b.NewHistogram(name, desc, buckets, labelNames...)
+		if err != nil {
+			return nil, fmt.Errorf("create histogram %q: %w", name, err)
+		}
+		return h, nil
+	}
+
+	var err error
+
+	wgInterfaceUp, err = newInt64Gauge("gerbil_wg_interface_up",
 		"Operational state of a WireGuard interface (1=up, 0=down)", "ifname", "instance")
-	wgPeersTotal = b.NewUpDownCounter("gerbil_wg_peers_total",
+	if err != nil {
+		return err
+	}
+	wgPeersTotal, err = newUpDownCounter("gerbil_wg_peers_total",
 		"Total number of configured peers per interface", "ifname")
-	wgPeerConnected = b.NewInt64Gauge("gerbil_wg_peer_connected",
+	if err != nil {
+		return err
+	}
+	wgPeerConnected, err = newInt64Gauge("gerbil_wg_peer_connected",
 		"Whether a specific peer is connected (1=connected, 0=disconnected)", "ifname", "peer")
-	allowedIPsCount = b.NewUpDownCounter("gerbil_allowed_ips_count",
+	if err != nil {
+		return err
+	}
+	allowedIPsCount, err = newUpDownCounter("gerbil_allowed_ips_count",
 		"Number of allowed IPs configured per peer", "ifname", "peer")
-	keyRotationTotal = b.NewCounter("gerbil_key_rotation_total",
+	if err != nil {
+		return err
+	}
+	keyRotationTotal, err = newCounter("gerbil_key_rotation_total",
 		"Key rotation events", "ifname", "reason")
-	wgHandshakesTotal = b.NewCounter("gerbil_wg_handshakes_total",
+	if err != nil {
+		return err
+	}
+	wgHandshakesTotal, err = newCounter("gerbil_wg_handshakes_total",
 		"Count of handshake attempts with their result status", "ifname", "peer", "result")
-	wgHandshakeLatency = b.NewHistogram("gerbil_wg_handshake_latency_seconds",
+	if err != nil {
+		return err
+	}
+	wgHandshakeLatency, err = newHistogram("gerbil_wg_handshake_latency_seconds",
 		"Distribution of handshake latencies in seconds", durationBuckets, "ifname", "peer")
-	wgPeerRTT = b.NewHistogram("gerbil_wg_peer_rtt_seconds",
+	if err != nil {
+		return err
+	}
+	wgPeerRTT, err = newHistogram("gerbil_wg_peer_rtt_seconds",
 		"Observed round-trip time to a peer in seconds", durationBuckets, "ifname", "peer")
-	wgBytesReceived = b.NewCounter("gerbil_wg_bytes_received_total",
+	if err != nil {
+		return err
+	}
+	wgBytesReceived, err = newCounter("gerbil_wg_bytes_received_total",
 		"Number of bytes received from a peer", "ifname", "peer")
-	wgBytesTransmitted = b.NewCounter("gerbil_wg_bytes_transmitted_total",
+	if err != nil {
+		return err
+	}
+	wgBytesTransmitted, err = newCounter("gerbil_wg_bytes_transmitted_total",
 		"Number of bytes transmitted to a peer", "ifname", "peer")
-	netlinkEventsTotal = b.NewCounter("gerbil_netlink_events_total",
+	if err != nil {
+		return err
+	}
+	netlinkEventsTotal, err = newCounter("gerbil_netlink_events_total",
 		"Number of netlink events processed", "event_type")
-	netlinkErrorsTotal = b.NewCounter("gerbil_netlink_errors_total",
+	if err != nil {
+		return err
+	}
+	netlinkErrorsTotal, err = newCounter("gerbil_netlink_errors_total",
 		"Count of netlink or kernel errors", "component", "error_type")
-	syncDuration = b.NewHistogram("gerbil_sync_duration_seconds",
+	if err != nil {
+		return err
+	}
+	syncDuration, err = newHistogram("gerbil_sync_duration_seconds",
 		"Duration of reconciliation/sync loops in seconds", durationBuckets, "component")
-	workqueueDepth = b.NewUpDownCounter("gerbil_workqueue_depth",
+	if err != nil {
+		return err
+	}
+	workqueueDepth, err = newUpDownCounter("gerbil_workqueue_depth",
 		"Current length of internal work queues", "queue")
-	kernelModuleLoads = b.NewCounter("gerbil_kernel_module_loads_total",
+	if err != nil {
+		return err
+	}
+	kernelModuleLoads, err = newCounter("gerbil_kernel_module_loads_total",
 		"Count of kernel module load attempts", "result")
-	firewallRulesApplied = b.NewCounter("gerbil_firewall_rules_applied_total",
+	if err != nil {
+		return err
+	}
+	firewallRulesApplied, err = newCounter("gerbil_firewall_rules_applied_total",
 		"IPTables/NFT rules applied", "result", "chain")
-	activeSessions = b.NewUpDownCounter("gerbil_active_sessions",
+	if err != nil {
+		return err
+	}
+	activeSessions, err = newUpDownCounter("gerbil_active_sessions",
 		"Number of active UDP relay sessions", "ifname")
-	activeProxyConnections = b.NewUpDownCounter("gerbil_active_proxy_connections",
+	if err != nil {
+		return err
+	}
+	activeProxyConnections, err = newUpDownCounter("gerbil_active_proxy_connections",
 		"Active SNI proxy connections")
-	proxyRouteLookups = b.NewCounter("gerbil_proxy_route_lookups_total",
+	if err != nil {
+		return err
+	}
+	proxyRouteLookups, err = newCounter("gerbil_proxy_route_lookups_total",
 		"Number of route lookups", "result")
-	proxyTLSHandshake = b.NewHistogram("gerbil_proxy_tls_handshake_seconds",
+	if err != nil {
+		return err
+	}
+	proxyTLSHandshake, err = newHistogram("gerbil_proxy_tls_handshake_seconds",
 		"TLS handshake duration for SNI proxy in seconds", durationBuckets)
-	proxyBytesTransmitted = b.NewCounter("gerbil_proxy_bytes_transmitted_total",
+	if err != nil {
+		return err
+	}
+	proxyBytesTransmitted, err = newCounter("gerbil_proxy_bytes_transmitted_total",
 		"Bytes sent/received by the SNI proxy", "direction")
-	configReloadsTotal = b.NewCounter("gerbil_config_reloads_total",
+	if err != nil {
+		return err
+	}
+	configReloadsTotal, err = newCounter("gerbil_config_reloads_total",
 		"Number of configuration reloads", "result")
-	restartTotal = b.NewCounter("gerbil_restart_total",
+	if err != nil {
+		return err
+	}
+	restartTotal, err = newCounter("gerbil_restart_total",
 		"Process restart count")
-	authFailuresTotal = b.NewCounter("gerbil_auth_failures_total",
+	if err != nil {
+		return err
+	}
+	authFailuresTotal, err = newCounter("gerbil_auth_failures_total",
 		"Count of authentication or peer validation failures", "peer", "reason")
-	aclDeniedTotal = b.NewCounter("gerbil_acl_denied_total",
+	if err != nil {
+		return err
+	}
+	aclDeniedTotal, err = newCounter("gerbil_acl_denied_total",
 		"Access control denied events", "ifname", "peer", "policy")
-	certificateExpiryDays = b.NewFloat64Gauge("gerbil_certificate_expiry_days",
+	if err != nil {
+		return err
+	}
+	certificateExpiryDays, err = newFloat64Gauge("gerbil_certificate_expiry_days",
 		"Days until certificate expiry", "cert_name", "ifname")
-	udpPacketsTotal = b.NewCounter("gerbil_udp_packets_total",
+	if err != nil {
+		return err
+	}
+	udpPacketsTotal, err = newCounter("gerbil_udp_packets_total",
 		"Count of UDP packets processed by relay workers", "ifname", "type", "direction")
-	udpPacketSizeBytes = b.NewHistogram("gerbil_udp_packet_size_bytes",
+	if err != nil {
+		return err
+	}
+	udpPacketSizeBytes, err = newHistogram("gerbil_udp_packet_size_bytes",
 		"Size distribution of packets forwarded through relay", sizeBuckets, "ifname", "type")
-	holePunchEventsTotal = b.NewCounter("gerbil_hole_punch_events_total",
+	if err != nil {
+		return err
+	}
+	holePunchEventsTotal, err = newCounter("gerbil_hole_punch_events_total",
 		"Count of hole punch messages processed", "ifname", "result")
-	proxyMappingActive = b.NewUpDownCounter("gerbil_proxy_mapping_active",
+	if err != nil {
+		return err
+	}
+	proxyMappingActive, err = newUpDownCounter("gerbil_proxy_mapping_active",
 		"Number of active proxy mappings", "ifname")
-	sessionActive = b.NewUpDownCounter("gerbil_session_active",
+	if err != nil {
-		"Number of active WireGuard sessions", "ifname")
+		return err
-	sessionRebuiltTotal = b.NewCounter("gerbil_session_rebuilt_total",
+	}
+	sessionRebuiltTotal, err = newCounter("gerbil_session_rebuilt_total",
 		"Count of sessions rebuilt from communication patterns", "ifname")
-	commPatternActive = b.NewUpDownCounter("gerbil_comm_pattern_active",
+	if err != nil {
+		return err
+	}
+	commPatternActive, err = newUpDownCounter("gerbil_comm_pattern_active",
 		"Number of active communication patterns", "ifname")
-	proxyCleanupRemovedTotal = b.NewCounter("gerbil_proxy_cleanup_removed_total",
+	if err != nil {
+		return err
+	}
+	proxyCleanupRemovedTotal, err = newCounter("gerbil_proxy_cleanup_removed_total",
 		"Count of items removed during cleanup routines", "ifname", "component")
-	proxyConnectionErrorsTotal = b.NewCounter("gerbil_proxy_connection_errors_total",
+	if err != nil {
+		return err
+	}
+	proxyConnectionErrorsTotal, err = newCounter("gerbil_proxy_connection_errors_total",
 		"Count of connection errors in proxy operations", "ifname", "error_type")
-	proxyInitialMappingsTotal = b.NewInt64Gauge("gerbil_proxy_initial_mappings",
+	if err != nil {
+		return err
+	}
+	proxyInitialMappingsTotal, err = newInt64Gauge("gerbil_proxy_initial_mappings",
 		"Number of initial proxy mappings loaded", "ifname")
-	proxyMappingUpdatesTotal = b.NewCounter("gerbil_proxy_mapping_updates_total",
+	if err != nil {
+		return err
+	}
+	proxyMappingUpdatesTotal, err = newCounter("gerbil_proxy_mapping_updates_total",
 		"Count of proxy mapping updates", "ifname")
-	proxyIdleCleanupDuration = b.NewHistogram("gerbil_proxy_idle_cleanup_duration_seconds",
+	if err != nil {
+		return err
+	}
+	proxyIdleCleanupDuration, err = newHistogram("gerbil_proxy_idle_cleanup_duration_seconds",
 		"Duration of cleanup cycles", durationBuckets, "ifname", "component")
-	sniConnectionsTotal = b.NewCounter("gerbil_sni_connections_total",
+	if err != nil {
+		return err
+	}
+	sniConnectionsTotal, err = newCounter("gerbil_sni_connections_total",
 		"Count of connections processed by SNI proxy", "result")
-	sniConnectionDuration = b.NewHistogram("gerbil_sni_connection_duration_seconds",
+	if err != nil {
+		return err
+	}
+	sniConnectionDuration, err = newHistogram("gerbil_sni_connection_duration_seconds",
 		"Lifetime distribution of proxied TLS connections", sniDurationBuckets)
-	sniActiveConnections = b.NewUpDownCounter("gerbil_sni_active_connections",
+	if err != nil {
+		return err
+	}
+	sniActiveConnections, err = newUpDownCounter("gerbil_sni_active_connections",
 		"Number of active SNI tunnels")
-	sniRouteCacheHitsTotal = b.NewCounter("gerbil_sni_route_cache_hits_total",
+	if err != nil {
+		return err
+	}
+	sniRouteCacheHitsTotal, err = newCounter("gerbil_sni_route_cache_hits_total",
 		"Count of route cache hits and misses", "result")
-	sniRouteAPIRequestsTotal = b.NewCounter("gerbil_sni_route_api_requests_total",
+	if err != nil {
+		return err
+	}
+	sniRouteAPIRequestsTotal, err = newCounter("gerbil_sni_route_api_requests_total",
 		"Count of route API requests", "result")
-	sniRouteAPILatency = b.NewHistogram("gerbil_sni_route_api_latency_seconds",
+	if err != nil {
+		return err
+	}
+	sniRouteAPILatency, err = newHistogram("gerbil_sni_route_api_latency_seconds",
 		"Distribution of route API call latencies", durationBuckets)
-	sniLocalOverrideTotal = b.NewCounter("gerbil_sni_local_override_total",
+	if err != nil {
+		return err
+	}
+	sniLocalOverrideTotal, err = newCounter("gerbil_sni_local_override_total",
 		"Count of routes using local overrides", "hit")
-	sniTrustedProxyEventsTotal = b.NewCounter("gerbil_sni_trusted_proxy_events_total",
+	if err != nil {
+		return err
+	}
+	sniTrustedProxyEventsTotal, err = newCounter("gerbil_sni_trusted_proxy_events_total",
 		"Count of PROXY protocol events", "event")
-	sniProxyProtocolParseErrorsTotal = b.NewCounter("gerbil_sni_proxy_protocol_parse_errors_total",
+	if err != nil {
+		return err
+	}
+	sniProxyProtocolParseErrorsTotal, err = newCounter("gerbil_sni_proxy_protocol_parse_errors_total",
 		"Count of PROXY protocol parse failures")
-	sniDataBytesTotal = b.NewCounter("gerbil_sni_data_bytes_total",
+	if err != nil {
+		return err
+	}
+	sniDataBytesTotal, err = newCounter("gerbil_sni_data_bytes_total",
 		"Count of bytes proxied through SNI tunnels", "direction")
-	sniTunnelTerminationsTotal = b.NewCounter("gerbil_sni_tunnel_terminations_total",
+	if err != nil {
+		return err
+	}
+	sniTunnelTerminationsTotal, err = newCounter("gerbil_sni_tunnel_terminations_total",
 		"Count of tunnel terminations by reason", "reason")
-	httpRequestsTotal = b.NewCounter("gerbil_http_requests_total",
+	if err != nil {
+		return err
+	}
+	httpRequestsTotal, err = newCounter("gerbil_http_requests_total",
 		"Count of HTTP requests to management API", "endpoint", "method", "status_code")
-	httpRequestDuration = b.NewHistogram("gerbil_http_request_duration_seconds",
+	if err != nil {
+		return err
+	}
+	httpRequestDuration, err = newHistogram("gerbil_http_request_duration_seconds",
 		"Distribution of HTTP request handling time", durationBuckets, "endpoint", "method")
-	peerOperationsTotal = b.NewCounter("gerbil_peer_operations_total",
+	if err != nil {
+		return err
+	}
+	peerOperationsTotal, err = newCounter("gerbil_peer_operations_total",
 		"Count of peer lifecycle operations", "operation", "result")
-	proxyMappingUpdateRequestsTotal = b.NewCounter("gerbil_proxy_mapping_update_requests_total",
+	if err != nil {
+		return err
+	}
+	proxyMappingUpdateRequestsTotal, err = newCounter("gerbil_proxy_mapping_update_requests_total",
 		"Count of proxy mapping update API calls", "result")
-	destinationsUpdateRequestsTotal = b.NewCounter("gerbil_destinations_update_requests_total",
+	if err != nil {
+		return err
+	}
+	destinationsUpdateRequestsTotal, err = newCounter("gerbil_destinations_update_requests_total",
 		"Count of destinations update API calls", "result")
-	remoteConfigFetchesTotal = b.NewCounter("gerbil_remote_config_fetches_total",
+	if err != nil {
+		return err
+	}
+	remoteConfigFetchesTotal, err = newCounter("gerbil_remote_config_fetches_total",
 		"Count of remote configuration fetch attempts", "result")
-	bandwidthReportsTotal = b.NewCounter("gerbil_bandwidth_reports_total",
+	if err != nil {
+		return err
+	}
+	bandwidthReportsTotal, err = newCounter("gerbil_bandwidth_reports_total",
 		"Count of bandwidth report transmissions", "result")
-	peerBandwidthBytesTotal = b.NewCounter("gerbil_peer_bandwidth_bytes_total",
+	if err != nil {
+		return err
+	}
+	peerBandwidthBytesTotal, err = newCounter("gerbil_peer_bandwidth_bytes_total",
 		"Bytes per peer tracked by bandwidth calculation", "peer", "direction")
-	memorySpikeTotal = b.NewCounter("gerbil_memory_spike_total",
+	if err != nil {
+		return err
+	}
+	memorySpikeTotal, err = newCounter("gerbil_memory_spike_total",
 		"Count of memory spikes detected", "severity")
-	heapProfilesWrittenTotal = b.NewCounter("gerbil_heap_profiles_written_total",
+	if err != nil {
+		return err
+	}
+	heapProfilesWrittenTotal, err = newCounter("gerbil_heap_profiles_written_total",
 		"Count of heap profile files generated")
+	if err != nil {
+		return err
+	}
 
