feat(metrics): enable TLS/mTLS serving support (#1576)

* feat(config): add metrics TLS configuration

Signed-off-by: Xe Iaso <me@xeiaso.net>

* feat(metrics): add naive TLS serving for metrics

Signed-off-by: Xe Iaso <me@xeiaso.net>

* feat(metrics): import keypairreloader from a private project

Signed-off-by: Xe Iaso <me@xeiaso.net>

* fix(metrics): properly surface errors with the metrics server

Signed-off-by: Xe Iaso <me@xeiaso.net>

* feat(config): add CA certificate config value

Signed-off-by: Xe Iaso <me@xeiaso.net>

* feat(metrics): enable mTLS support

Signed-off-by: Xe Iaso <me@xeiaso.net>

* doc(default-config): document how to set up TLS and mTLS

Signed-off-by: Xe Iaso <me@xeiaso.net>

* doc: document metrics TLS and mTLS

Signed-off-by: Xe Iaso <me@xeiaso.net>

* chore: spelling

Signed-off-by: Xe Iaso <me@xeiaso.net>

---------

Signed-off-by: Xe Iaso <me@xeiaso.net>

.github/actions/spelling/allow.txt

@@ -37,3 +37,4 @@ unipromos
 Samsung
 wenet
 qwertiko
+setuplistener

.github/actions/spelling/expect.txt

@@ -47,6 +47,7 @@ cachediptoasn
 Caddyfile
 caninetools
 Cardyb
+CAs
 celchecker
 celphase
 cerr
@@ -203,8 +204,10 @@ kagi
 kagibot
 Keyfunc
 keypair
+keypairreloader
 KHTML
 kinda
+kpr
 KUBECONFIG
 lcj
 ldflags
@@ -229,6 +232,7 @@ metarefresh
 metrix
 mimi
 Minfilia
+minica
 mistralai
 mnt
 Mojeek
@@ -313,6 +317,7 @@ searchbot
 searx
 sebest
 secretplans
+selfsigned
 Semrush
 Seo
 setsebool

cmd/anubis/main.go

@@ -17,12 +17,10 @@ import (
 	"net"
 	"net/http"
 	"net/http/httputil"
-	"net/http/pprof"
 	"net/url"
 	"os"
 	"os/signal"
 	"path/filepath"
-	"strconv"
 	"strings"
 	"sync"
 	"syscall"
@@ -33,12 +31,12 @@ import (
 	"github.com/TecharoHQ/anubis/internal"
 	libanubis "github.com/TecharoHQ/anubis/lib"
 	"github.com/TecharoHQ/anubis/lib/config"
+	"github.com/TecharoHQ/anubis/lib/metrics"
 	botPolicy "github.com/TecharoHQ/anubis/lib/policy"
 	"github.com/TecharoHQ/anubis/lib/thoth"
 	"github.com/TecharoHQ/anubis/web"
 	"github.com/facebookgo/flagenv"
 	_ "github.com/joho/godotenv/autoload"
-	"github.com/prometheus/client_golang/prometheus/promhttp"
 	healthv1 "google.golang.org/grpc/health/grpc_health_v1"
 )
 
@@ -119,33 +117,6 @@ func doHealthCheck() error {
 	return nil
 }
 
-// parseBindNetFromAddr determine bind network and address based on the given network and address.
-func parseBindNetFromAddr(address string) (string, string) {
-	defaultScheme := "http://"
-	if !strings.Contains(address, "://") {
-		if strings.HasPrefix(address, ":") {
-			address = defaultScheme + "localhost" + address
-		} else {
-			address = defaultScheme + address
-		}
-	}
-
-	bindUri, err := url.Parse(address)
-	if err != nil {
-		log.Fatal(fmt.Errorf("failed to parse bind URL: %w", err))
-	}
-
-	switch bindUri.Scheme {
-	case "unix":
-		return "unix", bindUri.Path
-	case "tcp", "http", "https":
-		return "tcp", bindUri.Host
-	default:
-		log.Fatal(fmt.Errorf("unsupported network scheme %s in address %s", bindUri.Scheme, address))
-	}
-	return "", address
-}
-
 func parseSameSite(s string) http.SameSite {
 	switch strings.ToLower(s) {
 	case "none":
@@ -162,53 +133,6 @@ func parseSameSite(s string) http.SameSite {
 	return http.SameSiteDefaultMode
 }
 
-func setupListener(network string, address string) (net.Listener, string) {
-	formattedAddress := ""
-
-	if network == "" {
-		// keep compatibility
-		network, address = parseBindNetFromAddr(address)
-	}
-
-	switch network {
-	case "unix":
-		formattedAddress = "unix:" + address
-	case "tcp":
-		if strings.HasPrefix(address, ":") { // assume it's just a port e.g. :4259
-			formattedAddress = "http://localhost" + address
-		} else {
-			formattedAddress = "http://" + address
-		}
-	default:
-		formattedAddress = fmt.Sprintf(`(%s) %s`, network, address)
-	}
-
-	listener, err := net.Listen(network, address)
-	if err != nil {
-		log.Fatal(fmt.Errorf("failed to bind to %s: %w", formattedAddress, err))
-	}
-
-	// additional permission handling for unix sockets
-	if network == "unix" {
-		mode, err := strconv.ParseUint(*socketMode, 8, 0)
-		if err != nil {
-			listener.Close()
-			log.Fatal(fmt.Errorf("could not parse socket mode %s: %w", *socketMode, err))
-		}
-
-		err = os.Chmod(address, os.FileMode(mode))
-		if err != nil {
-			err := listener.Close()
-			if err != nil {
-				log.Printf("failed to close listener: %v", err)
-			}
-			log.Fatal(fmt.Errorf("could not change socket mode: %w", err))
-		}
-	}
-
-	return listener, formattedAddress
-}
-
 func makeReverseProxy(target string, targetSNI string, targetHost string, insecureSkipVerify bool, targetDisableKeepAlive bool) (http.Handler, error) {
 	targetUri, err := url.Parse(target)
 	if err != nil {
@@ -304,11 +228,6 @@ func main() {
 
 	wg := new(sync.WaitGroup)
 
-	if *metricsBind != "" {
-		wg.Add(1)
-		go metricsServer(ctx, *lg.With("subsystem", "metrics"), wg.Done)
-	}
-
 	var rp http.Handler
 	// when using anubis via Systemd and environment variables, then it is not possible to set targe to an empty string but only to space
 	if strings.TrimSpace(*target) != "" {
@@ -348,6 +267,26 @@ func main() {
 	lg.Debug("swapped to new logger")
 	slog.SetDefault(lg)
 
+	if *metricsBind != "" || policy.Metrics != nil {
+		wg.Add(1)
+
+		ms := &metrics.Server{
+			Config: policy.Metrics,
+			Log:    lg,
+		}
+
+		if policy.Metrics == nil {
+			lg.Debug("migrating flags to metrics config", "bind", *metricsBind, "network", *metricsBindNetwork, "socket-mode", *socketMode)
+			ms.Config = &config.Metrics{
+				Bind:       *metricsBind,
+				Network:    *metricsBindNetwork,
+				SocketMode: *socketMode,
+			}
+		}
+
+		go ms.Run(ctx, wg.Done)
+	}
+
 	// Warn if persistent storage is used without a configured signing key
 	if policy.Store.IsPersistent() {
 		if *hs512Secret == "" && *ed25519PrivateKeyHex == "" && *ed25519PrivateKeyHexFile == "" {
@@ -484,7 +423,11 @@ func main() {
 	h = internal.JA4H(h)
 
 	srv := http.Server{Handler: h, ErrorLog: internal.GetFilteredHTTPLogger()}
-	listener, listenerUrl := setupListener(*bindNetwork, *bind)
+	listener, listenerUrl, err := internal.SetupListener(*bindNetwork, *bind, *socketMode)
+	if err != nil {
+		log.Fatalf("SetupListener(%q, %q, %q): %v", *bindNetwork, *bind, *socketMode, err)
+	}
+
 	lg.Info(
 		"listening",
 		"url", listenerUrl,
@@ -519,53 +462,6 @@ func main() {
 	wg.Wait()
 }
 