 	return nil
 }
 
 func RecordInterfaceUp(ifname, instance string, up bool) {
+	if wgInterfaceUp == nil {
+		return
+	}
 	value := int64(0)
 	if up {
 		value = 1
@@ -266,10 +498,16 @@ func RecordInterfaceUp(ifname, instance string, up bool) {
 }
 
 func RecordPeersTotal(ifname string, delta int64) {
+	if wgPeersTotal == nil {
+		return
+	}
 	wgPeersTotal.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
 }
 
 func RecordPeerConnected(ifname, peer string, connected bool) {
+	if wgPeerConnected == nil {
+		return
+	}
 	value := int64(0)
 	if connected {
 		value = 1
@@ -278,229 +516,393 @@ func RecordPeerConnected(ifname, peer string, connected bool) {
 }
 
 func RecordHandshake(ifname, peer, result string) {
+	if wgHandshakesTotal == nil {
+		return
+	}
 	wgHandshakesTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "peer": peer, "result": result})
 }
 
 func RecordHandshakeLatency(ifname, peer string, seconds float64) {
+	if wgHandshakeLatency == nil {
+		return
+	}
 	wgHandshakeLatency.Record(context.Background(), seconds, observability.Labels{"ifname": ifname, "peer": peer})
 }
 
 func RecordPeerRTT(ifname, peer string, seconds float64) {
+	if wgPeerRTT == nil {
+		return
+	}
 	wgPeerRTT.Record(context.Background(), seconds, observability.Labels{"ifname": ifname, "peer": peer})
 }
 
 func RecordBytesReceived(ifname, peer string, bytes int64) {
+	if wgBytesReceived == nil {
+		return
+	}
 	wgBytesReceived.Add(context.Background(), bytes, observability.Labels{"ifname": ifname, "peer": peer})
 }
 
 func RecordBytesTransmitted(ifname, peer string, bytes int64) {
+	if wgBytesTransmitted == nil {
+		return
+	}
 	wgBytesTransmitted.Add(context.Background(), bytes, observability.Labels{"ifname": ifname, "peer": peer})
 }
 
 func RecordAllowedIPsCount(ifname, peer string, delta int64) {
+	if allowedIPsCount == nil {
+		return
+	}
 	allowedIPsCount.Add(context.Background(), delta, observability.Labels{"ifname": ifname, "peer": peer})
 }
 
 func RecordKeyRotation(ifname, reason string) {
+	if keyRotationTotal == nil {
+		return
+	}
 	keyRotationTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "reason": reason})
 }
 
 func RecordNetlinkEvent(eventType string) {
+	if netlinkEventsTotal == nil {
+		return
+	}
 	netlinkEventsTotal.Add(context.Background(), 1, observability.Labels{"event_type": eventType})
 }
 
 func RecordNetlinkError(component, errorType string) {
+	if netlinkErrorsTotal == nil {
+		return
+	}
 	netlinkErrorsTotal.Add(context.Background(), 1, observability.Labels{"component": component, "error_type": errorType})
 }
 
 func RecordSyncDuration(component string, seconds float64) {
+	if syncDuration == nil {
+		return
+	}
 	syncDuration.Record(context.Background(), seconds, observability.Labels{"component": component})
 }
 
 func RecordWorkqueueDepth(queue string, delta int64) {
+	if workqueueDepth == nil {
+		return
+	}
 	workqueueDepth.Add(context.Background(), delta, observability.Labels{"queue": queue})
 }
 
 func RecordKernelModuleLoad(result string) {
+	if kernelModuleLoads == nil {
+		return
+	}
 	kernelModuleLoads.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordFirewallRuleApplied(result, chain string) {
+	if firewallRulesApplied == nil {
+		return
+	}
 	firewallRulesApplied.Add(context.Background(), 1, observability.Labels{"result": result, "chain": chain})
 }
 
 func RecordActiveSession(ifname string, delta int64) {
+	if activeSessions == nil {
+		return
+	}
 	activeSessions.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
 }
 
-func RecordActiveProxyConnection(hostname string, delta int64) {
+func RecordActiveProxyConnection(delta int64) {
-	_ = hostname
+	if activeProxyConnections == nil {
+		return
+	}
 	activeProxyConnections.Add(context.Background(), delta, nil)
 }
 
-func RecordProxyRouteLookup(result, hostname string) {
+func RecordProxyRouteLookup(result string) {
-	_ = hostname
+	if proxyRouteLookups == nil {
+		return
+	}
 	proxyRouteLookups.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
-func RecordProxyTLSHandshake(hostname string, seconds float64) {
+func RecordProxyTLSHandshake(seconds float64) {
-	_ = hostname
+	if proxyTLSHandshake == nil {
+		return
+	}
 	proxyTLSHandshake.Record(context.Background(), seconds, nil)
 }
 
-func RecordProxyBytesTransmitted(hostname, direction string, bytes int64) {
+func RecordProxyBytesTransmitted(direction string, bytes int64) {
-	_ = hostname
+	if proxyBytesTransmitted == nil {
+		return
+	}
 	proxyBytesTransmitted.Add(context.Background(), bytes, observability.Labels{"direction": direction})
 }
 
 func RecordConfigReload(result string) {
+	if configReloadsTotal == nil {
+		return
+	}
 	configReloadsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordRestart() {
+	if restartTotal == nil {
+		return
+	}
 	restartTotal.Add(context.Background(), 1, nil)
 }
 
 func RecordAuthFailure(peer, reason string) {
+	if authFailuresTotal == nil {
+		return
+	}
 	authFailuresTotal.Add(context.Background(), 1, observability.Labels{"peer": peer, "reason": reason})
 }
 
 func RecordACLDenied(ifname, peer, policy string) {
+	if aclDeniedTotal == nil {
+		return
+	}
 	aclDeniedTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "peer": peer, "policy": policy})
 }
 
 func RecordCertificateExpiry(certName, ifname string, days float64) {
+	if certificateExpiryDays == nil {
+		return
+	}
 	certificateExpiryDays.Record(context.Background(), days, observability.Labels{"cert_name": certName, "ifname": ifname})
 }
 
 func RecordUDPPacket(ifname, packetType, direction string) {
+	if udpPacketsTotal == nil {
+		return
+	}
 	udpPacketsTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "type": packetType, "direction": direction})
 }
 
 func RecordUDPPacketSize(ifname, packetType string, bytes float64) {
+	if udpPacketSizeBytes == nil {
+		return
+	}
 	udpPacketSizeBytes.Record(context.Background(), bytes, observability.Labels{"ifname": ifname, "type": packetType})
 }
 
 func RecordHolePunchEvent(ifname, result string) {
+	if holePunchEventsTotal == nil {
+		return
+	}
 	holePunchEventsTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "result": result})
 }
 
 func RecordProxyMapping(ifname string, delta int64) {
+	if proxyMappingActive == nil {
+		return
+	}
 	proxyMappingActive.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
 }
 
 func RecordSession(ifname string, delta int64) {
-	sessionActive.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
+	if activeSessions == nil {
+		return
+	}
+	activeSessions.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
 }
 
 func RecordSessionRebuilt(ifname string) {
+	if sessionRebuiltTotal == nil {
+		return
+	}
 	sessionRebuiltTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname})
 }
 
 func RecordCommPattern(ifname string, delta int64) {
+	if commPatternActive == nil {
+		return
+	}
 	commPatternActive.Add(context.Background(), delta, observability.Labels{"ifname": ifname})
 }
 
 func RecordProxyCleanupRemoved(ifname, component string, count int64) {
+	if proxyCleanupRemovedTotal == nil {
+		return
+	}
 	proxyCleanupRemovedTotal.Add(context.Background(), count, observability.Labels{"ifname": ifname, "component": component})
 }
 
 func RecordProxyConnectionError(ifname, errorType string) {
+	if proxyConnectionErrorsTotal == nil {
+		return
+	}
 	proxyConnectionErrorsTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname, "error_type": errorType})
 }
 
 func RecordProxyInitialMappings(ifname string, count int64) {
+	if proxyInitialMappingsTotal == nil {
+		return
+	}
 	proxyInitialMappingsTotal.Record(context.Background(), count, observability.Labels{"ifname": ifname})
 }
 
 func RecordProxyMappingUpdate(ifname string) {
+	if proxyMappingUpdatesTotal == nil {
+		return
+	}
 	proxyMappingUpdatesTotal.Add(context.Background(), 1, observability.Labels{"ifname": ifname})
 }
 