-func metricsServer(ctx context.Context, lg slog.Logger, done func()) {
-	defer done()
-
-	mux := http.NewServeMux()
-	mux.HandleFunc("GET /debug/pprof/", pprof.Index)
-	mux.HandleFunc("GET /debug/pprof/cmdline", pprof.Cmdline)
-	mux.HandleFunc("GET /debug/pprof/profile", pprof.Profile)
-	mux.HandleFunc("GET /debug/pprof/symbol", pprof.Symbol)
-	mux.HandleFunc("GET /debug/pprof/trace", pprof.Trace)
-	mux.Handle("/metrics", promhttp.Handler())
-	mux.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
-		st, ok := internal.GetHealth("anubis")
-		if !ok {
-			slog.Error("health service anubis does not exist, file a bug")
-		}
-
-		switch st {
-		case healthv1.HealthCheckResponse_NOT_SERVING:
-			http.Error(w, "NOT OK", http.StatusInternalServerError)
-			return
-		case healthv1.HealthCheckResponse_SERVING:
-			fmt.Fprintln(w, "OK")
-			return
-		default:
-			http.Error(w, "UNKNOWN", http.StatusFailedDependency)
-			return
-		}
-	})
-
-	srv := http.Server{Handler: mux, ErrorLog: internal.GetFilteredHTTPLogger()}
-	listener, metricsUrl := setupListener(*metricsBindNetwork, *metricsBind)
-	lg.Debug("listening for metrics", "url", metricsUrl)
-
-	go func() {
-		<-ctx.Done()
-		c, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-		defer cancel()
-		if err := srv.Shutdown(c); err != nil {
-			log.Printf("cannot shut down: %v", err)
-		}
-	}()
-
-	if err := srv.Serve(listener); !errors.Is(err, http.ErrServerClosed) {
-		log.Fatal(err)
-	}
-}
-
 func extractEmbedFS(fsys embed.FS, root string, destDir string) error {
 	return fs.WalkDir(fsys, root, func(path string, d fs.DirEntry, err error) error {
 		if err != nil {

data/botPolicies.yaml

@@ -166,6 +166,29 @@ status_codes:
   CHALLENGE: 200
   DENY: 200
 
+# # Configuration for the metrics server. See the docs for more information:
+# #
+# # https://anubis.techaro.lol/docs/admin/policies#metrics-server
+# #
+# # This is commented out by default so that command line flags take precedence.
+# metrics:
+#   bind: ":9090"
+#   network: "tcp"
+#
+# # To serve metrics over TLS, set the path to the right TLS certificate and key
+# # here. When the files change on disk, they will automatically be reloaded.
+# #
+# # https://anubis.techaro.lol/docs/admin/policies#tls
+#   tls:
+#     certificate: /path/to/tls.crt
+#     key: /path/to/tls.key
+#
+#     # If you want to secure your metrics endpoint using mutual TLS (mTLS), set
+#     # the path to a certificate authority public certificate here.
+#     #
+#     # https://anubis.techaro.lol/docs/admin/policies#mtls
+#     ca: /path/to/ca.crt
+
 # Anubis can store temporary data in one of a few backends. See the storage
 # backends section of the docs for more information:
 #

data/crawlers/ai-search.yaml

@@ -4,5 +4,5 @@
 #  - Claude-SearchBot: No published IP allowlist
 - name: "ai-crawlers-search"
   user_agent_regex: >-
-    OAI-SearchBot|Claude-SearchBot|PerplexityBot
+    OAI-SearchBot|Claude-SearchBot|PerplexityBot|meta-webindexer
   action: DENY

docs/docs/CHANGELOG.md

@@ -11,15 +11,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+<!-- This changes the project to: -->
+
+- Move metrics server configuration to [the policy file](./admin/policies.mdx#metrics-server).
 - Expose [pprof endpoints](https://pkg.go.dev/net/http/pprof) on the metrics listener to enable profiling Anubis in production.
 - fix: prevent nil pointer panic in challenge validation when threshold rules match during PassChallenge (#1463)
 - Instruct reverse proxies to not cache error pages.
 - Fixed mixed tab/space indentation in Caddy documentation code block
 - Improve error messages and fix broken REDIRECT_DOMAINS link in docs ([#1193](https://github.com/TecharoHQ/anubis/issues/1193))
 - Add Bulgarian locale ([#1394](https://github.com/TecharoHQ/anubis/pull/1394))
-
-<!-- This changes the project to: -->
 - Fix CEL internal errors when iterating `headers`/`query` map wrappers by implementing map iterators for `HTTPHeaders` and `URLValues` ([#1465](https://github.com/TecharoHQ/anubis/pull/1465)).
+- Enable [metrics serving via TLS](./admin/policies.mdx#tls), including [mutual TLS (mTLS)](./admin/policies.mdx#mtls).
 
 ## v1.25.0: Necron
 

docs/docs/admin/installation.mdx

@@ -87,8 +87,8 @@ Anubis uses these environment variables for configuration:
 | `ED25519_PRIVATE_KEY_HEX_FILE` | unset                   | Path to a file containing the hex-encoded ed25519 private key. Only one of this or its sister option may be set. **Required when using persistent storage backends** (like bbolt) to ensure challenges survive service restarts. When running multiple instances on the same base domain, the key must be the same across all instances.                                                                                                                                                                                                       |
 | `ERROR_TITLE`                  | unset                   | <EO /> If set, override the translation stack to show a custom title for error pages such as "Something went wrong!". See [Customizing messages](./botstopper.mdx#customizing-messages) for more details.                                                                                                                                                                                                                                                                                                                                      |
 | `JWT_RESTRICTION_HEADER`       | `X-Real-IP`             | If set, the JWT is only valid if the current value of this header matches the value when the JWT was created. You can use it e.g. to restrict a JWT to the source IP of the user using `X-Real-IP`.                                                                                                                                                                                                                                                                                                                                            |
-| `METRICS_BIND`                 | `:9090`                 | The network address that Anubis serves Prometheus metrics on. See `BIND` for more information.                                                                                                                                                                                                                                                                                                                                                                                                                                                 |
+| `METRICS_BIND`                 | `:9090`                 | The legacy configuration value for the network address that Anubis serves Prometheus metrics on. Please migrate this to [the policy file](./policies.mdx#metrics-server) as soon as possible.                                                                                                                                                                                                                                                                                                                                                  |
-| `METRICS_BIND_NETWORK`         | `tcp`                   | The address family that the Anubis metrics server listens on. See `BIND_NETWORK` for more information.                                                                                                                                                                                                                                                                                                                                                                                                                                         |
+| `METRICS_BIND_NETWORK`         | `tcp`                   | The legacy configuration value for the address family that Anubis serves Prometheus metrics on. Please migrate this to [the policy file](./policies.mdx#metrics-server) as soon as possible.                                                                                                                                                                                                                                                                                                                                                   |
 | `OG_EXPIRY_TIME`               | `24h`                   | The expiration time for the Open Graph tag cache. Prefer using [the policy file](./configuration/open-graph.mdx) to configure the Open Graph subsystem.                                                                                                                                                                                                                                                                                                                                                                                        |
 | `OG_PASSTHROUGH`               | `false`                 | If set to `true`, Anubis will enable Open Graph tag passthrough. Prefer using [the policy file](./configuration/open-graph.mdx) to configure the Open Graph subsystem.                                                                                                                                                                                                                                                                                                                                                                         |
 | `OG_CACHE_CONSIDER_HOST`       | `false`                 | If set to `true`, Anubis will consider the host in the Open Graph tag cache key. Prefer using [the policy file](./configuration/open-graph.mdx) to configure the Open Graph subsystem.                                                                                                                                                                                                                                                                                                                                                         |