 func RecordProxyIdleCleanupDuration(ifname, component string, seconds float64) {
+	if proxyIdleCleanupDuration == nil {
+		return
+	}
 	proxyIdleCleanupDuration.Record(context.Background(), seconds, observability.Labels{"ifname": ifname, "component": component})
 }
 
 func RecordSNIConnection(result string) {
+	if sniConnectionsTotal == nil {
+		return
+	}
 	sniConnectionsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordSNIConnectionDuration(seconds float64) {
+	if sniConnectionDuration == nil {
+		return
+	}
 	sniConnectionDuration.Record(context.Background(), seconds, nil)
 }
 
 func RecordSNIActiveConnection(delta int64) {
+	if sniActiveConnections == nil {
+		return
+	}
 	sniActiveConnections.Add(context.Background(), delta, nil)
 }
 
 func RecordSNIRouteCacheHit(result string) {
+	if sniRouteCacheHitsTotal == nil {
+		return
+	}
 	sniRouteCacheHitsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordSNIRouteAPIRequest(result string) {
+	if sniRouteAPIRequestsTotal == nil {
+		return
+	}
 	sniRouteAPIRequestsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordSNIRouteAPILatency(seconds float64) {
+	if sniRouteAPILatency == nil {
+		return
+	}
 	sniRouteAPILatency.Record(context.Background(), seconds, nil)
 }
 
 func RecordSNILocalOverride(hit string) {
+	if sniLocalOverrideTotal == nil {
+		return
+	}
 	sniLocalOverrideTotal.Add(context.Background(), 1, observability.Labels{"hit": hit})
 }
 
 func RecordSNITrustedProxyEvent(event string) {
+	if sniTrustedProxyEventsTotal == nil {
+		return
+	}
 	sniTrustedProxyEventsTotal.Add(context.Background(), 1, observability.Labels{"event": event})
 }
 
 func RecordSNIProxyProtocolParseError() {
+	if sniProxyProtocolParseErrorsTotal == nil {
+		return
+	}
 	sniProxyProtocolParseErrorsTotal.Add(context.Background(), 1, nil)
 }
 
 func RecordSNIDataBytes(direction string, bytes int64) {
+	if sniDataBytesTotal == nil {
+		return
+	}
 	sniDataBytesTotal.Add(context.Background(), bytes, observability.Labels{"direction": direction})
 }
 
 func RecordSNITunnelTermination(reason string) {
+	if sniTunnelTerminationsTotal == nil {
+		return
+	}
 	sniTunnelTerminationsTotal.Add(context.Background(), 1, observability.Labels{"reason": reason})
 }
 
 func RecordHTTPRequest(endpoint, method, statusCode string) {
+	if httpRequestsTotal == nil {
+		return
+	}
 	httpRequestsTotal.Add(context.Background(), 1, observability.Labels{"endpoint": endpoint, "method": method, "status_code": statusCode})
 }
 
 func RecordHTTPRequestDuration(endpoint, method string, seconds float64) {
+	if httpRequestDuration == nil {
+		return
+	}
 	httpRequestDuration.Record(context.Background(), seconds, observability.Labels{"endpoint": endpoint, "method": method})
 }
 
 func RecordPeerOperation(operation, result string) {
+	if peerOperationsTotal == nil {
+		return
+	}
 	peerOperationsTotal.Add(context.Background(), 1, observability.Labels{"operation": operation, "result": result})
 }
 
 func RecordProxyMappingUpdateRequest(result string) {
+	if proxyMappingUpdateRequestsTotal == nil {
+		return
+	}
 	proxyMappingUpdateRequestsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordDestinationsUpdateRequest(result string) {
+	if destinationsUpdateRequestsTotal == nil {
+		return
+	}
 	destinationsUpdateRequestsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordRemoteConfigFetch(result string) {
+	if remoteConfigFetchesTotal == nil {
+		return
+	}
 	remoteConfigFetchesTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordBandwidthReport(result string) {
+	if bandwidthReportsTotal == nil {
+		return
+	}
 	bandwidthReportsTotal.Add(context.Background(), 1, observability.Labels{"result": result})
 }
 
 func RecordPeerBandwidthBytes(peer, direction string, bytes int64) {
+	if peerBandwidthBytesTotal == nil {
+		return
+	}
 	peerBandwidthBytesTotal.Add(context.Background(), bytes, observability.Labels{"peer": peer, "direction": direction})
 }
 
 func RecordMemorySpike(severity string) {
+	if memorySpikeTotal == nil {
+		return
+	}
 	memorySpikeTotal.Add(context.Background(), 1, observability.Labels{"severity": severity})
 }
 
 func RecordHeapProfileWritten() {
+	if heapProfilesWrittenTotal == nil {
+		return
+	}
 	heapProfilesWrittenTotal.Add(context.Background(), 1, nil)
 }

internal/metrics/metrics_test.go

@@ -89,6 +89,9 @@ func TestDefaultConfig(t *testing.T) {
 }
 
 func TestShutdownNoInit(t *testing.T) {
+	// Ensure a known clean global state before testing no-init shutdown behavior.
+	_ = metrics.Shutdown(context.Background())
+
 	// Shutdown without Initialize should not panic or error.
 	if err := metrics.Shutdown(context.Background()); err != nil {
 		t.Errorf("unexpected error: %v", err)
@@ -168,6 +171,7 @@ func TestRecordRelay(t *testing.T) {
 	body := scrape(t, h)
 	assertContains(t, body, "gerbil_udp_packets_total")
 	assertContains(t, body, "gerbil_proxy_mapping_active")
+	assertContains(t, body, "gerbil_active_sessions")
 }
 
 func TestRecordWireGuard(t *testing.T) {
@@ -216,10 +220,10 @@ func TestRecordNetlink(t *testing.T) {
 	metrics.RecordKernelModuleLoad("success")
 	metrics.RecordFirewallRuleApplied("success", "INPUT")
 	metrics.RecordActiveSession("wg0", 1)
-	metrics.RecordActiveProxyConnection(exampleHostname, 1)
+	metrics.RecordActiveProxyConnection(1)
-	metrics.RecordProxyRouteLookup("hit", exampleHostname)
+	metrics.RecordProxyRouteLookup("hit")
-	metrics.RecordProxyTLSHandshake(exampleHostname, 0.05)
+	metrics.RecordProxyTLSHandshake(0.05)
-	metrics.RecordProxyBytesTransmitted(exampleHostname, "tx", 1024)
+	metrics.RecordProxyBytesTransmitted("tx", 1024)
 	body := scrape(t, h)
 	assertContains(t, body, "gerbil_netlink_events_total")
 	assertContains(t, body, "gerbil_active_sessions")

internal/observability/config.go

@@ -60,6 +60,10 @@ type OTelConfig struct {
 	// ExportInterval is how often metrics are pushed to the collector.
 	// Defaults to 60 s.
 	ExportInterval time.Duration
+
+	// Timeout bounds OTLP exporter construction calls.
+	// Defaults to 10 s.
+	Timeout time.Duration
 }
 
 // DefaultMetricsConfig returns a MetricsConfig with sensible defaults.
@@ -75,6 +79,7 @@ func DefaultMetricsConfig() MetricsConfig {
 			Endpoint:       "localhost:4317",
 			Insecure:       true,
 			ExportInterval: 60 * time.Second,
+			Timeout:        10 * time.Second,
 		},
 		ServiceName:    "gerbil",
 		ServiceVersion: "1.0.0",
@@ -88,8 +93,10 @@ func (c *MetricsConfig) Validate() error {
 	}
 
 	switch c.Backend {
-	case "prometheus", "none", "":
+	case "prometheus", "none":
 		// valid
+	case "":
+		return fmt.Errorf("metrics: enabled requires a non-empty backend")
 	case "otel":
 		if c.OTel.Endpoint == "" {
 			return fmt.Errorf("metrics: backend=otel requires a non-empty OTel endpoint")
@@ -100,6 +107,9 @@ func (c *MetricsConfig) Validate() error {
 		if c.OTel.ExportInterval <= 0 {
 			return fmt.Errorf("metrics: otel export interval must be positive")
 		}
+		if c.OTel.Timeout <= 0 {
+			return fmt.Errorf("metrics: otel timeout must be positive")
+		}
 	default:
 		return fmt.Errorf("metrics: unknown backend %q (must be \"prometheus\", \"otel\", or \"none\")", c.Backend)
 	}

internal/observability/metrics.go

@@ -43,20 +43,20 @@ type Histogram interface {
 type Backend interface {
 	// NewCounter creates a counter metric.
 	// labelNames declares the set of label keys that will be passed at observation time.
-	NewCounter(name, desc string, labelNames ...string) Counter
+	NewCounter(name, desc string, labelNames ...string) (Counter, error)
 
 	// NewUpDownCounter creates an up-down counter metric.
-	NewUpDownCounter(name, desc string, labelNames ...string) UpDownCounter
+	NewUpDownCounter(name, desc string, labelNames ...string) (UpDownCounter, error)
 
 	// NewInt64Gauge creates an integer gauge metric.
-	NewInt64Gauge(name, desc string, labelNames ...string) Int64Gauge
+	NewInt64Gauge(name, desc string, labelNames ...string) (Int64Gauge, error)
 
 	// NewFloat64Gauge creates a float gauge metric.
-	NewFloat64Gauge(name, desc string, labelNames ...string) Float64Gauge
+	NewFloat64Gauge(name, desc string, labelNames ...string) (Float64Gauge, error)
 
 	// NewHistogram creates a histogram metric.
 	// buckets are the explicit upper-bound bucket boundaries.
-	NewHistogram(name, desc string, buckets []float64, labelNames ...string) Histogram
+	NewHistogram(name, desc string, buckets []float64, labelNames ...string) (Histogram, error)
 
 	// HTTPHandler returns the /metrics HTTP handler.
 	// Implementations that do not expose an HTTP endpoint return nil.
@@ -88,6 +88,7 @@ func New(cfg MetricsConfig) (Backend, error) {
 			Endpoint:              cfg.OTel.Endpoint,
 			Insecure:              cfg.OTel.Insecure,
 			ExportInterval:        cfg.OTel.ExportInterval,
+			Timeout:               cfg.OTel.Timeout,
 			ServiceName:           cfg.ServiceName,
 			ServiceVersion:        cfg.ServiceVersion,
 			DeploymentEnvironment: cfg.DeploymentEnvironment,
@@ -110,19 +111,19 @@ type promAdapter struct {
 	b *obsprom.Backend
 }
 
-func (a *promAdapter) NewCounter(name, desc string, labelNames ...string) Counter {
+func (a *promAdapter) NewCounter(name, desc string, labelNames ...string) (Counter, error) {
 	return a.b.NewCounter(name, desc, labelNames...)
 }
-func (a *promAdapter) NewUpDownCounter(name, desc string, labelNames ...string) UpDownCounter {
+func (a *promAdapter) NewUpDownCounter(name, desc string, labelNames ...string) (UpDownCounter, error) {
 	return a.b.NewUpDownCounter(name, desc, labelNames...)
 }
-func (a *promAdapter) NewInt64Gauge(name, desc string, labelNames ...string) Int64Gauge {
+func (a *promAdapter) NewInt64Gauge(name, desc string, labelNames ...string) (Int64Gauge, error) {
 	return a.b.NewInt64Gauge(name, desc, labelNames...)
 }
-func (a *promAdapter) NewFloat64Gauge(name, desc string, labelNames ...string) Float64Gauge {
+func (a *promAdapter) NewFloat64Gauge(name, desc string, labelNames ...string) (Float64Gauge, error) {
 	return a.b.NewFloat64Gauge(name, desc, labelNames...)
 }
-func (a *promAdapter) NewHistogram(name, desc string, buckets []float64, labelNames ...string) Histogram {
+func (a *promAdapter) NewHistogram(name, desc string, buckets []float64, labelNames ...string) (Histogram, error) {
 	return a.b.NewHistogram(name, desc, buckets, labelNames...)
 }
 func (a *promAdapter) HTTPHandler() http.Handler          { return a.b.HTTPHandler() }
@@ -133,19 +134,19 @@ type otelAdapter struct {
 	b *obsotel.Backend
 }
 