docs/docs/admin/policies.mdx

@@ -117,6 +117,60 @@ remote_addresses:
   - 100.64.0.0/10
 ```
 
+## Metrics server
+
+Anubis includes support for [Prometheus-style metrics](https://prometheus.io/docs/introduction/overview/), allowing systems administrators to monitor Anubis' performance and effectiveness. This is a separate HTTP server with metrics, health checking, and debug routes.
+
+Anubis' metrics server is configured with the `metrics` block in the configuration file:
+
+```yaml
+metrics:
+  bind: ":9090"
+  network: "tcp"
+```
+
+If you want to bind metrics to a Unix socket, make sure to set the network to `unix` and add a socket mode:
+
+```yaml
+metrics:
+  bind: "/tmp/anubis_metrics.sock"
+  network: unix
+  socketMode: "0700" # must be a string
+```
+
+### TLS
+
+If you want to serve the metrics server over TLS, use the `tls` block:
+
+```yaml
+metrics:
+  bind: ":9090"
+  network: "tcp"
+
+  tls:
+    certificate: /path/to/tls.crt
+    key: /path/to/tls.key
+```
+
+The certificate and key will automatically be reloaded when the respective files change.
+
+### mTLS
+
+If you want to validate requests to ensure that they use a client certificate signed by a certificate authority (mutual TLS or mTLS), set the `ca` value in the `tls` block:
+
+```yaml
+metrics:
+  bind: ":9090"
+  network: "tcp"
+
+  tls:
+    certificate: /path/to/tls.crt
+    key: /path/to/tls.key
+    ca: /path/to/ca.crt
+```
+
+As it is not expected for certificate authority certificates to change often, the CA certificate will NOT be automatically reloaded when the respective file changes.
+
 ## Imprint / Impressum support
 
 Anubis has support for showing imprint / impressum information. This is defined in the `impressum` block of your configuration. See [Imprint / Impressum configuration](./configuration/impressum.mdx) for more information.

internal/setuplistener.go

@@ -0,0 +1,92 @@
+package internal
+
+import (
+	"errors"
+	"fmt"
+	"net"
+	"net/url"
+	"os"
+	"strconv"
+	"strings"
+)
+
+// parseBindNetFromAddr determine bind network and address based on the given network and address.
+func parseBindNetFromAddr(address string) (string, string, error) {
+	defaultScheme := "http://"
+	if !strings.Contains(address, "://") {
+		if strings.HasPrefix(address, ":") {
+			address = defaultScheme + "localhost" + address
+		} else {
+			address = defaultScheme + address
+		}
+	}
+
+	bindUri, err := url.Parse(address)
+	if err != nil {
+		return "", "", fmt.Errorf("failed to parse bind URL: %w", err)
+	}
+
+	switch bindUri.Scheme {
+	case "unix":
+		return "unix", bindUri.Path, nil
+	case "tcp", "http", "https":
+		return "tcp", bindUri.Host, nil
+	default:
+		return "", "", fmt.Errorf("unsupported network scheme %s in address %s", bindUri.Scheme, address)
+	}
+}
+
+// SetupListener sets up a network listener based on the input from configuration
+// envvars. It returns a network listener and the URL to that listener or an error.
+func SetupListener(network, address, socketMode string) (net.Listener, string, error) {
+	formattedAddress := ""
+	var err error
+
+	if network == "" {
+		// keep compatibility
+		network, address, err = parseBindNetFromAddr(address)
+	}
+
+	if err != nil {
+		return nil, "", fmt.Errorf("can't parse bind and network: %w", err)
+	}
+
+	switch network {
+	case "unix":
+		formattedAddress = "unix:" + address
+	case "tcp":
+		if strings.HasPrefix(address, ":") { // assume it's just a port e.g. :4259
+			formattedAddress = "http://localhost" + address
+		} else {
+			formattedAddress = "http://" + address
+		}
+	default:
+		formattedAddress = fmt.Sprintf(`(%s) %s`, network, address)
+	}
+
+	ln, err := net.Listen(network, address)
+	if err != nil {
+		return nil, "", fmt.Errorf("failed to bind to %s: %w", formattedAddress, err)
+	}
+
+	// additional permission handling for unix sockets
+	if network == "unix" {
+		mode, err := strconv.ParseUint(socketMode, 8, 0)
+		if err != nil {
+			ln.Close()
+			return nil, "", fmt.Errorf("could not parse socket mode %s: %w", socketMode, err)
+		}
+
+		err = os.Chmod(address, os.FileMode(mode))
+		if err != nil {
+			err := fmt.Errorf("could not change socket mode: %w", err)
+			clErr := ln.Close()
+			if clErr != nil {
+				return nil, "", errors.Join(err, clErr)
+			}
+			return nil, "", err
+		}
+	}
+
+	return ln, formattedAddress, nil
+}

internal/setuplistener_test.go

@@ -0,0 +1,180 @@
+package internal
+
+import (
+	"io/fs"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"testing"
+)
+
+func TestParseBindNetFromAddr(t *testing.T) {
+	for _, tt := range []struct {
+		name    string
+		address string
+		wantErr bool
+		network string
+		bind    string
+	}{
+		{
+			name:    "simple tcp",
+			address: "localhost:9090",
+			wantErr: false,
+			network: "tcp",
+			bind:    "localhost:9090",
+		},
+		{
+			name:    "simple unix",
+			address: "unix:///tmp/foo.sock",
+			wantErr: false,
+			network: "unix",
+			bind:    "/tmp/foo.sock",
+		},
+		{
+			name:    "invalid network",
+			address: "foo:///tmp/bar.sock",
+			wantErr: true,
+		},
+		{
+			name:    "tcp uri",
+			address: "tcp://[::]:9090",
+			wantErr: false,
+			network: "tcp",
+			bind:    "[::]:9090",
+		},
+		{
+			name:    "http uri",
+			address: "http://[::]:9090",
+			wantErr: false,
+			network: "tcp",
+			bind:    "[::]:9090",
+		},
+		{
+			name:    "https uri",
+			address: "https://[::]:9090",
+			wantErr: false,
+			network: "tcp",
+			bind:    "[::]:9090",
+		},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			network, bind, err := parseBindNetFromAddr(tt.address)
+
+			switch {
+			case tt.wantErr && err == nil:
+				t.Errorf("parseBindNetFromAddr(%q) should have errored but did not", tt.address)
+			case !tt.wantErr && err != nil:
+				t.Errorf("parseBindNetFromAddr(%q) threw an error: %v", tt.address, err)
+			}
+
+			if network != tt.network {
+				t.Errorf("parseBindNetFromAddr(%q) wanted network: %q, got: %q", tt.address, tt.network, network)
+			}
+
+			if bind != tt.bind {
+				t.Errorf("parseBindNetFromAddr(%q) wanted bind: %q, got: %q", tt.address, tt.bind, bind)
+			}
+		})
+	}
+}
+
+func TestSetupListener(t *testing.T) {
+	td := t.TempDir()
+
+	for _, tt := range []struct {
+		name                         string
+		network, address, socketMode string