-func (a *otelAdapter) NewCounter(name, desc string, labelNames ...string) Counter {
+func (a *otelAdapter) NewCounter(name, desc string, labelNames ...string) (Counter, error) {
 	return a.b.NewCounter(name, desc, labelNames...)
 }
-func (a *otelAdapter) NewUpDownCounter(name, desc string, labelNames ...string) UpDownCounter {
+func (a *otelAdapter) NewUpDownCounter(name, desc string, labelNames ...string) (UpDownCounter, error) {
 	return a.b.NewUpDownCounter(name, desc, labelNames...)
 }
-func (a *otelAdapter) NewInt64Gauge(name, desc string, labelNames ...string) Int64Gauge {
+func (a *otelAdapter) NewInt64Gauge(name, desc string, labelNames ...string) (Int64Gauge, error) {
 	return a.b.NewInt64Gauge(name, desc, labelNames...)
 }
-func (a *otelAdapter) NewFloat64Gauge(name, desc string, labelNames ...string) Float64Gauge {
+func (a *otelAdapter) NewFloat64Gauge(name, desc string, labelNames ...string) (Float64Gauge, error) {
 	return a.b.NewFloat64Gauge(name, desc, labelNames...)
 }
-func (a *otelAdapter) NewHistogram(name, desc string, buckets []float64, labelNames ...string) Histogram {
+func (a *otelAdapter) NewHistogram(name, desc string, buckets []float64, labelNames ...string) (Histogram, error) {
 	return a.b.NewHistogram(name, desc, buckets, labelNames...)
 }
 func (a *otelAdapter) HTTPHandler() http.Handler          { return a.b.HTTPHandler() }

internal/observability/metrics_test.go

@@ -2,6 +2,8 @@ package observability_test
 
 import (
 	"context"
+	"net"
+	"os"
 	"testing"
 	"time"
 
@@ -39,20 +41,19 @@ func TestValidateValidConfigs(t *testing.T) {
 	}{
 		{name: "disabled", cfg: observability.MetricsConfig{Enabled: false}},
 		{name: "backend none", cfg: observability.MetricsConfig{Enabled: true, Backend: "none"}},
-		{name: "backend empty", cfg: observability.MetricsConfig{Enabled: true, Backend: ""}},
 		{name: "prometheus", cfg: observability.MetricsConfig{Enabled: true, Backend: "prometheus"}},
 		{
 			name: "otel grpc",
 			cfg: observability.MetricsConfig{
 				Enabled: true, Backend: "otel",
-				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, ExportInterval: 10 * time.Second},
+				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, ExportInterval: 10 * time.Second, Timeout: 2 * time.Second},
 			},
 		},
 		{
 			name: "otel http",
 			cfg: observability.MetricsConfig{
 				Enabled: true, Backend: "otel",
-				OTel: observability.OTelConfig{Protocol: "http", Endpoint: "localhost:4318", ExportInterval: 30 * time.Second},
+				OTel: observability.OTelConfig{Protocol: "http", Endpoint: "localhost:4318", ExportInterval: 30 * time.Second, Timeout: 2 * time.Second},
 			},
 		},
 	}
@@ -71,25 +72,36 @@ func TestValidateInvalidConfigs(t *testing.T) {
 		cfg  observability.MetricsConfig
 	}{
 		{name: "unknown backend", cfg: observability.MetricsConfig{Enabled: true, Backend: "datadog"}},
+		{
+			name: "backend empty while enabled",
+			cfg:  observability.MetricsConfig{Enabled: true, Backend: ""},
+		},
 		{
 			name: "otel missing endpoint",
 			cfg: observability.MetricsConfig{
 				Enabled: true, Backend: "otel",
-				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: "", ExportInterval: 10 * time.Second},
+				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: "", ExportInterval: 10 * time.Second, Timeout: 2 * time.Second},
 			},
 		},
 		{
 			name: "otel invalid protocol",
 			cfg: observability.MetricsConfig{
 				Enabled: true, Backend: "otel",
-				OTel: observability.OTelConfig{Protocol: "tcp", Endpoint: otelGRPCEndpoint, ExportInterval: 10 * time.Second},
+				OTel: observability.OTelConfig{Protocol: "tcp", Endpoint: otelGRPCEndpoint, ExportInterval: 10 * time.Second, Timeout: 2 * time.Second},
 			},
 		},
 		{
 			name: "otel zero interval",
 			cfg: observability.MetricsConfig{
 				Enabled: true, Backend: "otel",
-				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, ExportInterval: 0},
+				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, ExportInterval: 0, Timeout: 2 * time.Second},
+			},
+		},
+		{
+			name: "otel zero timeout",
+			cfg: observability.MetricsConfig{
+				Enabled: true, Backend: "otel",
+				OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, ExportInterval: 10 * time.Second, Timeout: 0},
 			},
 		},
 	}
@@ -157,11 +169,32 @@ func TestPrometheusAdapterAllInstruments(t *testing.T) {
 	ctx := context.Background()
 	labels := observability.Labels{"k": "v"}
 
-	b.NewCounter("prom_adapter_counter_total", "desc", "k").Add(ctx, 1, labels)
+	c, err := b.NewCounter("prom_adapter_counter_total", "desc", "k")
-	b.NewUpDownCounter("prom_adapter_updown", "desc", "k").Add(ctx, 2, labels)
+	if err != nil {
-	b.NewInt64Gauge("prom_adapter_int_gauge", "desc", "k").Record(ctx, 99, labels)
+		t.Fatalf("NewCounter error: %v", err)
-	b.NewFloat64Gauge("prom_adapter_float_gauge", "desc", "k").Record(ctx, 1.23, labels)
+	}
-	b.NewHistogram("prom_adapter_histogram", "desc", []float64{0.1, 1.0}, "k").Record(ctx, 0.5, labels)
+	u, err := b.NewUpDownCounter("prom_adapter_updown", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewUpDownCounter error: %v", err)
+	}
+	ig, err := b.NewInt64Gauge("prom_adapter_int_gauge", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewInt64Gauge error: %v", err)
+	}
+	fg, err := b.NewFloat64Gauge("prom_adapter_float_gauge", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewFloat64Gauge error: %v", err)
+	}
+	h, err := b.NewHistogram("prom_adapter_histogram", "desc", []float64{0.1, 1.0}, "k")
+	if err != nil {
+		t.Fatalf("NewHistogram error: %v", err)
+	}
+
+	c.Add(ctx, 1, labels)
+	u.Add(ctx, 2, labels)
+	ig.Record(ctx, 99, labels)
+	fg.Record(ctx, 1.23, labels)
+	h.Record(ctx, 0.5, labels)
 
 	if b.HTTPHandler() == nil {
 		t.Error("prometheus adapter HTTPHandler should not be nil")
@@ -172,9 +205,20 @@ func TestPrometheusAdapterAllInstruments(t *testing.T) {
 }
 
 func TestOtelAdapterAllInstruments(t *testing.T) {
+	if os.Getenv("SKIP_OTEL_INTEGRATION") != "" {
+		t.Skip("skipping OTel integration test because SKIP_OTEL_INTEGRATION is set")
+	}
+
+	dialTimeout := 300 * time.Millisecond
+	conn, err := net.DialTimeout("tcp", otelGRPCEndpoint, dialTimeout)
+	if err != nil {
+		t.Skipf("skipping OTel integration test; collector %s not reachable: %v", otelGRPCEndpoint, err)
+	}
+	_ = conn.Close()
+
 	b, err := observability.New(observability.MetricsConfig{
 		Enabled: true, Backend: "otel",
-		OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, Insecure: true, ExportInterval: 100 * time.Millisecond},
+		OTel: observability.OTelConfig{Protocol: "grpc", Endpoint: otelGRPCEndpoint, Insecure: true, ExportInterval: 100 * time.Millisecond, Timeout: 2 * time.Second},
 	})
 	if err != nil {
 		t.Fatalf("failed to create otel backend: %v", err)
@@ -182,11 +226,32 @@ func TestOtelAdapterAllInstruments(t *testing.T) {
 	ctx := context.Background()
 	labels := observability.Labels{"k": "v"}
 
-	b.NewCounter("otel_adapter_counter_total", "desc", "k").Add(ctx, 1, labels)
+	c, err := b.NewCounter("otel_adapter_counter_total", "desc", "k")
-	b.NewUpDownCounter("otel_adapter_updown", "desc", "k").Add(ctx, 2, labels)
+	if err != nil {
-	b.NewInt64Gauge("otel_adapter_int_gauge", "desc", "k").Record(ctx, 99, labels)
+		t.Fatalf("NewCounter error: %v", err)
-	b.NewFloat64Gauge("otel_adapter_float_gauge", "desc", "k").Record(ctx, 1.23, labels)
+	}
-	b.NewHistogram("otel_adapter_histogram", "desc", []float64{0.1, 1.0}, "k").Record(ctx, 0.5, labels)
+	u, err := b.NewUpDownCounter("otel_adapter_updown", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewUpDownCounter error: %v", err)
+	}
+	ig, err := b.NewInt64Gauge("otel_adapter_int_gauge", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewInt64Gauge error: %v", err)
+	}
+	fg, err := b.NewFloat64Gauge("otel_adapter_float_gauge", "desc", "k")
+	if err != nil {
+		t.Fatalf("NewFloat64Gauge error: %v", err)
+	}
+	h, err := b.NewHistogram("otel_adapter_histogram", "desc", []float64{0.1, 1.0}, "k")
+	if err != nil {
+		t.Fatalf("NewHistogram error: %v", err)
+	}
+
+	c.Add(ctx, 1, labels)
+	u.Add(ctx, 2, labels)
+	ig.Record(ctx, 99, labels)
+	fg.Record(ctx, 1.23, labels)
+	h.Record(ctx, 0.5, labels)
 
 	if b.HTTPHandler() != nil {
 		t.Error("OTel adapter HTTPHandler should be nil")

internal/observability/noop.go

@@ -13,38 +13,31 @@ type NoopBackend struct{}
 // Compile-time interface check.
 var _ Backend = (*NoopBackend)(nil)
 
-func (n *NoopBackend) NewCounter(_ string, _ string, _ ...string) Counter {
+func (n *NoopBackend) NewCounter(_ string, _ string, _ ...string) (Counter, error) {
-	_ = n
+	return noopCounter{}, nil
-	return noopCounter{}
 }
 
-func (n *NoopBackend) NewUpDownCounter(_ string, _ string, _ ...string) UpDownCounter {
+func (n *NoopBackend) NewUpDownCounter(_ string, _ string, _ ...string) (UpDownCounter, error) {
-	_ = n
+	return noopUpDownCounter{}, nil
-	return noopUpDownCounter{}
 }
 
-func (n *NoopBackend) NewInt64Gauge(_ string, _ string, _ ...string) Int64Gauge {
+func (n *NoopBackend) NewInt64Gauge(_ string, _ string, _ ...string) (Int64Gauge, error) {
-	_ = n
+	return noopInt64Gauge{}, nil
-	return noopInt64Gauge{}
 }
 
-func (n *NoopBackend) NewFloat64Gauge(_ string, _ string, _ ...string) Float64Gauge {
+func (n *NoopBackend) NewFloat64Gauge(_ string, _ string, _ ...string) (Float64Gauge, error) {
-	_ = n
+	return noopFloat64Gauge{}, nil
-	return noopFloat64Gauge{}
 }
 
-func (n *NoopBackend) NewHistogram(_ string, _ string, _ []float64, _ ...string) Histogram {
+func (n *NoopBackend) NewHistogram(_ string, _ string, _ []float64, _ ...string) (Histogram, error) {
-	_ = n
+	return noopHistogram{}, nil
-	return noopHistogram{}
 }
 
 func (n *NoopBackend) HTTPHandler() http.Handler {
-	_ = n
 	return nil
 }
 
 func (n *NoopBackend) Shutdown(_ context.Context) error {
-	_ = n
 	return nil
 }
 

internal/observability/noop_test.go

@@ -13,32 +13,32 @@ func TestNoopBackendAllInstruments(t *testing.T) {
 	ctx := context.Background()
 	labels := observability.Labels{"k": "v"}
 
-	t.Run("Counter", func(_ *testing.T) {
+	t.Run("Counter", func(t *testing.T) {
-		c := n.NewCounter("test_counter", "desc")
+		c, _ := n.NewCounter("test_counter", "desc")
 		c.Add(ctx, 1, labels)
 		c.Add(ctx, 0, nil)
 	})
 
-	t.Run("UpDownCounter", func(_ *testing.T) {
+	t.Run("UpDownCounter", func(t *testing.T) {
-		u := n.NewUpDownCounter("test_updown", "desc")
+		u, _ := n.NewUpDownCounter("test_updown", "desc")
 		u.Add(ctx, 1, labels)
 		u.Add(ctx, -1, nil)
 	})
 
-	t.Run("Int64Gauge", func(_ *testing.T) {
+	t.Run("Int64Gauge", func(t *testing.T) {
-		g := n.NewInt64Gauge("test_int64gauge", "desc")
+		g, _ := n.NewInt64Gauge("test_int64gauge", "desc")
 		g.Record(ctx, 42, labels)
 		g.Record(ctx, 0, nil)
 	})
 
-	t.Run("Float64Gauge", func(_ *testing.T) {
+	t.Run("Float64Gauge", func(t *testing.T) {
-		g := n.NewFloat64Gauge("test_float64gauge", "desc")
+		g, _ := n.NewFloat64Gauge("test_float64gauge", "desc")
 		g.Record(ctx, 3.14, labels)
 		g.Record(ctx, 0, nil)
 	})
 
-	t.Run("Histogram", func(_ *testing.T) {
+	t.Run("Histogram", func(t *testing.T) {
-		h := n.NewHistogram("test_histogram", "desc", []float64{1, 5, 10})
+		h, _ := n.NewHistogram("test_histogram", "desc", []float64{1, 5, 10})
 		h.Record(ctx, 2.5, labels)
 		h.Record(ctx, 0, nil)
 	})
@@ -56,12 +56,47 @@ func TestNoopBackendAllInstruments(t *testing.T) {
 	})
 }
 
-func TestNoopBackendLabelNames(_ *testing.T) {
+func TestNoopBackendLabelNames(t *testing.T) {
 	// Verify that label names passed at creation time are accepted without panic.
 	n := &observability.NoopBackend{}
-	n.NewCounter("c", "d", "label1", "label2")
+
-	n.NewUpDownCounter("u", "d", "l1")
+	assertNoPanic := func(t *testing.T, constructor string, fn func()) {
-	n.NewInt64Gauge("g1", "d", "l1", "l2", "l3")
+		t.Helper()
-	n.NewFloat64Gauge("g2", "d")
+		defer func() {
-	n.NewHistogram("h", "d", []float64{0.1, 1.0}, "l1")
+			if r := recover(); r != nil {
+				t.Fatalf("%s panicked: %v", constructor, r)
+			}
+		}()
+		fn()
+	}
+
+	t.Run("NewCounter", func(t *testing.T) {
+		assertNoPanic(t, "NewCounter", func() {
+			_, _ = n.NewCounter("c", "d", "label1", "label2")
+		})
+	})
+
+	t.Run("NewUpDownCounter", func(t *testing.T) {
+		assertNoPanic(t, "NewUpDownCounter", func() {
+			_, _ = n.NewUpDownCounter("u", "d", "l1")
+		})
+	})
+
+	t.Run("NewInt64Gauge", func(t *testing.T) {
+		assertNoPanic(t, "NewInt64Gauge", func() {
+			_, _ = n.NewInt64Gauge("g1", "d", "l1", "l2", "l3")
+		})
+	})
+
+	t.Run("NewFloat64Gauge", func(t *testing.T) {
+		assertNoPanic(t, "NewFloat64Gauge", func() {
+			_, _ = n.NewFloat64Gauge("g2", "d")
+		})
+	})
+
+	t.Run("NewHistogram", func(t *testing.T) {
+		assertNoPanic(t, "NewHistogram", func() {
+			_, _ = n.NewHistogram("h", "d", []float64{0.1, 1.0}, "l1")
+		})
+	})
 }

internal/observability/otel/backend.go

@@ -9,7 +9,10 @@ package otel
 import (
 	"context"
 	"fmt"
+	"log"
 	"net/http"
+	"regexp"
+	"strings"
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
@@ -17,6 +20,8 @@ import (
 	sdkmetric "go.opentelemetry.io/otel/sdk/metric"
 )
 
+var metricLabelNameRE = regexp.MustCompile(`^[a-zA-Z_][a-zA-Z0-9_]*$`)
+
 // Config holds OTel backend configuration.
 type Config struct {
 	// Protocol is "grpc" (default) or "http".
@@ -31,6 +36,9 @@ type Config struct {
 	// ExportInterval is the period between pushes to the collector.
 	ExportInterval time.Duration
 
+	// Timeout bounds exporter construction calls.
+	Timeout time.Duration
+
 	ServiceName           string
 	ServiceVersion        string
 	DeploymentEnvironment string
@@ -57,9 +65,15 @@ func New(cfg Config) (*Backend, error) {
 	if cfg.Protocol == "" {
 		cfg.Protocol = "grpc"
 	}
+	if strings.TrimSpace(cfg.Endpoint) == "" {
+		return nil, fmt.Errorf("otel backend: empty cfg.Endpoint")
+	}
 	if cfg.ExportInterval <= 0 {
 		cfg.ExportInterval = 60 * time.Second
 	}
+	if cfg.Timeout <= 0 {
+		cfg.Timeout = 10 * time.Second
+	}
 	if cfg.ServiceName == "" {
 		cfg.ServiceName = "gerbil"
 	}
@@ -100,111 +114,196 @@ func (b *Backend) Shutdown(ctx context.Context) error {
 }
 
 // NewCounter creates an OTel Int64Counter.
-func (b *Backend) NewCounter(name, desc string, _ ...string) *Counter {
+func (b *Backend) NewCounter(name, desc string, labelNames ...string) (*Counter, error) {
+	normalizedLabelNames, err := validateLabelNames(labelNames)
+	if err != nil {
+		return nil, fmt.Errorf("otel: create counter %q: %w", name, err)
+	}
 	c, err := b.meter.Int64Counter(name, metric.WithDescription(desc))
 	if err != nil {
-		panic(fmt.Sprintf("otel: create counter %q: %v", name, err))
+		return nil, fmt.Errorf("otel: create counter %q: %w", name, err)
 	}
-	return &Counter{c: c}
+	return &Counter{c: c, labelNames: normalizedLabelNames}, nil
 }
 
 // NewUpDownCounter creates an OTel Int64UpDownCounter.
-func (b *Backend) NewUpDownCounter(name, desc string, _ ...string) *UpDownCounter {
+func (b *Backend) NewUpDownCounter(name, desc string, labelNames ...string) (*UpDownCounter, error) {
+	normalizedLabelNames, err := validateLabelNames(labelNames)
+	if err != nil {
+		return nil, fmt.Errorf("otel: create up-down counter %q: %w", name, err)
+	}
 	c, err := b.meter.Int64UpDownCounter(name, metric.WithDescription(desc))
 	if err != nil {
-		panic(fmt.Sprintf("otel: create up-down counter %q: %v", name, err))
+		return nil, fmt.Errorf("otel: create up-down counter %q: %w", name, err)
 	}
-	return &UpDownCounter{c: c}
+	return &UpDownCounter{c: c, labelNames: normalizedLabelNames}, nil
 }
 
 // NewInt64Gauge creates an OTel Int64Gauge.
-func (b *Backend) NewInt64Gauge(name, desc string, _ ...string) *Int64Gauge {
+func (b *Backend) NewInt64Gauge(name, desc string, labelNames ...string) (*Int64Gauge, error) {
+	normalizedLabelNames, err := validateLabelNames(labelNames)
+	if err != nil {
+		return nil, fmt.Errorf("otel: create int64 gauge %q: %w", name, err)
+	}
 	g, err := b.meter.Int64Gauge(name, metric.WithDescription(desc))
 	if err != nil {
-		panic(fmt.Sprintf("otel: create int64 gauge %q: %v", name, err))
+		return nil, fmt.Errorf("otel: create int64 gauge %q: %w", name, err)
 	}
-	return &Int64Gauge{g: g}
+	return &Int64Gauge{g: g, labelNames: normalizedLabelNames}, nil
 }
 
 // NewFloat64Gauge creates an OTel Float64Gauge.
-func (b *Backend) NewFloat64Gauge(name, desc string, _ ...string) *Float64Gauge {
+func (b *Backend) NewFloat64Gauge(name, desc string, labelNames ...string) (*Float64Gauge, error) {
+	normalizedLabelNames, err := validateLabelNames(labelNames)
+	if err != nil {
+		return nil, fmt.Errorf("otel: create float64 gauge %q: %w", name, err)
+	}
 	g, err := b.meter.Float64Gauge(name, metric.WithDescription(desc))
 	if err != nil {
-		panic(fmt.Sprintf("otel: create float64 gauge %q: %v", name, err))
+		return nil, fmt.Errorf("otel: create float64 gauge %q: %w", name, err)
 	}
-	return &Float64Gauge{g: g}
+	return &Float64Gauge{g: g, labelNames: normalizedLabelNames}, nil
 }
 
 // NewHistogram creates an OTel Float64Histogram with explicit bucket boundaries.
-func (b *Backend) NewHistogram(name, desc string, buckets []float64, _ ...string) *Histogram {
+func (b *Backend) NewHistogram(name, desc string, buckets []float64, labelNames ...string) (*Histogram, error) {
+	normalizedLabelNames, err := validateLabelNames(labelNames)
+	if err != nil {
+		return nil, fmt.Errorf("otel: create histogram %q: %w", name, err)
+	}
 	h, err := b.meter.Float64Histogram(name,
 		metric.WithDescription(desc),
 		metric.WithExplicitBucketBoundaries(buckets...),
 	)
 	if err != nil {
-		panic(fmt.Sprintf("otel: create histogram %q: %v", name, err))
+		return nil, fmt.Errorf("otel: create histogram %q: %w", name, err)
 	}
-	return &Histogram{h: h}
+	return &Histogram{h: h, labelNames: normalizedLabelNames}, nil
 }
 
-// labelsToAttrs converts a Labels map to OTel attribute key-value pairs.
+func validateLabelNames(labelNames []string) ([]string, error) {
-func labelsToAttrs(labels map[string]string) []attribute.KeyValue {
+	if len(labelNames) == 0 {
-	if len(labels) == 0 {
+		return nil, nil
-		return nil
 	}
-	attrs := make([]attribute.KeyValue, 0, len(labels))
+
-	for k, v := range labels {
+	normalized := make([]string, len(labelNames))
-		attrs = append(attrs, attribute.String(k, v))
+	seen := make(map[string]struct{}, len(labelNames))
+	for i, name := range labelNames {
+		if !metricLabelNameRE.MatchString(name) {
+			return nil, fmt.Errorf("invalid label name %q", name)
+		}
+		if _, exists := seen[name]; exists {
+			return nil, fmt.Errorf("duplicate label name %q", name)
+		}
+		seen[name] = struct{}{}
+		normalized[i] = name
 	}
+
+	return normalized, nil
+}
+
+func labelsToAttrs(labelNames []string, labels map[string]string) []attribute.KeyValue {
+	if len(labelNames) == 0 {
+		if len(labels) > 0 {
+			log.Printf("WARN: dropping otel metric sample due to unexpected labels: got=%v expected=none", labels)
+			return nil
+		}
+		return []attribute.KeyValue{}
+	}
+
+	attrs := make([]attribute.KeyValue, 0, len(labelNames))
+	for _, labelName := range labelNames {
+		attrs = append(attrs, attribute.String(labelName, labels[labelName]))
+	}
+
+	for got := range labels {
+		found := false
+		for _, expected := range labelNames {
+			if got == expected {
+				found = true
+				break
+			}
+		}
+		if !found {
+			log.Printf("WARN: dropping otel metric sample due to unexpected label key %q (expected=%v)", got, labelNames)
+			return nil
+		}
+	}
+
 	return attrs
 }
 
 // Counter wraps an OTel Int64Counter.
 type Counter struct {
-	c metric.Int64Counter
+	c          metric.Int64Counter
+	labelNames []string
 }
 
 // Add increments the counter by value.
 func (c *Counter) Add(ctx context.Context, value int64, labels map[string]string) {
-	c.c.Add(ctx, value, metric.WithAttributes(labelsToAttrs(labels)...))
+	attrs := labelsToAttrs(c.labelNames, labels)
+	if attrs == nil {
+		return
+	}
+	c.c.Add(ctx, value, metric.WithAttributes(attrs...))
 }
 
 // UpDownCounter wraps an OTel Int64UpDownCounter.
 type UpDownCounter struct {
-	c metric.Int64UpDownCounter
+	c          metric.Int64UpDownCounter
+	labelNames []string
 }
 
 // Add adjusts the up-down counter by value.
 func (u *UpDownCounter) Add(ctx context.Context, value int64, labels map[string]string) {
-	u.c.Add(ctx, value, metric.WithAttributes(labelsToAttrs(labels)...))
+	attrs := labelsToAttrs(u.labelNames, labels)
+	if attrs == nil {
+		return
+	}
+	u.c.Add(ctx, value, metric.WithAttributes(attrs...))
 }
 
 // Int64Gauge wraps an OTel Int64Gauge.
 type Int64Gauge struct {
-	g metric.Int64Gauge
+	g          metric.Int64Gauge
+	labelNames []string
 }
 
 // Record sets the gauge to value.
 func (g *Int64Gauge) Record(ctx context.Context, value int64, labels map[string]string) {
-	g.g.Record(ctx, value, metric.WithAttributes(labelsToAttrs(labels)...))
+	attrs := labelsToAttrs(g.labelNames, labels)
+	if attrs == nil {
+		return
+	}
+	g.g.Record(ctx, value, metric.WithAttributes(attrs...))
 }
 
 // Float64Gauge wraps an OTel Float64Gauge.
 type Float64Gauge struct {
-	g metric.Float64Gauge
+	g          metric.Float64Gauge
+	labelNames []string
 }
 
 // Record sets the gauge to value.
 func (g *Float64Gauge) Record(ctx context.Context, value float64, labels map[string]string) {
-	g.g.Record(ctx, value, metric.WithAttributes(labelsToAttrs(labels)...))
+	attrs := labelsToAttrs(g.labelNames, labels)
+	if attrs == nil {
+		return
+	}
+	g.g.Record(ctx, value, metric.WithAttributes(attrs...))
 }
 