+		wantErr                      bool
+		socketURLPrefix              string
+	}{
+		{
+			name:            "simple tcp",
+			network:         "",
+			address:         ":0",
+			wantErr:         false,
+			socketURLPrefix: "http://localhost:",
+		},
+		{
+			name:            "simple unix",
+			network:         "",
+			address:         "unix://" + filepath.Join(td, "a"),
+			socketMode:      "0770",
+			wantErr:         false,
+			socketURLPrefix: "unix:" + filepath.Join(td, "a"),
+		},
+		{
+			name:            "tcp",
+			network:         "tcp",
+			address:         ":0",
+			wantErr:         false,
+			socketURLPrefix: "http://localhost:",
+		},
+		{
+			name:            "udp",
+			network:         "udp",
+			address:         ":0",
+			wantErr:         true,
+			socketURLPrefix: "http://localhost:",
+		},
+		{
+			name:            "unix socket",
+			network:         "unix",
+			socketMode:      "0770",
+			address:         filepath.Join(td, "a"),
+			wantErr:         false,
+			socketURLPrefix: "unix:" + filepath.Join(td, "a"),
+		},
+		{
+			name:            "invalid socket mode",
+			network:         "unix",
+			socketMode:      "taco bell",
+			address:         filepath.Join(td, "a"),
+			wantErr:         true,
+			socketURLPrefix: "unix:" + filepath.Join(td, "a"),
+		},
+		{
+			name:            "empty socket mode",
+			network:         "unix",
+			socketMode:      "",
+			address:         filepath.Join(td, "a"),
+			wantErr:         true,
+			socketURLPrefix: "unix:" + filepath.Join(td, "a"),
+		},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			ln, socketURL, err := SetupListener(tt.network, tt.address, tt.socketMode)
+			switch {
+			case tt.wantErr && err == nil:
+				t.Errorf("SetupListener(%q, %q, %q) should have errored but did not", tt.network, tt.address, tt.socketMode)
+			case !tt.wantErr && err != nil:
+				t.Fatalf("SetupListener(%q, %q, %q) threw an error: %v", tt.network, tt.address, tt.socketMode, err)
+			}
+
+			if ln != nil {
+				defer ln.Close()
+			}
+
+			if !tt.wantErr && !strings.HasPrefix(socketURL, tt.socketURLPrefix) {
+				t.Errorf("SetupListener(%q, %q, %q) should have returned a URL with prefix %q but got: %q", tt.network, tt.address, tt.socketMode, tt.socketURLPrefix, socketURL)
+			}
+
+			if tt.socketMode != "" {
+				mode, err := strconv.ParseUint(tt.socketMode, 8, 0)
+				if err != nil {
+					return
+				}
+
+				sockPath := strings.TrimPrefix(socketURL, "unix:")
+				st, err := os.Stat(sockPath)
+				if err != nil {
+					t.Fatalf("can't os.Stat(%q): %v", sockPath, err)
+				}
+
+				if st.Mode().Perm() != fs.FileMode(mode) {
+					t.Errorf("file mode of %q should be %s but is actually %s", sockPath, strconv.FormatUint(mode, 8), strconv.FormatUint(uint64(st.Mode()), 8))
+				}
+			}
+		})
+	}
+}

lib/config/config.go

@@ -334,6 +334,7 @@ type fileConfig struct {
 	DNSBL       bool                `json:"dnsbl"`
 	DNSTTL      DnsTTL              `json:"dns_ttl"`
 	Logging     *Logging            `json:"logging"`
+	Metrics     *Metrics            `json:"metrics,omitempty"`
 }
 
 func (c *fileConfig) Valid() error {
@@ -375,6 +376,12 @@ func (c *fileConfig) Valid() error {
 		}
 	}
 
+	if c.Metrics != nil {
+		if err := c.Metrics.Valid(); err != nil {
+			errs = append(errs, err)
+		}
+	}
+
 	if len(errs) != 0 {
 		return fmt.Errorf("config is not valid:\n%w", errors.Join(errs...))
 	}
@@ -417,6 +424,7 @@ func Load(fin io.Reader, fname string) (*Config, error) {
 		StatusCodes: c.StatusCodes,
 		Store:       c.Store,
 		Logging:     c.Logging,
+		Metrics:     c.Metrics,
 	}
 
 	if c.OpenGraph.TimeToLive != "" {
@@ -508,6 +516,7 @@ type Config struct {
 	Logging     *Logging
 	DNSBL       bool
 	DNSTTL      DnsTTL
+	Metrics     *Metrics
 }
 
 func (c Config) Valid() error {

lib/config/metrics.go

@@ -0,0 +1,133 @@
+package config
+
+import (
+	"crypto/tls"
+	"crypto/x509"
+	"errors"
+	"fmt"
+	"os"
+	"strconv"
+)
+
+var (
+	ErrInvalidMetricsConfig        = errors.New("config: invalid metrics configuration")
+	ErrInvalidMetricsTLSConfig     = errors.New("config: invalid metrics TLS configuration")
+	ErrNoMetricsBind               = errors.New("config.Metrics: must define bind")
+	ErrNoMetricsNetwork            = errors.New("config.Metrics: must define network")
+	ErrNoMetricsSocketMode         = errors.New("config.Metrics: must define socket mode when using unix sockets")
+	ErrInvalidMetricsSocketMode    = errors.New("config.Metrics: invalid unix socket mode")
+	ErrInvalidMetricsNetwork       = errors.New("config.Metrics: invalid metrics network")
+	ErrNoMetricsTLSCertificate     = errors.New("config.Metrics.TLS: must define certificate file")
+	ErrNoMetricsTLSKey             = errors.New("config.Metrics.TLS: must define key file")
+	ErrInvalidMetricsTLSKeypair    = errors.New("config.Metrics.TLS: keypair is invalid")
+	ErrInvalidMetricsCACertificate = errors.New("config.Metrics.TLS: invalid CA certificate")
+	ErrCantReadFile                = errors.New("config: can't read required file")
+)
+
+type Metrics struct {
+	Bind       string      `json:"bind" yaml:"bind"`
+	Network    string      `json:"network" yaml:"network"`
+	SocketMode string      `json:"socketMode" yaml:"socketMode"`
+	TLS        *MetricsTLS `json:"tls" yaml:"tls"`
+}
+
+func (m *Metrics) Valid() error {
+	var errs []error
+
+	if m.Bind == "" {
+		errs = append(errs, ErrNoMetricsBind)
+	}
+
+	if m.Network == "" {
+		errs = append(errs, ErrNoMetricsNetwork)
+	}
+
+	switch m.Network {
+	case "tcp", "tcp4", "tcp6": // https://pkg.go.dev/net#Listen
+	case "unix":
+		if m.SocketMode == "" {
+			errs = append(errs, ErrNoMetricsSocketMode)
+		}
+
+		if _, err := strconv.ParseUint(m.SocketMode, 8, 0); err != nil {
+			errs = append(errs, fmt.Errorf("%w: %w", ErrInvalidMetricsSocketMode, err))
+		}
+	default:
+		errs = append(errs, ErrInvalidMetricsNetwork)
+	}
+
+	if m.TLS != nil {
+		if err := m.TLS.Valid(); err != nil {
+			errs = append(errs, err)
+		}
+	}
+
+	if len(errs) != 0 {
+		return errors.Join(ErrInvalidMetricsConfig, errors.Join(errs...))
+	}
+
+	return nil
+}
+
+type MetricsTLS struct {
+	Certificate string `json:"certificate" yaml:"certificate"`
+	Key         string `json:"key" yaml:"key"`
+	CA          string `json:"ca" yaml:"ca"`
+}
+
+func (mt *MetricsTLS) Valid() error {