 // Histogram wraps an OTel Float64Histogram.
 type Histogram struct {
-	h metric.Float64Histogram
+	h          metric.Float64Histogram
+	labelNames []string
 }
 
 // Record observes value in the histogram.
 func (h *Histogram) Record(ctx context.Context, value float64, labels map[string]string) {
-	h.h.Record(ctx, value, metric.WithAttributes(labelsToAttrs(labels)...))
+	attrs := labelsToAttrs(h.labelNames, labels)
+	if attrs == nil {
+		return
+	}
+	h.h.Record(ctx, value, metric.WithAttributes(attrs...))
 }

internal/observability/otel/backend_test.go

@@ -55,7 +55,10 @@ func TestOtelBackendCounter(t *testing.T) {
 	b := newInMemoryBackend(t)
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 
-	c := b.NewCounter("gerbil_test_counter_total", "test counter", "result")
+	c, err := b.NewCounter("gerbil_test_counter_total", "test counter", "result")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
 	// Should not panic
 	c.Add(context.Background(), 1, map[string]string{"result": "ok"})
 	c.Add(context.Background(), 5, nil)
@@ -65,7 +68,10 @@ func TestOtelBackendUpDownCounter(t *testing.T) {
 	b := newInMemoryBackend(t)
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 
-	u := b.NewUpDownCounter("gerbil_test_updown", "test updown", "state")
+	u, err := b.NewUpDownCounter("gerbil_test_updown", "test updown", "state")
+	if err != nil {
+		t.Fatalf("NewUpDownCounter returned error: %v", err)
+	}
 	u.Add(context.Background(), 3, map[string]string{"state": "active"})
 	u.Add(context.Background(), -1, map[string]string{"state": "active"})
 }
@@ -74,7 +80,10 @@ func TestOtelBackendInt64Gauge(t *testing.T) {
 	b := newInMemoryBackend(t)
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 
-	g := b.NewInt64Gauge("gerbil_test_int_gauge", "test gauge")
+	g, err := b.NewInt64Gauge("gerbil_test_int_gauge", "test gauge")
+	if err != nil {
+		t.Fatalf("NewInt64Gauge returned error: %v", err)
+	}
 	g.Record(context.Background(), 42, nil)
 }
 
@@ -82,7 +91,10 @@ func TestOtelBackendFloat64Gauge(t *testing.T) {
 	b := newInMemoryBackend(t)
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 
-	g := b.NewFloat64Gauge("gerbil_test_float_gauge", "test float gauge")
+	g, err := b.NewFloat64Gauge("gerbil_test_float_gauge", "test float gauge")
+	if err != nil {
+		t.Fatalf("NewFloat64Gauge returned error: %v", err)
+	}
 	g.Record(context.Background(), 3.14, nil)
 }
 
@@ -90,8 +102,11 @@ func TestOtelBackendHistogram(t *testing.T) {
 	b := newInMemoryBackend(t)
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 
-	h := b.NewHistogram("gerbil_test_duration_seconds", "test histogram",
+	h, err := b.NewHistogram("gerbil_test_duration_seconds", "test histogram",
 		[]float64{0.1, 0.5, 1.0}, "method")
+	if err != nil {
+		t.Fatalf("NewHistogram returned error: %v", err)
+	}
 	h.Record(context.Background(), 0.3, map[string]string{"method": "GET"})
 }
 
@@ -139,3 +154,22 @@ func TestOtelBackendDeploymentEnvironment(t *testing.T) {
 	}
 	defer b.Shutdown(context.Background()) //nolint:errcheck
 }
+
+func TestOtelBackendRejectsInvalidLabelNames(t *testing.T) {
+	b := newInMemoryBackend(t)
+	defer b.Shutdown(context.Background()) //nolint:errcheck
+
+	t.Run("duplicate labels", func(t *testing.T) {
+		_, err := b.NewCounter("gerbil_test_invalid_labels_total", "test counter", "result", "result")
+		if err == nil {
+			t.Fatal("expected error for duplicate label names")
+		}
+	})
+
+	t.Run("invalid label name", func(t *testing.T) {
+		_, err := b.NewHistogram("gerbil_test_invalid_histogram", "test histogram", []float64{0.1, 1.0}, "status-code")
+		if err == nil {
+			t.Fatal("expected error for invalid label name")
+		}
+	})
+}

internal/observability/otel/exporter.go

@@ -3,6 +3,8 @@ package otel
 import (
 	"context"
 	"fmt"
+	"net/url"
+	"strings"
 
 	"go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc"
 	"go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetrichttp"
@@ -11,6 +13,10 @@ import (
 
 // newExporter creates the appropriate OTLP exporter based on cfg.Protocol.
 func newExporter(ctx context.Context, cfg Config) (sdkmetric.Exporter, error) {
+	if strings.TrimSpace(cfg.Endpoint) == "" {
+		return nil, fmt.Errorf("otel: cfg.Endpoint is empty")
+	}
+
 	switch cfg.Protocol {
 	case "grpc", "":
 		return newGRPCExporter(ctx, cfg)
@@ -36,8 +42,20 @@ func newGRPCExporter(ctx context.Context, cfg Config) (sdkmetric.Exporter, error
 }
 
 func newHTTPExporter(ctx context.Context, cfg Config) (sdkmetric.Exporter, error) {
-	opts := []otlpmetrichttp.Option{
+	endpoint := strings.TrimSpace(cfg.Endpoint)
-		otlpmetrichttp.WithEndpoint(cfg.Endpoint),
+
+	opts := make([]otlpmetrichttp.Option, 0, 3)
+	if strings.Contains(endpoint, "://") {
+		parsed, err := url.Parse(endpoint)
+		if err != nil {
+			return nil, fmt.Errorf("otlp http exporter: parse endpoint URL %q: %w", endpoint, err)
+		}
+		opts = append(opts, otlpmetrichttp.WithEndpointURL(parsed.String()))
+	} else {
+		opts = append(opts,
+			otlpmetrichttp.WithEndpoint(endpoint),
+			otlpmetrichttp.WithURLPath("/v1/metrics"),
+		)
 	}
 	if cfg.Insecure {
 		opts = append(opts, otlpmetrichttp.WithInsecure())

internal/observability/otel/resource.go

@@ -3,7 +3,7 @@ package otel
 import (
 	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/sdk/resource"
-	semconv "go.opentelemetry.io/otel/semconv/v1.40.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.18.0"
 )
 
 // newResource builds an OTel resource for the Gerbil service.
@@ -15,11 +15,11 @@ func newResource(serviceName, serviceVersion, deploymentEnv string) (*resource.R
 		attrs = append(attrs, semconv.ServiceVersion(serviceVersion))
 	}
 	if deploymentEnv != "" {
-		attrs = append(attrs, semconv.DeploymentEnvironmentName(deploymentEnv))
+		attrs = append(attrs, semconv.DeploymentEnvironment(deploymentEnv))
 	}
 
 	return resource.Merge(
 		resource.Default(),
-		resource.NewWithAttributes(semconv.SchemaURL, attrs...),
+		resource.NewSchemaless(attrs...),
 	)
 }

internal/observability/prometheus/backend.go

@@ -7,6 +7,7 @@ package prometheus
 
 import (
 	"context"
+	"log"
 	"net/http"
 
 	"github.com/prometheus/client_golang/prometheus"
@@ -30,9 +31,10 @@ type Config struct {
 // in the backend-specific instrument types that implement the observability
 // instrument interfaces.
 type Backend struct {
-	cfg      Config
+	cfg                   Config
-	registry *prometheus.Registry
+	registry              *prometheus.Registry
-	handler  http.Handler
+	handler               http.Handler
+	droppedSamplesCounter prometheus.Counter
 }
 
 // New creates and initialises a Prometheus backend.
@@ -48,6 +50,11 @@ func New(cfg Config) (*Backend, error) {
 	}
 
 	registry := prometheus.NewRegistry()
+	droppedSamplesCounter := prometheus.NewCounter(prometheus.CounterOpts{
+		Name: "gerbil_dropped_metric_samples_total",
+		Help: "Total number of metric samples dropped due to invalid labels or unsupported label sets",
+	})
+	registry.MustRegister(droppedSamplesCounter)
 
 	// Include Go and process metrics by default.
 	includeGo := cfg.IncludeGoMetrics == nil || *cfg.IncludeGoMetrics
@@ -62,7 +69,7 @@ func New(cfg Config) (*Backend, error) {
 		EnableOpenMetrics: false,
 	})
 
-	return &Backend{cfg: cfg, registry: registry, handler: handler}, nil
+	return &Backend{cfg: cfg, registry: registry, handler: handler, droppedSamplesCounter: droppedSamplesCounter}, nil
 }
 
 // HTTPHandler returns the Prometheus /metrics HTTP handler.
@@ -78,60 +85,107 @@ func (b *Backend) Shutdown(_ context.Context) error {
 }
 
 // NewCounter creates a Prometheus CounterVec registered on the backend's registry.
-func (b *Backend) NewCounter(name, desc string, labelNames ...string) *Counter {
+func (b *Backend) NewCounter(name, desc string, labelNames ...string) (*Counter, error) {
 	vec := prometheus.NewCounterVec(prometheus.CounterOpts{
 		Name: name,
 		Help: desc,
 	}, labelNames)
-	b.registry.MustRegister(vec)
+	if err := b.registry.Register(vec); err != nil {
-	return &Counter{vec: vec}
+		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
+			existing, ok := are.ExistingCollector.(*prometheus.CounterVec)
+			if !ok {
+				return nil, err
+			}
+			return &Counter{vec: existing, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
+		}
+		return nil, err
+	}
+	return &Counter{vec: vec, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
 }
 
 // NewUpDownCounter creates a Prometheus GaugeVec (Prometheus gauges are
 // bidirectional) registered on the backend's registry.
-func (b *Backend) NewUpDownCounter(name, desc string, labelNames ...string) *UpDownCounter {
+func (b *Backend) NewUpDownCounter(name, desc string, labelNames ...string) (*UpDownCounter, error) {
 	vec := prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Name: name,
 		Help: desc,
 	}, labelNames)
-	b.registry.MustRegister(vec)
+	if err := b.registry.Register(vec); err != nil {
-	return &UpDownCounter{vec: vec}
+		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
+			existing, ok := are.ExistingCollector.(*prometheus.GaugeVec)
+			if !ok {
+				return nil, err
+			}
+			return &UpDownCounter{vec: existing, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
+		}
+		return nil, err
+	}
+	return &UpDownCounter{vec: vec, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
 }
 
 // NewInt64Gauge creates a Prometheus GaugeVec registered on the backend's registry.
-func (b *Backend) NewInt64Gauge(name, desc string, labelNames ...string) *Int64Gauge {
+func (b *Backend) NewInt64Gauge(name, desc string, labelNames ...string) (*Int64Gauge, error) {
 	vec := prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Name: name,
 		Help: desc,
 	}, labelNames)
-	b.registry.MustRegister(vec)
+	if err := b.registry.Register(vec); err != nil {
-	return &Int64Gauge{vec: vec}
+		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
+			existing, ok := are.ExistingCollector.(*prometheus.GaugeVec)
+			if !ok {
+				return nil, err
+			}
+			return &Int64Gauge{vec: existing, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
+		}
+		return nil, err
+	}
+	return &Int64Gauge{vec: vec, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
 }
 
 // NewFloat64Gauge creates a Prometheus GaugeVec registered on the backend's registry.
-func (b *Backend) NewFloat64Gauge(name, desc string, labelNames ...string) *Float64Gauge {
+func (b *Backend) NewFloat64Gauge(name, desc string, labelNames ...string) (*Float64Gauge, error) {
 	vec := prometheus.NewGaugeVec(prometheus.GaugeOpts{
 		Name: name,
 		Help: desc,
 	}, labelNames)
-	b.registry.MustRegister(vec)
+	if err := b.registry.Register(vec); err != nil {
-	return &Float64Gauge{vec: vec}
+		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
+			existing, ok := are.ExistingCollector.(*prometheus.GaugeVec)
+			if !ok {
+				return nil, err
+			}
+			return &Float64Gauge{vec: existing, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
+		}
+		return nil, err
+	}
+	return &Float64Gauge{vec: vec, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
 }
 