+	var errs []error
+
+	if mt.Certificate == "" {
+		errs = append(errs, ErrNoMetricsTLSCertificate)
+	}
+
+	if err := canReadFile(mt.Certificate); err != nil {
+		errs = append(errs, fmt.Errorf("%w %s: %w", ErrCantReadFile, mt.Certificate, err))
+	}
+
+	if mt.Key == "" {
+		errs = append(errs, ErrNoMetricsTLSKey)
+	}
+
+	if err := canReadFile(mt.Key); err != nil {
+		errs = append(errs, fmt.Errorf("%w %s: %w", ErrCantReadFile, mt.Key, err))
+	}
+
+	if _, err := tls.LoadX509KeyPair(mt.Certificate, mt.Key); err != nil {
+		errs = append(errs, fmt.Errorf("%w: %w", ErrInvalidMetricsTLSKeypair, err))
+	}
+
+	if mt.CA != "" {
+		caCert, err := os.ReadFile(mt.CA)
+		if err != nil {
+			errs = append(errs, fmt.Errorf("%w %s: %w", ErrCantReadFile, mt.CA, err))
+		}
+
+		certPool := x509.NewCertPool()
+		if !certPool.AppendCertsFromPEM(caCert) {
+			errs = append(errs, fmt.Errorf("%w %s", ErrInvalidMetricsCACertificate, mt.CA))
+		}
+	}
+
+	if len(errs) != 0 {
+		return errors.Join(ErrInvalidMetricsTLSConfig, errors.Join(errs...))
+	}
+
+	return nil
+}
+
+func canReadFile(fname string) error {
+	fin, err := os.Open(fname)
+	if err != nil {
+		return err
+	}
+	defer fin.Close()
+
+	data := make([]byte, 64)
+	if _, err := fin.Read(data); err != nil {
+		return fmt.Errorf("can't read %s: %w", fname, err)
+	}
+
+	return nil
+}

lib/config/metrics_test.go

@@ -0,0 +1,169 @@
+package config
+
+import (
+	"errors"
+	"testing"
+)
+
+func TestMetricsValid(t *testing.T) {
+	for _, tt := range []struct {
+		name  string
+		input *Metrics
+		err   error
+	}{
+		{
+			name: "basic TCP",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+			},
+		},
+		{
+			name: "basic TCP4",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp4",
+			},
+		},
+		{
+			name: "basic TCP6",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp6",
+			},
+		},
+		{
+			name: "basic unix",
+			input: &Metrics{
+				Bind:       "/tmp/anubis-metrics.sock",
+				Network:    "unix",
+				SocketMode: "0770",
+			},
+		},
+		{
+			name:  "no bind",
+			input: &Metrics{},
+			err:   ErrNoMetricsBind,
+		},
+		{
+			name:  "no network",
+			input: &Metrics{},
+			err:   ErrNoMetricsNetwork,
+		},
+		{
+			name: "no unix socket mode",
+			input: &Metrics{
+				Bind:    "/tmp/anubis-metrics.sock",
+				Network: "unix",
+			},
+			err: ErrNoMetricsSocketMode,
+		},
+		{
+			name: "invalid unix socket mode",
+			input: &Metrics{
+				Bind:       "/tmp/anubis-metrics.sock",
+				Network:    "unix",
+				SocketMode: "taco bell",
+			},
+			err: ErrInvalidMetricsSocketMode,
+		},
+		{
+			name: "invalid network",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "taco",
+			},
+			err: ErrInvalidMetricsNetwork,
+		},
+		{
+			name: "invalid TLS config",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS:     &MetricsTLS{},
+			},
+			err: ErrInvalidMetricsTLSConfig,
+		},
+		{
+			name: "selfsigned TLS cert",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls/selfsigned.crt",
+					Key:         "./testdata/tls/selfsigned.key",
+				},
+			},
+		},
+		{
+			name: "wrong path to selfsigned TLS cert",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls2/selfsigned.crt",
+					Key:         "./testdata/tls2/selfsigned.key",
+				},
+			},
+			err: ErrCantReadFile,
+		},
+		{
+			name: "unparseable TLS cert",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls/invalid.crt",
+					Key:         "./testdata/tls/invalid.key",
+				},
+			},
+			err: ErrInvalidMetricsTLSKeypair,
+		},
+		{
+			name: "mTLS with CA",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls/selfsigned.crt",
+					Key:         "./testdata/tls/selfsigned.key",
+					CA:          "./testdata/tls/minica.pem",
+				},
+			},
+		},
+		{
+			name: "mTLS with nonexistent CA",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls/selfsigned.crt",
+					Key:         "./testdata/tls/selfsigned.key",
+					CA:          "./testdata/tls/nonexistent.crt",
+				},
+			},
+			err: ErrCantReadFile,
+		},
+		{
+			name: "mTLS with invalid CA",
+			input: &Metrics{
+				Bind:    ":9090",
+				Network: "tcp",
+				TLS: &MetricsTLS{
+					Certificate: "./testdata/tls/selfsigned.crt",
+					Key:         "./testdata/tls/selfsigned.key",
+					CA:          "./testdata/tls/invalid.crt",
+				},
+			},
+			err: ErrInvalidMetricsCACertificate,
+		},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			if err := tt.input.Valid(); !errors.Is(err, tt.err) {
+				t.Logf("wanted error: %v", tt.err)
+				t.Logf("got error:    %v", err)
+				t.Error("validation failed")
+			}
+		})
+	}
+}

lib/config/testdata/bad/metrics-invalid-net.yaml

@@ -0,0 +1,3 @@
+metrics:
+  bind: ":9090"
+  network: taco

lib/config/testdata/good/allow_everyone.json

@@ -5,5 +5,9 @@
       "remote_addresses": ["0.0.0.0/0", "::/0"],
       "action": "ALLOW"
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/allow_everyone.yaml

@@ -4,3 +4,7 @@ bots:
       - "0.0.0.0/0"
       - "::/0"
     action: ALLOW
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/block_cf_workers.json

@@ -8,5 +8,9 @@
       "action": "DENY"
     }
   ],
-  "dnsbl": false
+  "dnsbl": false,
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/block_cf_workers.yaml

@@ -3,3 +3,7 @@ bots:
     headers_regex:
       CF-Worker: .*
     action: DENY
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/challenge_cloudflare.yaml

@@ -4,3 +4,7 @@ bots:
     asns:
       match:
         - 13335 # Cloudflare
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/challengemozilla.json

@@ -5,5 +5,9 @@
       "user_agent_regex": "Mozilla",
       "action": "CHALLENGE"
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/challengemozilla.yaml

@@ -2,3 +2,7 @@ bots:
   - name: generic-browser
     user_agent_regex: Mozilla
     action: CHALLENGE
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/dns-ttl-custom.yaml

@@ -6,3 +6,7 @@ bots:
   - name: "test"
     user_agent_regex: ".*"
     action: "DENY"
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/entropy.yaml

@@ -6,3 +6,7 @@ bots:
         - '"Accept" in headers'
         - headers["Accept"].contains("text/html")
         - randInt(1) == 0
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/everything_blocked.json

@@ -6,5 +6,9 @@
       "action": "DENY"
     }
   ],
-  "dnsbl": false
+  "dnsbl": false,
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/everything_blocked.yaml

@@ -2,3 +2,7 @@ bots:
   - name: everything
     user_agent_regex: .*
     action: DENY
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/geoip_us.yaml

@@ -4,3 +4,7 @@ bots:
     geoip:
       countries:
         - US
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/git_client.json

@@ -10,5 +10,9 @@
         ]
       }
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/git_client.yaml

@@ -6,3 +6,7 @@ bots:
         - userAgent.startsWith("git/") || userAgent.contains("libgit")
         - >
           "Git-Protocol" in headers && headers["Git-Protocol"] == "version=2"
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/import_filesystem.json

@@ -3,5 +3,9 @@
     {
       "import": "./testdata/hack-test.json"
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/import_filesystem.yaml

@@ -1,2 +1,6 @@
 bots:
   - import: ./testdata/hack-test.yaml
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/import_keep_internet_working.json

@@ -3,5 +3,9 @@
     {
       "import": "(data)/common/keep-internet-working.yaml"
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/import_keep_internet_working.yaml

@@ -1,2 +1,6 @@
 bots:
   - import: (data)/common/keep-internet-working.yaml
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/impressum.yaml

@@ -8,3 +8,7 @@ impressum:
   page:
     title: Test
     body: <p>This is a test</p>
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/logging-file.yaml

@@ -13,3 +13,7 @@ logs:
     oldFileTimeFormat: 2006-01-02T15-04-05 # RFC 3339-ish
     compress: true
     useLocalTime: false # timezone for rotated files is UTC
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/logging-stdio.yaml

@@ -5,3 +5,7 @@ bots:
 
 logging:
   sink: "stdio"
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/no-thresholds.yaml

@@ -6,3 +6,7 @@ bots:
       adjust: 5
 
 thresholds: []
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/old_xesite.json

@@ -75,5 +75,9 @@
       "user_agent_regex": "Mozilla",
       "action": "CHALLENGE"
     }
-  ]
+  ],
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
+  }
 }

lib/config/testdata/good/opengraph_all_good.yaml

@@ -10,3 +10,7 @@ openGraph:
   default:
     "og:title": "Xe's magic land of fun"
     "og:description": "We're no strangers to love, you know the rules and so do I"
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/simple-weight.yaml

@@ -4,3 +4,7 @@ bots:
     user_agent_regex: Mozilla
     weight:
       adjust: 5
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/status-codes-paranoid.json

@@ -9,5 +9,9 @@
   "status_codes": {
     "CHALLENGE": 200,
     "DENY": 200
+  },
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
   }
 }

lib/config/testdata/good/status-codes-paranoid.yaml

@@ -6,3 +6,7 @@ bots:
 status_codes:
   CHALLENGE: 200
   DENY: 200
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/status-codes-rfc.json

@@ -9,5 +9,9 @@
   "status_codes": {
     "CHALLENGE": 403,
     "DENY": 403
+  },
+  "metrics": {
+    "bind": ":9090",
+    "network": "tcp"
   }
 }

lib/config/testdata/good/status-codes-rfc.yaml

@@ -6,3 +6,7 @@ bots:
 status_codes:
   CHALLENGE: 403
   DENY: 403
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/thresholds.yaml

@@ -33,3 +33,7 @@ thresholds:
     challenge:
       algorithm: fast
       difficulty: 4
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/good/weight-no-weight.yaml

@@ -2,3 +2,7 @@ bots:
   - name: weight
     action: WEIGH
     user_agent_regex: Mozilla
+
+metrics:
+  bind: ":9090"
+  network: "tcp"

lib/config/testdata/tls/1.1.1.1/cert.pem

@@ -0,0 +1,12 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----

lib/config/testdata/tls/1.1.1.1/key.pem

@@ -0,0 +1,6 @@
+-----BEGIN PRIVATE KEY-----
+MIG2AgEAMBAGByqGSM49AgEGBSuBBAAiBIGeMIGbAgEBBDBN8QsHxxHGJpStu8K7
+D/FmaBBNo6c514KGFSIfqGFuREF5aOL3gN/W11yk2OIibdWhZANiAAS7LgNiyhmx
+AbkgOC05tSminO/0ApFDrIqV0ePg+f97FgWYgB+4llIB8KceWGX9qpGZ+qEXWvpL
+NCS00OlYbXkHnZm+5fOmzHgzEfKeC4Nm9/Jf8RK+0tuVH1AyrnV6zgI=
+-----END PRIVATE KEY-----

lib/config/testdata/tls/minica-key.pem

@@ -0,0 +1,6 @@
+-----BEGIN PRIVATE KEY-----
+MIG2AgEAMBAGByqGSM49AgEGBSuBBAAiBIGeMIGbAgEBBDDr9QQo7ZaTgUL6d73G
+2BG7+YRTFJHAZa0FogRglfc+jYttL1J4/xTig3RmHoqSgrehZANiAASDhijM9Xe0
+G9Vam6AJMeKC6aWDNSLwrxNVmPxemsY/yJ1urBgnxRd9GFH6YW1ki/B8rS+Xl1UX
+NnhBrukLaXvgAQQq782/5IUYGsvK5jw8+dSscYVMCQJwGfmQuaNeczQ=
+-----END PRIVATE KEY-----

lib/config/testdata/tls/minica.pem

@@ -0,0 +1,13 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----

lib/config/testdata/tls/selfsigned.crt

@@ -0,0 +1,11 @@
+-----BEGIN CERTIFICATE-----
+MIIBnzCCAVGgAwIBAgIUK39B3Ft+kU5o81IuISs79O4u1ncwBQYDK2VwMEUxCzAJ
+BgNVBAYTAkFVMRMwEQYDVQQIDApTb21lLVN0YXRlMSEwHwYDVQQKDBhJbnRlcm5l
+dCBXaWRnaXRzIFB0eSBMdGQwHhcNMjYwNDIyMTQyNjE4WhcNMjYwNTIyMTQyNjE4
+WjBFMQswCQYDVQQGEwJBVTETMBEGA1UECAwKU29tZS1TdGF0ZTEhMB8GA1UECgwY
+SW50ZXJuZXQgV2lkZ2l0cyBQdHkgTHRkMCowBQYDK2VwAyEAfgpAUpp8MIOOdQpH
+fxaw3R7mFKQRMR6Kmxzk1Xn/2VujUzBRMB0GA1UdDgQWBBSmkBmzo0RiZ2iocMR8
+uIIpz9cZyTAfBgNVHSMEGDAWgBSmkBmzo0RiZ2iocMR8uIIpz9cZyTAPBgNVHRMB
+Af8EBTADAQH/MAUGAytlcANBAG37XXZrVUUzGyy3T9qsPIzvJQAGpGhdjJ7bt9O6
+sBhzrliTONPrudYuyUggWsHgFb0JlN2xs4/2HhKU+PY7AAQ=
+-----END CERTIFICATE-----

lib/config/testdata/tls/selfsigned.key

@@ -0,0 +1,3 @@
+-----BEGIN PRIVATE KEY-----
+MC4CAQAwBQYDK2VwBCIEIL0HxjjfVlg6zQPB9/zTLq0IBzfp8gEoifEYzQZYIj+T
+-----END PRIVATE KEY-----

lib/metrics/keypairreloader.go

@@ -0,0 +1,81 @@
+package metrics
+
+import (
+	"crypto/tls"
+	"fmt"
+	"log/slog"
+	"os"
+	"sync"
+	"time"
+)
+
+type KeypairReloader struct {
+	certMu   sync.RWMutex
+	cert     *tls.Certificate
+	certPath string
+	keyPath  string
+	modTime  time.Time
+	lg       *slog.Logger
+}
+
+func NewKeypairReloader(certPath, keyPath string, lg *slog.Logger) (*KeypairReloader, error) {
+	result := &KeypairReloader{
+		certPath: certPath,
+		keyPath:  keyPath,
+		lg:       lg,
+	}
+	cert, err := tls.LoadX509KeyPair(certPath, keyPath)
+	if err != nil {
+		return nil, err
+	}
+	result.cert = &cert
+
+	st, err := os.Stat(certPath)
+	if err != nil {
+		return nil, err
+	}
+	result.modTime = st.ModTime()
+
+	return result, nil
+}
+
+func (kpr *KeypairReloader) maybeReload() error {