 // NewHistogram creates a Prometheus HistogramVec registered on the backend's registry.
-func (b *Backend) NewHistogram(name, desc string, buckets []float64, labelNames ...string) *Histogram {
+func (b *Backend) NewHistogram(name, desc string, buckets []float64, labelNames ...string) (*Histogram, error) {
 	vec := prometheus.NewHistogramVec(prometheus.HistogramOpts{
 		Name:    name,
 		Help:    desc,
 		Buckets: buckets,
 	}, labelNames)
-	b.registry.MustRegister(vec)
+	if err := b.registry.Register(vec); err != nil {
-	return &Histogram{vec: vec}
+		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
+			existing, ok := are.ExistingCollector.(*prometheus.HistogramVec)
+			if !ok {
+				return nil, err
+			}
+			return &Histogram{vec: existing, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
+		}
+		return nil, err
+	}
+	return &Histogram{vec: vec, labelNames: append([]string(nil), labelNames...), droppedSamplesCounter: b.droppedSamplesCounter}, nil
 }
 
 // Counter is a native Prometheus counter instrument.
 type Counter struct {
-	vec *prometheus.CounterVec
+	vec                   *prometheus.CounterVec
+	labelNames            []string
+	droppedSamplesCounter prometheus.Counter
 }
 
 // Add increments the counter by value for the given labels.
@@ -139,47 +193,118 @@ type Counter struct {
 // value must be non-negative. Negative values are ignored.
 func (c *Counter) Add(_ context.Context, value int64, labels map[string]string) {
 	if value < 0 {
+		log.Printf("WARN: counter add called with negative value=%d labels=%v expected_labels=%v", value, labels, c.labelNames)
 		return
 	}
-	c.vec.With(prometheus.Labels(labels)).Add(float64(value))
+	normalized, ok := normalizeLabels(c.labelNames, labels, c.droppedSamplesCounter)
+	if !ok {
+		return
+	}
+	defer guardMetricPanic("counter", c.labelNames, labels)
+	c.vec.With(normalized).Add(float64(value))
 }
 
 // UpDownCounter is a native Prometheus gauge used as a bidirectional counter.
 type UpDownCounter struct {
-	vec *prometheus.GaugeVec
+	vec                   *prometheus.GaugeVec
+	labelNames            []string
+	droppedSamplesCounter prometheus.Counter
 }
 
 // Add adjusts the gauge by value for the given labels.
 func (u *UpDownCounter) Add(_ context.Context, value int64, labels map[string]string) {
-	u.vec.With(prometheus.Labels(labels)).Add(float64(value))
+	normalized, ok := normalizeLabels(u.labelNames, labels, u.droppedSamplesCounter)
+	if !ok {
+		return
+	}
+	defer guardMetricPanic("updown", u.labelNames, labels)
+	u.vec.With(normalized).Add(float64(value))
 }
 
 // Int64Gauge is a native Prometheus gauge recording integer snapshot values.
 type Int64Gauge struct {
-	vec *prometheus.GaugeVec
+	vec                   *prometheus.GaugeVec
+	labelNames            []string
+	droppedSamplesCounter prometheus.Counter
 }
 
 // Record sets the gauge to value for the given labels.
 func (g *Int64Gauge) Record(_ context.Context, value int64, labels map[string]string) {
-	g.vec.With(prometheus.Labels(labels)).Set(float64(value))
+	normalized, ok := normalizeLabels(g.labelNames, labels, g.droppedSamplesCounter)
+	if !ok {
+		return
+	}
+	defer guardMetricPanic("int64-gauge", g.labelNames, labels)
+	g.vec.With(normalized).Set(float64(value))
 }
 
 // Float64Gauge is a native Prometheus gauge recording float snapshot values.
 type Float64Gauge struct {
-	vec *prometheus.GaugeVec
+	vec                   *prometheus.GaugeVec
+	labelNames            []string
+	droppedSamplesCounter prometheus.Counter
 }
 
 // Record sets the gauge to value for the given labels.
 func (g *Float64Gauge) Record(_ context.Context, value float64, labels map[string]string) {
-	g.vec.With(prometheus.Labels(labels)).Set(value)
+	normalized, ok := normalizeLabels(g.labelNames, labels, g.droppedSamplesCounter)
+	if !ok {
+		return
+	}
+	defer guardMetricPanic("float64-gauge", g.labelNames, labels)
+	g.vec.With(normalized).Set(value)
 }
 
 // Histogram is a native Prometheus histogram instrument.
 type Histogram struct {
-	vec *prometheus.HistogramVec
+	vec                   *prometheus.HistogramVec
+	labelNames            []string
+	droppedSamplesCounter prometheus.Counter
 }
 
 // Record observes value for the given labels.
 func (h *Histogram) Record(_ context.Context, value float64, labels map[string]string) {
-	h.vec.With(prometheus.Labels(labels)).Observe(value)
+	normalized, ok := normalizeLabels(h.labelNames, labels, h.droppedSamplesCounter)
+	if !ok {
+		return
+	}
+	defer guardMetricPanic("histogram", h.labelNames, labels)
+	h.vec.With(normalized).Observe(value)
+}
+
+func normalizeLabels(labelNames []string, labels map[string]string, droppedSamplesCounter prometheus.Counter) (prometheus.Labels, bool) {
+	if len(labelNames) == 0 {
+		if len(labels) > 0 {
+			if droppedSamplesCounter != nil {
+				droppedSamplesCounter.Inc()
+			}
+			log.Printf("WARN: dropping metric sample due to unexpected labels: got=%v expected=none", labels)
+			return nil, false
+		}
+		return nil, true
+	}
+
+	normalized := make(prometheus.Labels, len(labelNames))
+	for _, name := range labelNames {
+		normalized[name] = ""
+	}
+
+	for k, v := range labels {
+		if _, ok := normalized[k]; !ok {
+			if droppedSamplesCounter != nil {
+				droppedSamplesCounter.Inc()
+			}
+			log.Printf("WARN: dropping metric sample due to unexpected label key %q (expected=%v)", k, labelNames)
+			return nil, false
+		}
+		normalized[k] = v
+	}
+
+	return normalized, true
+}
+
+func guardMetricPanic(kind string, expected []string, labels map[string]string) {
+	if recovered := recover(); recovered != nil {
+		log.Printf("WARN: dropped %s metric sample due to label panic: expected=%v got=%v err=%v", kind, expected, labels, recovered)
+	}
 }

internal/observability/prometheus/backend_test.go

@@ -36,7 +36,10 @@ func TestPrometheusBackendShutdown(t *testing.T) {
 
 func TestPrometheusBackendCounter(t *testing.T) {
 	b := newTestBackend(t)
-	c := b.NewCounter("test_counter_total", "A test counter", "result")
+	c, err := b.NewCounter("test_counter_total", "A test counter", "result")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
 	c.Add(context.Background(), 3, map[string]string{"result": "ok"})
 
 	body := scrapeMetrics(t, b)
@@ -45,7 +48,10 @@ func TestPrometheusBackendCounter(t *testing.T) {
 
 func TestPrometheusBackendUpDownCounter(t *testing.T) {
 	b := newTestBackend(t)
-	u := b.NewUpDownCounter("test_gauge_total", "A test up-down counter", "state")
+	u, err := b.NewUpDownCounter("test_gauge_total", "A test up-down counter", "state")
+	if err != nil {
+		t.Fatalf("NewUpDownCounter returned error: %v", err)
+	}
 	u.Add(context.Background(), 5, map[string]string{"state": "active"})
 	u.Add(context.Background(), -2, map[string]string{"state": "active"})
 
@@ -55,7 +61,10 @@ func TestPrometheusBackendUpDownCounter(t *testing.T) {
 
 func TestPrometheusBackendInt64Gauge(t *testing.T) {
 	b := newTestBackend(t)
-	g := b.NewInt64Gauge("test_int_gauge", "An integer gauge", "ifname")
+	g, err := b.NewInt64Gauge("test_int_gauge", "An integer gauge", "ifname")
+	if err != nil {
+		t.Fatalf("NewInt64Gauge returned error: %v", err)
+	}
 	g.Record(context.Background(), 42, map[string]string{"ifname": "wg0"})
 
 	body := scrapeMetrics(t, b)
@@ -64,7 +73,10 @@ func TestPrometheusBackendInt64Gauge(t *testing.T) {
 
 func TestPrometheusBackendFloat64Gauge(t *testing.T) {
 	b := newTestBackend(t)
-	g := b.NewFloat64Gauge("test_float_gauge", "A float gauge", "cert")
+	g, err := b.NewFloat64Gauge("test_float_gauge", "A float gauge", "cert")
+	if err != nil {
+		t.Fatalf("NewFloat64Gauge returned error: %v", err)
+	}
 	g.Record(context.Background(), 7.5, map[string]string{"cert": "example.com"})
 
 	body := scrapeMetrics(t, b)
@@ -74,7 +86,10 @@ func TestPrometheusBackendFloat64Gauge(t *testing.T) {
 func TestPrometheusBackendHistogram(t *testing.T) {
 	b := newTestBackend(t)
 	buckets := []float64{0.1, 0.5, 1.0, 5.0}
-	h := b.NewHistogram("test_duration_seconds", "A test histogram", buckets, "method")
+	h, err := b.NewHistogram("test_duration_seconds", "A test histogram", buckets, "method")
+	if err != nil {
+		t.Fatalf("NewHistogram returned error: %v", err)
+	}
 	h.Record(context.Background(), 0.3, map[string]string{"method": "GET"})
 
 	body := scrapeMetrics(t, b)
@@ -85,7 +100,10 @@ func TestPrometheusBackendHistogram(t *testing.T) {
 
 func TestPrometheusBackendMultipleLabels(t *testing.T) {
 	b := newTestBackend(t)
-	c := b.NewCounter("multi_label_total", "Multi-label counter", "method", "route", "status_code")
+	c, err := b.NewCounter("multi_label_total", "Multi-label counter", "method", "route", "status_code")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
 	c.Add(context.Background(), 1, map[string]string{
 		"method":      "POST",
 		"route":       "/api/peers",
@@ -122,23 +140,29 @@ func TestPrometheusBackendNoGoMetrics(t *testing.T) {
 func TestPrometheusBackendNilLabels(t *testing.T) {
 	// Adding with nil labels should not panic (treated as empty map).
 	b := newTestBackend(t)
-	c := b.NewCounter("nil_labels_total", "counter with no labels")
+	c, err := b.NewCounter("nil_labels_total", "counter with no labels")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
 	// nil labels with no label names declared should be safe
 	c.Add(context.Background(), 1, nil)
 }
 
 func TestPrometheusBackendConcurrentAdd(t *testing.T) {
 	b := newTestBackend(t)
-	c := b.NewCounter("concurrent_total", "concurrent counter", "worker")
+	c, err := b.NewCounter("concurrent_total", "concurrent counter", "worker")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
 
 	done := make(chan struct{})
 	for i := 0; i < 10; i++ {
-		go func(_ int) {
+		go func() {
 			for j := 0; j < 100; j++ {
 				c.Add(context.Background(), 1, map[string]string{"worker": "w"})
 			}
 			done <- struct{}{}
-		}(i)
+		}()
 	}
 	for i := 0; i < 10; i++ {
 		<-done
@@ -148,6 +172,40 @@ func TestPrometheusBackendConcurrentAdd(t *testing.T) {
 	assertMetricPresent(t, body, `concurrent_total{worker="w"} 1000`)
 }
 
+func TestPrometheusBackendAlreadyRegisteredCounter(t *testing.T) {
+	b := newTestBackend(t)
+	c1, err := b.NewCounter("dupe_counter_total", "duplicate counter", "result")
+	if err != nil {
+		t.Fatalf("first NewCounter returned error: %v", err)
+	}
+	c2, err := b.NewCounter("dupe_counter_total", "duplicate counter", "result")
+	if err != nil {
+		t.Fatalf("second NewCounter returned error: %v", err)
+	}
+
+	c1.Add(context.Background(), 1, map[string]string{"result": "ok"})
+	c2.Add(context.Background(), 2, map[string]string{"result": "ok"})
+
+	body := scrapeMetrics(t, b)
+	assertMetricPresent(t, body, `dupe_counter_total{result="ok"} 3`)
+}
+
+func TestPrometheusBackendInvalidLabelsNoPanic(t *testing.T) {
+	b := newTestBackend(t)
+	c, err := b.NewCounter("invalid_labels_total", "invalid labels test", "result")
+	if err != nil {
+		t.Fatalf("NewCounter returned error: %v", err)
+	}
+
+	// Extra label key should be dropped and must not panic.
+	c.Add(context.Background(), 5, map[string]string{"result": "ok", "unexpected": "x"})
+
+	body := scrapeMetrics(t, b)
+	if strings.Contains(body, `invalid_labels_total{result="ok"}`) {
+		t.Error("invalid label sample should have been dropped")
+	}
+}
+
 // --- helpers ---
 
 func scrapeMetrics(t *testing.T, b *obsprom.Backend) string {

main.go

@@ -175,6 +175,7 @@ func main() {
 		otelMetricsEndpoint       string
 		otelMetricsInsecure       bool
 		otelMetricsExportInterval time.Duration
+		otelMetricsTimeout        time.Duration
 	)
 
 	interfaceName = os.Getenv("INTERFACE")
@@ -229,6 +230,14 @@ func main() {
 			log.Printf("WARN: invalid OTEL_METRICS_EXPORT_INTERVAL=%q: %v", v, err2)
 		}
 	}
+	otelMetricsTimeout = 10 * time.Second // default
+	if v := os.Getenv("OTEL_METRICS_TIMEOUT"); v != "" {
+		if d, err2 := time.ParseDuration(v); err2 == nil {
+			otelMetricsTimeout = d
+		} else {
+			log.Printf("WARN: invalid OTEL_METRICS_TIMEOUT=%q: %v", v, err2)
+		}
+	}
 
 	if interfaceName == "" {
 		flag.StringVar(&interfaceName, "interface", "wg0", "Name of the WireGuard interface")
@@ -322,6 +331,7 @@ func main() {
 	flag.StringVar(&otelMetricsEndpoint, "otel-metrics-endpoint", otelMetricsEndpoint, "OTLP collector endpoint (e.g. localhost:4317)")
 	flag.BoolVar(&otelMetricsInsecure, "otel-metrics-insecure", otelMetricsInsecure, "Disable TLS for OTLP connection")
 	flag.DurationVar(&otelMetricsExportInterval, "otel-metrics-export-interval", otelMetricsExportInterval, "Interval between OTLP metric pushes")
+	flag.DurationVar(&otelMetricsTimeout, "otel-metrics-timeout", otelMetricsTimeout, "Timeout for OTLP exporter setup")
 
 	flag.Parse()
 