+	kpr.lg.Debug("loading new keypair", "cert", kpr.certPath, "key", kpr.keyPath)
+	newCert, err := tls.LoadX509KeyPair(kpr.certPath, kpr.keyPath)
+	if err != nil {
+		return err
+	}
+
+	st, err := os.Stat(kpr.certPath)
+	if err != nil {
+		return err
+	}
+
+	kpr.certMu.Lock()
+	defer kpr.certMu.Unlock()
+	kpr.cert = &newCert
+	kpr.modTime = st.ModTime()
+
+	return nil
+}
+
+func (kpr *KeypairReloader) GetCertificate(*tls.ClientHelloInfo) (*tls.Certificate, error) {
+	st, err := os.Stat(kpr.certPath)
+	if err != nil {
+		return nil, fmt.Errorf("stat(%q): %w", kpr.certPath, err)
+	}
+
+	kpr.certMu.RLock()
+	needsReload := st.ModTime().After(kpr.modTime)
+	kpr.certMu.RUnlock()
+
+	if needsReload {
+		if err := kpr.maybeReload(); err != nil {
+			return nil, fmt.Errorf("reload cert: %w", err)
+		}
+	}
+
+	kpr.certMu.RLock()
+	defer kpr.certMu.RUnlock()
+	return kpr.cert, nil
+}

lib/metrics/keypairreloader_test.go

@@ -0,0 +1,265 @@
+package metrics
+
+import (
+	"bytes"
+	"crypto/rand"
+	"crypto/rsa"
+	"crypto/x509"
+	"crypto/x509/pkix"
+	"encoding/pem"
+	"errors"
+	"log/slog"
+	"math/big"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+)
+
+func discardLogger() *slog.Logger {
+	return slog.New(slog.DiscardHandler)
+}
+
+// writeKeypair generates a fresh self-signed cert + RSA key and writes them
+// as PEM files in dir. Returns the paths and the cert's DER bytes so callers
+// can identify which pair was loaded.
+func writeKeypair(t *testing.T, dir, prefix string) (certPath, keyPath string, certDER []byte) {
+	t.Helper()
+
+	priv, err := rsa.GenerateKey(rand.Reader, 2048)
+	if err != nil {
+		t.Fatalf("rsa.GenerateKey: %v", err)
+	}
+
+	tmpl := &x509.Certificate{
+		SerialNumber:          big.NewInt(time.Now().UnixNano()),
+		Subject:               pkix.Name{CommonName: "keypairreloader-test"},
+		NotBefore:             time.Now().Add(-time.Hour),
+		NotAfter:              time.Now().Add(time.Hour),
+		KeyUsage:              x509.KeyUsageDigitalSignature | x509.KeyUsageKeyEncipherment,
+		ExtKeyUsage:           []x509.ExtKeyUsage{x509.ExtKeyUsageServerAuth},
+		BasicConstraintsValid: true,
+		DNSNames:              []string{"keypairreloader-test"},
+	}
+
+	der, err := x509.CreateCertificate(rand.Reader, tmpl, tmpl, &priv.PublicKey, priv)
+	if err != nil {
+		t.Fatalf("x509.CreateCertificate: %v", err)
+	}
+
+	certPath = filepath.Join(dir, prefix+"cert.pem")
+	keyPath = filepath.Join(dir, prefix+"key.pem")
+
+	certPEM := pem.EncodeToMemory(&pem.Block{Type: "CERTIFICATE", Bytes: der})
+	keyPEM := pem.EncodeToMemory(&pem.Block{Type: "RSA PRIVATE KEY", Bytes: x509.MarshalPKCS1PrivateKey(priv)})
+
+	if err := os.WriteFile(certPath, certPEM, 0o600); err != nil {
+		t.Fatalf("write cert: %v", err)
+	}
+	if err := os.WriteFile(keyPath, keyPEM, 0o600); err != nil {
+		t.Fatalf("write key: %v", err)
+	}
+
+	return certPath, keyPath, der
+}
+
+func TestNewKeypairReloader(t *testing.T) {
+	dir := t.TempDir()
+	goodCert, goodKey, _ := writeKeypair(t, dir, "good-")
+
+	garbagePath := filepath.Join(dir, "garbage.pem")
+	if err := os.WriteFile(garbagePath, []byte("not a pem file"), 0o600); err != nil {
+		t.Fatalf("write garbage: %v", err)
+	}
+
+	tests := []struct {
+		name     string
+		certPath string
+		keyPath  string
+		wantErr  error
+		wantNil  bool
+	}{
+		{
+			name:     "valid cert and key",
+			certPath: goodCert,
+			keyPath:  goodKey,
+		},
+		{
+			name:     "missing cert file",
+			certPath: filepath.Join(dir, "does-not-exist.pem"),
+			keyPath:  goodKey,
+			wantErr:  os.ErrNotExist,
+			wantNil:  true,
+		},
+		{
+			name:     "missing key file",
+			certPath: goodCert,
+			keyPath:  filepath.Join(dir, "does-not-exist-key.pem"),
+			wantErr:  os.ErrNotExist,
+			wantNil:  true,
+		},
+		{
+			name:     "cert file is garbage",
+			certPath: garbagePath,
+			keyPath:  goodKey,
+			wantNil:  true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			kpr, err := NewKeypairReloader(tt.certPath, tt.keyPath, discardLogger())
+
+			if tt.wantErr != nil && !errors.Is(err, tt.wantErr) {
+				t.Errorf("err = %v, want errors.Is(..., %v)", err, tt.wantErr)
+			}
+			if tt.wantErr == nil && !tt.wantNil && err != nil {
+				t.Errorf("unexpected err: %v", err)
+			}
+			if tt.wantNil && kpr != nil {
+				t.Errorf("kpr = %+v, want nil", kpr)
+			}
+			if !tt.wantNil && kpr == nil {
+				t.Errorf("kpr is nil, want non-nil")
+			}
+		})
+	}
+}
+
+func TestKeypairReloader_GetCertificate(t *testing.T) {
+	tests := []struct {
+		name string
+		run  func(t *testing.T)
+	}{
+		{
+			name: "returns loaded cert",
+			run: func(t *testing.T) {
+				dir := t.TempDir()
+				certPath, keyPath, wantDER := writeKeypair(t, dir, "a-")
+
+				kpr, err := NewKeypairReloader(certPath, keyPath, discardLogger())
+				if err != nil {
+					t.Fatalf("NewKeypairReloader: %v", err)
+				}
+
+				got, err := kpr.GetCertificate(nil)
+				if err != nil {
+					t.Fatalf("GetCertificate: %v", err)
+				}
+				if len(got.Certificate) == 0 || !bytes.Equal(got.Certificate[0], wantDER) {
+					t.Errorf("GetCertificate returned wrong cert bytes")
+				}
+			},
+		},
+		{
+			name: "reloads when mtime advances",
+			run: func(t *testing.T) {
+				dir := t.TempDir()
+				certPath, keyPath, _ := writeKeypair(t, dir, "a-")
+
+				kpr, err := NewKeypairReloader(certPath, keyPath, discardLogger())
+				if err != nil {
+					t.Fatalf("NewKeypairReloader: %v", err)
+				}
+
+				// Overwrite with a new pair at the same paths and bump mtime.
+				newCertPath, newKeyPath, newDER := writeKeypair(t, dir, "b-")
+				mustRename(t, newCertPath, certPath)
+				mustRename(t, newKeyPath, keyPath)
+				future := time.Now().Add(time.Hour)
+				if err := os.Chtimes(certPath, future, future); err != nil {
+					t.Fatalf("Chtimes: %v", err)
+				}
+
+				got, err := kpr.GetCertificate(nil)
+				if err != nil {
+					t.Fatalf("GetCertificate: %v", err)
+				}
+				if len(got.Certificate) == 0 || !bytes.Equal(got.Certificate[0], newDER) {
+					t.Errorf("GetCertificate did not return reloaded cert")
+				}
+			},
+		},
+		{
+			name: "does not reload when mtime unchanged",
+			run: func(t *testing.T) {
+				dir := t.TempDir()
+				certPath, keyPath, originalDER := writeKeypair(t, dir, "a-")
+
+				kpr, err := NewKeypairReloader(certPath, keyPath, discardLogger())
+				if err != nil {
+					t.Fatalf("NewKeypairReloader: %v", err)
+				}
+
+				// Overwrite the cert/key files with a *different* keypair, then
+				// rewind mtime so the reloader must not pick up the change.
+				newCertPath, newKeyPath, newDER := writeKeypair(t, dir, "b-")
+				mustRename(t, newCertPath, certPath)
+				mustRename(t, newKeyPath, keyPath)
+				past := time.Unix(0, 0)
+				if err := os.Chtimes(certPath, past, past); err != nil {
+					t.Fatalf("Chtimes: %v", err)
+				}
+
+				got, err := kpr.GetCertificate(nil)
+				if err != nil {
+					t.Fatalf("GetCertificate: %v", err)
+				}
+				if len(got.Certificate) == 0 {
+					t.Fatal("empty cert chain")
+				}
+				if bytes.Equal(got.Certificate[0], newDER) {
+					t.Errorf("GetCertificate reloaded despite unchanged mtime")
+				}
+				if !bytes.Equal(got.Certificate[0], originalDER) {
+					t.Errorf("GetCertificate did not return original cert")
+				}
+			},
+		},
+		{
+			name: "does not panic when reload fails after mtime bump",
+			run: func(t *testing.T) {
+				dir := t.TempDir()
+				certPath, keyPath, _ := writeKeypair(t, dir, "a-")
+
+				kpr, err := NewKeypairReloader(certPath, keyPath, discardLogger())
+				if err != nil {
+					t.Fatalf("NewKeypairReloader: %v", err)
+				}
+
+				// Corrupt the cert file and bump mtime. maybeReload will fail.
+				if err := os.WriteFile(certPath, []byte("not a pem file"), 0o600); err != nil {
+					t.Fatalf("corrupt cert: %v", err)
+				}
+				future := time.Now().Add(time.Hour)
+				if err := os.Chtimes(certPath, future, future); err != nil {
+					t.Fatalf("Chtimes: %v", err)
+				}
+
+				defer func() {
+					if r := recover(); r != nil {
+						t.Fatalf("GetCertificate panicked on reload failure: %v", r)
+					}
+				}()
+
+				got, err := kpr.GetCertificate(nil)
+				if err == nil {
+					t.Errorf("GetCertificate returned nil err for corrupt cert; got %+v", got)
+				}
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			tt.run(t)
+		})
+	}
+}
+
+func mustRename(t *testing.T, from, to string) {
+	t.Helper()
+	if err := os.Rename(from, to); err != nil {
+		t.Fatalf("rename %q -> %q: %v", from, to, err)
+	}
+}

lib/metrics/metrics.go

@@ -0,0 +1,123 @@
+package metrics
+
+import (
+	"context"
+	"crypto/tls"
+	"crypto/x509"
+	"errors"
+	"fmt"
+	"log/slog"
+	"net/http"
+	"net/http/pprof"
+	"os"
+	"time"
+
+	"github.com/TecharoHQ/anubis/internal"
+	"github.com/TecharoHQ/anubis/lib/config"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+	healthv1 "google.golang.org/grpc/health/grpc_health_v1"
+)
+
+type Server struct {
+	Config *config.Metrics
+	Log    *slog.Logger
+}
+
+func (s *Server) Run(ctx context.Context, done func()) {
+	defer done()
+	lg := s.Log.With("subsystem", "metrics")
+
+	if err := s.run(ctx, lg); err != nil {
+		lg.Error("can't serve metrics server", "err", err)
+	}
+}
+
+func (s *Server) run(ctx context.Context, lg *slog.Logger) error {
+	mux := http.NewServeMux()
+	mux.HandleFunc("GET /debug/pprof/", pprof.Index)
+	mux.HandleFunc("GET /debug/pprof/cmdline", pprof.Cmdline)
+	mux.HandleFunc("GET /debug/pprof/profile", pprof.Profile)
+	mux.HandleFunc("GET /debug/pprof/symbol", pprof.Symbol)
+	mux.HandleFunc("GET /debug/pprof/trace", pprof.Trace)
+	mux.Handle("/metrics", promhttp.Handler())
+	mux.HandleFunc("/healthz", func(w http.ResponseWriter, r *http.Request) {
+		st, ok := internal.GetHealth("anubis")
+		if !ok {
+			slog.Error("health service anubis does not exist, file a bug")
+		}
+
+		switch st {
+		case healthv1.HealthCheckResponse_NOT_SERVING:
+			http.Error(w, "NOT OK", http.StatusInternalServerError)
+			return
+		case healthv1.HealthCheckResponse_SERVING:
+			fmt.Fprintln(w, "OK")
+			return
+		default:
+			http.Error(w, "UNKNOWN", http.StatusFailedDependency)
+			return
+		}
+	})
+
+	srv := http.Server{
+		Handler:  mux,
+		ErrorLog: internal.GetFilteredHTTPLogger(),
+	}
+
+	ln, metricsURL, err := internal.SetupListener(s.Config.Bind, s.Config.Network, s.Config.SocketMode)
+	if err != nil {
+		return fmt.Errorf("can't setup listener: %w", err)
+	}
+
+	defer ln.Close()
+
+	if s.Config.TLS != nil {
+		kpr, err := NewKeypairReloader(s.Config.TLS.Certificate, s.Config.TLS.Key, lg)
+		if err != nil {
+			return fmt.Errorf("can't setup keypair reloader: %w", err)
+		}
+
+		srv.TLSConfig = &tls.Config{
+			GetCertificate: kpr.GetCertificate,
+		}
+
+		if s.Config.TLS.CA != "" {
+			caCert, err := os.ReadFile(s.Config.TLS.CA)
+			if err != nil {
+				return fmt.Errorf("%w %s: %w", config.ErrCantReadFile, s.Config.TLS.CA, err)
+			}
+
+			certPool := x509.NewCertPool()
+			if !certPool.AppendCertsFromPEM(caCert) {
+				return fmt.Errorf("%w %s", config.ErrInvalidMetricsCACertificate, s.Config.TLS.CA)
+			}
+
+			srv.TLSConfig.ClientCAs = certPool
+			srv.TLSConfig.ClientAuth = tls.RequireAndVerifyClientCert
+		}
+	}
+
+	lg.Debug("listening for metrics", "url", metricsURL)
+
+	go func() {
+		<-ctx.Done()
+		c, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		defer cancel()
+		if err := srv.Shutdown(c); err != nil {
+			lg.Error("can't shut down metrics server", "err", err)
+		}
+	}()
+
+	switch s.Config.TLS != nil {
+	case true:
+		if err := srv.ServeTLS(ln, "", ""); !errors.Is(err, http.ErrServerClosed) {
+			return fmt.Errorf("can't serve TLS metrics server: %w", err)
+		}
+	case false:
+		if err := srv.Serve(ln); !errors.Is(err, http.ErrServerClosed) {
+			return fmt.Errorf("can't serve metrics server: %w", err)
+		}
+	}
+
+	return nil
+}

lib/policy/policy.go

@@ -46,6 +46,7 @@ type ParsedConfig struct {
 	DnsCache          *dns.DnsCache
 	Dns               *dns.Dns
 	Logger            *slog.Logger
+	Metrics           *config.Metrics
 }
 
 func newParsedConfig(orig *config.Config) *ParsedConfig {
@@ -53,6 +54,7 @@ func newParsedConfig(orig *config.Config) *ParsedConfig {
 		orig:        orig,
 		OpenGraph:   orig.OpenGraph,
 		StatusCodes: orig.StatusCodes,
+		Metrics:     orig.Metrics,
 	}
 }
 

web/static/robots.txt

@@ -54,6 +54,7 @@ User-agent: meta-externalagent
 User-agent: Meta-ExternalAgent
 User-agent: meta-externalfetcher
 User-agent: Meta-ExternalFetcher
+User-agent: meta-webindexer
 User-agent: MistralAI-User
 User-agent: MistralAI-User/1.0
 User-agent: MyCentralAIScraperBot