@@ -347,6 +357,7 @@ func main() {
 			Endpoint:       otelMetricsEndpoint,
 			Insecure:       otelMetricsInsecure,
 			ExportInterval: otelMetricsExportInterval,
+			Timeout:        otelMetricsTimeout,
 		},
 		ServiceName:           "gerbil",
 		ServiceVersion:        "1.0.0",
@@ -543,6 +554,8 @@ func main() {
 
 	// Register metrics endpoint only for Prometheus backend.
 	// OTel backend pushes to a collector; no /metrics endpoint needed.
+	// Note: metricsPath is registered directly without httpMetricsMiddleware to prevent infinite recursion.
+	// The metricsHandler must not be wrapped by the middleware, as it would observe its own observation calls.
 	if metricsHandler != nil {
 		http.Handle(metricsPath, metricsHandler)
 		logger.Info("Metrics endpoint enabled at %s", metricsPath)
@@ -1162,10 +1175,12 @@ func removePeerInternal(publicKey string) error {
 
 	// Get current peer info before removing to clear relay connections and bandwidth limits
 	var wgIPs []string
+	allowedIPsCount := 0
 	device, err := wgClient.Device(interfaceName)
 	if err == nil {
 		for _, peer := range device.Peers {
 			if peer.PublicKey.String() == publicKey {
+				allowedIPsCount = len(peer.AllowedIPs)
 				// Extract WireGuard IPs from this peer's allowed IPs
 				for _, allowedIP := range peer.AllowedIPs {
 					wgIPs = append(wgIPs, allowedIP.IP.String())
@@ -1208,7 +1223,7 @@ func removePeerInternal(publicKey string) error {
 
 	// Record metrics
 	metrics.RecordPeersTotal(interfaceName, -1)
-	metrics.RecordAllowedIPsCount(interfaceName, publicKey, -int64(len(wgIPs)))
+	metrics.RecordAllowedIPsCount(interfaceName, publicKey, -int64(allowedIPsCount))
 
 	return nil
 }

proxy/proxy.go

@@ -19,7 +19,6 @@ import (
 	"github.com/fosrl/gerbil/internal/metrics"
 	"github.com/fosrl/gerbil/logger"
 	"github.com/patrickmn/go-cache"
-	"golang.org/x/sync/errgroup"
 )
 
 // RouteRecord represents a routing configuration
@@ -80,9 +79,7 @@ type SNIProxy struct {
 }
 
 type activeTunnel struct {
-	ctx    context.Context
+	conns []net.Conn
-	cancel context.CancelFunc
-	count  int // protected by activeTunnelsLock
 }
 
 // readOnlyConn is a wrapper for io.Reader that implements net.Conn
@@ -551,7 +548,7 @@ func (p *SNIProxy) handleConnection(clientConn net.Conn) {
 		logger.Debug("SNI extraction failed: %v", err)
 		return
 	}
-	metrics.RecordProxyTLSHandshake(hostname, time.Since(clientHelloStart).Seconds())
+	metrics.RecordProxyTLSHandshake(time.Since(clientHelloStart).Seconds())
 
 	if hostname == "" {
 		log.Println("No SNI hostname found")
@@ -599,8 +596,8 @@ func (p *SNIProxy) handleConnection(clientConn net.Conn) {
 	defer targetConn.Close()
 
 	logger.Debug("Connected to target: %s:%d", route.TargetHost, route.TargetPort)
-	metrics.RecordActiveProxyConnection(hostname, 1)
+	metrics.RecordActiveProxyConnection(1)
-	defer metrics.RecordActiveProxyConnection(hostname, -1)
+	defer metrics.RecordActiveProxyConnection(-1)
 
 	// Send PROXY protocol header if enabled
 	if p.proxyProtocol {
@@ -620,32 +617,37 @@ func (p *SNIProxy) handleConnection(clientConn net.Conn) {
 		}
 	}
 
-	// Track this tunnel by SNI using context for cancellation
+	// Track this tunnel by SNI
 	p.activeTunnelsLock.Lock()
 	tunnel, ok := p.activeTunnels[hostname]
 	if !ok {
-		ctx, cancel := context.WithCancel(p.ctx)
+		tunnel = &activeTunnel{}
-		tunnel = &activeTunnel{ctx: ctx, cancel: cancel}
 		p.activeTunnels[hostname] = tunnel
 	}
-	tunnel.count++
+	tunnel.conns = append(tunnel.conns, actualClientConn)
-	tunnelCtx := tunnel.ctx
 	p.activeTunnelsLock.Unlock()
 
 	defer func() {
+		// Remove this conn from active tunnels
 		p.activeTunnelsLock.Lock()
-		tunnel.count--
+		if tunnel, ok := p.activeTunnels[hostname]; ok {
-		if tunnel.count == 0 {
+			newConns := make([]net.Conn, 0, len(tunnel.conns))
-			tunnel.cancel()
+			for _, c := range tunnel.conns {
-			if p.activeTunnels[hostname] == tunnel {
+				if c != actualClientConn {
+					newConns = append(newConns, c)
+				}
+			}
+			if len(newConns) == 0 {
 				delete(p.activeTunnels, hostname)
+			} else {
+				tunnel.conns = newConns
 			}
 		}
 		p.activeTunnelsLock.Unlock()
 	}()
 
-	// Start bidirectional data transfer with tunnel-level cancellation context.
+	// Start bidirectional data transfer
-	p.pipe(tunnelCtx, hostname, actualClientConn, targetConn, clientReader)
+	p.pipe(hostname, actualClientConn, targetConn, clientReader)
 }
 
 // getRoute retrieves routing information for a hostname
@@ -653,7 +655,7 @@ func (p *SNIProxy) getRoute(hostname, clientAddr string) (*RouteRecord, error) {
 	// Check local overrides first
 	if _, isOverride := p.localOverrides[hostname]; isOverride {
 		logger.Debug("Local override matched for hostname: %s", hostname)
-		metrics.RecordProxyRouteLookup("local_override", hostname)
+		metrics.RecordProxyRouteLookup("local_override")
 		return &RouteRecord{
 			Hostname:   hostname,
 			TargetHost: p.localProxyAddr,
@@ -666,7 +668,7 @@ func (p *SNIProxy) getRoute(hostname, clientAddr string) (*RouteRecord, error) {
 	_, isLocal := p.localSNIs[hostname]
 	p.localSNIsLock.RUnlock()
 	if isLocal {
-		metrics.RecordProxyRouteLookup("local", hostname)
+		metrics.RecordProxyRouteLookup("local")
 		return &RouteRecord{
 			Hostname:   hostname,
 			TargetHost: p.localProxyAddr,
@@ -677,16 +679,16 @@ func (p *SNIProxy) getRoute(hostname, clientAddr string) (*RouteRecord, error) {
 	// Check cache first
 	if cached, found := p.cache.Get(hostname); found {
 		if cached == nil {
-			metrics.RecordProxyRouteLookup("cached_not_found", hostname)
+			metrics.RecordProxyRouteLookup("cached_not_found")
 			return nil, nil // Cached negative result
 		}
 		logger.Debug("Cache hit for hostname: %s", hostname)
-		metrics.RecordProxyRouteLookup("cache_hit", hostname)
+		metrics.RecordProxyRouteLookup("cache_hit")
 		return cached.(*RouteRecord), nil
 	}
 
 	logger.Debug("Cache miss for hostname: %s, querying API", hostname)
-	metrics.RecordProxyRouteLookup("cache_miss", hostname)
+	metrics.RecordProxyRouteLookup("cache_miss")
 
 	// Query API with timeout
 	ctx, cancel := context.WithTimeout(p.ctx, 5*time.Second)
@@ -792,48 +794,61 @@ func (p *SNIProxy) selectStickyEndpoint(clientAddr string, endpoints []string) s
 }
 
 // pipe handles bidirectional data transfer between connections
-func (p *SNIProxy) pipe(ctx context.Context, hostname string, clientConn, targetConn net.Conn, clientReader io.Reader) {
+func (p *SNIProxy) pipe(hostname string, clientConn, targetConn net.Conn, clientReader io.Reader) {
-	g, gCtx := errgroup.WithContext(ctx)
+	var wg sync.WaitGroup
+	wg.Add(2)
 
-	// Close connections when context cancels to unblock io.Copy operations
+	// closeOnce ensures we only close connections once
-	context.AfterFunc(gCtx, func() {
+	var closeOnce sync.Once
-		clientConn.Close()
+	closeConns := func() {
-		targetConn.Close()
+		closeOnce.Do(func() {
-	})
+			// Close both connections to unblock any pending reads
+			clientConn.Close()
+			targetConn.Close()
+		})
+	}
 
-	// Copy data from client to target (using buffered reader and pooled memory).
+	// Copy data from client to target (using the buffered reader)
-	g.Go(func() error {
+	go func() {
+		defer wg.Done()
+		defer closeConns()
+
+		// Get buffer from pool and return when done
 		bufPtr := p.bufferPool.Get().(*[]byte)
 		defer func() {
+			// Clear buffer before returning to pool to prevent data leakage
 			clear(*bufPtr)
 			p.bufferPool.Put(bufPtr)
 		}()
 
 		bytesCopied, err := io.CopyBuffer(targetConn, clientReader, *bufPtr)
-		metrics.RecordProxyBytesTransmitted(hostname, "client_to_target", bytesCopied)
+		metrics.RecordProxyBytesTransmitted("client_to_target", bytesCopied)
 		if err != nil && err != io.EOF {
 			logger.Debug("Copy client->target error: %v", err)
 		}
-		return err
+	}()
-	})
 
 	// Copy data from target to client
-	g.Go(func() error {
+	go func() {
+		defer wg.Done()
+		defer closeConns()
+
+		// Get buffer from pool and return when done
 		bufPtr := p.bufferPool.Get().(*[]byte)
 		defer func() {
+			// Clear buffer before returning to pool to prevent data leakage
 			clear(*bufPtr)
 			p.bufferPool.Put(bufPtr)
 		}()
 
 		bytesCopied, err := io.CopyBuffer(clientConn, targetConn, *bufPtr)
-		metrics.RecordProxyBytesTransmitted(hostname, "target_to_client", bytesCopied)
+		metrics.RecordProxyBytesTransmitted("target_to_client", bytesCopied)
 		if err != nil && err != io.EOF {
 			logger.Debug("Copy target->client error: %v", err)
 		}
-		return err
+	}()
-	})
 
-	_ = g.Wait()
+	wg.Wait()
 }
 
 // GetCacheStats returns cache statistics
@@ -869,14 +884,16 @@ func (p *SNIProxy) UpdateLocalSNIs(fullDomains []string) {
 
 	logger.Debug("Updated local SNIs, added %d, removed %d", len(newSNIs), len(removed))
 
-	// Terminate tunnels for removed SNIs via context cancellation
+	// Terminate tunnels for removed SNIs
 	if len(removed) > 0 {
 		p.activeTunnelsLock.Lock()
 		for _, sni := range removed {
-			if tunnel, ok := p.activeTunnels[sni]; ok {
+			if tunnels, ok := p.activeTunnels[sni]; ok {
-				tunnel.cancel()
+				for _, conn := range tunnels.conns {
+					conn.Close()
+				}
 				delete(p.activeTunnels, sni)
-				logger.Debug("Cancelled tunnel context for SNI target change: %s", sni)
+				logger.Debug("Closed tunnels for SNI target change: %s", sni)
 			}
 		}
 		p.activeTunnelsLock.Unlock()