Add production observability manifests

deploy/prod/kustomization.yaml

@@ -5,6 +5,7 @@ resources:
   - webshare-api-externalsecret.yaml
   - site-analyzer-deployment.yaml
   - site-analyzer-service.yaml
+  - site-analyzer-servicemonitor.yaml
   - site-analyzer-hpa.yaml
   - lightpanda-cdp-deployment.yaml
   - lightpanda-cdp-service.yaml
@@ -14,7 +15,10 @@ resources:
   - lightpanda-mcp-hpa.yaml
   - scraper-deployment.yaml
   - scraper-service.yaml
+  - scraper-servicemonitor.yaml
   - scraper-hpa.yaml
+  - microservices-workload-prometheusrule.yaml
+  - microservices-http-prometheusrule.yaml
   - traefik-middleware-strip-prefix.yaml
   - ingressroute-microservices.yaml
 images:

deploy/prod/lightpanda-cdp-service.yaml

@@ -3,6 +3,8 @@ kind: Service
 metadata:
   name: lightpanda-cdp
   namespace: microservices
+  labels:
+    app: lightpanda-cdp
 spec:
   type: ClusterIP
   selector:

deploy/prod/lightpanda-mcp-service.yaml

@@ -3,6 +3,8 @@ kind: Service
 metadata:
   name: lightpanda-mcp
   namespace: microservices
+  labels:
+    app: lightpanda-mcp
 spec:
   type: ClusterIP
   selector:

deploy/prod/microservices-http-prometheusrule.yaml

@@ -0,0 +1,27 @@
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  name: microservices-http
+  namespace: microservices
+  labels:
+    app: microservices
+spec:
+  groups:
+    - name: microservices.http
+      rules:
+        - alert: MicroserviceHigh5xxRate
+          expr: sum by (service) (rate(http_requests_total{service=~"scraper|site-analyzer",status=~"5.."}[5m])) / clamp_min(sum by (service) (rate(http_requests_total{service=~"scraper|site-analyzer"}[5m])), 0.001) > 0.05
+          for: 10m
+          labels:
+            severity: warning
+          annotations:
+            summary: "Microservice has elevated 5xx errors"
+            description: "{{ $labels.service }} has more than 5% 5xx responses over 10 minutes."
+        - alert: MicroserviceHighP95Latency
+          expr: histogram_quantile(0.95, sum by (service, le) (rate(http_request_duration_seconds_bucket{service=~"scraper|site-analyzer",path!~"/health|/ready"}[5m]))) > 2
+          for: 10m
+          labels:
+            severity: warning
+          annotations:
+            summary: "Microservice has high p95 latency"
+            description: "{{ $labels.service }} p95 HTTP latency is above 2 seconds over 10 minutes."

deploy/prod/microservices-workload-prometheusrule.yaml

@@ -0,0 +1,27 @@
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  name: microservices-workloads
+  namespace: microservices
+  labels:
+    app: microservices
+spec:
+  groups:
+    - name: microservices.workloads
+      rules:
+        - alert: MicroserviceDeploymentUnavailable
+          expr: kube_deployment_status_replicas_unavailable{namespace="microservices",deployment=~"scraper|site-analyzer|lightpanda-cdp|lightpanda-mcp"} > 0
+          for: 5m
+          labels:
+            severity: warning
+          annotations:
+            summary: "Microservice deployment has unavailable replicas"
+            description: "{{ $labels.deployment }} has {{ $value }} unavailable replica(s) in namespace {{ $labels.namespace }}."
+        - alert: MicroservicePodRestarting
+          expr: increase(kube_pod_container_status_restarts_total{namespace="microservices",container=~"scraper|site-analyzer|lightpanda-cdp|lightpanda-mcp"}[15m]) > 2
+          for: 5m
+          labels:
+            severity: warning
+          annotations:
+            summary: "Microservice pod is restarting"
+            description: "{{ $labels.pod }}/{{ $labels.container }} restarted {{ $value }} times in the last 15 minutes."

deploy/prod/scraper-deployment.yaml

@@ -56,7 +56,7 @@ spec:
             periodSeconds: 10
           readinessProbe:
             httpGet:
-              path: /health
+              path: /ready
               port: 8080
             initialDelaySeconds: 2
             periodSeconds: 5

deploy/prod/scraper-service.yaml

@@ -3,6 +3,8 @@ kind: Service
 metadata:
   name: scraper
   namespace: microservices
+  labels:
+    app: scraper
 spec:
   type: ClusterIP
   selector:

deploy/prod/scraper-servicemonitor.yaml

@@ -0,0 +1,16 @@
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: scraper
+  namespace: microservices
+  labels:
+    app: scraper
+spec:
+  selector:
+    matchLabels:
+      app: scraper
+  endpoints:
+    - port: http
+      path: /metrics
+      interval: 30s
+      scrapeTimeout: 10s

deploy/prod/site-analyzer-service.yaml

@@ -3,6 +3,8 @@ kind: Service
 metadata:
   name: site-analyzer
   namespace: microservices
+  labels:
+    app: site-analyzer
 spec:
   type: ClusterIP
   selector:

deploy/prod/site-analyzer-servicemonitor.yaml

@@ -0,0 +1,16 @@
+apiVersion: monitoring.coreos.com/v1
+kind: ServiceMonitor
+metadata:
+  name: site-analyzer
+  namespace: microservices
+  labels:
+    app: site-analyzer
+spec:
+  selector:
+    matchLabels:
+      app: site-analyzer
+  endpoints:
+    - port: http
+      path: /metrics
+      interval: 30s
+      scrapeTimeout: 10s

services/scraper/main.go

@@ -8,9 +8,11 @@ import (
 	"net/http"
 	"net/url"
 	"os"
+	"runtime"
 	"sort"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/PuerkitoBio/goquery"
@@ -46,6 +48,8 @@ type ErrorResponse struct {
 
 const webshareProxyCacheTTL = 5 * time.Minute
 
+var metricsState = newMetricsState()
+
 type webshareProxy struct {
 	Username         string `json:"username"`
 	Password         string `json:"password"`
@@ -75,31 +79,218 @@ func main() {
 		port = "8080"
 	}
 
-	http.HandleFunc("/scrape", handleScrape)
+	http.Handle("/scrape", instrumentHandler("/scrape", http.HandlerFunc(handleScrape)))
-	http.HandleFunc("/health", handleHealth)
+	http.Handle("/health", instrumentHandler("/health", http.HandlerFunc(handleHealth)))
+	http.HandleFunc("/ready", handleHealth)
+	http.HandleFunc("/metrics", handleMetrics)
 
 	log.Printf("scraper listening on :%s", port)
 	log.Fatal(http.ListenAndServe(":"+port, nil))
 }
 
+type metricsStateData struct {
+	startedAt        time.Time
+	inFlight         atomic.Int64
+	proxyErrorsTotal atomic.Int64
+	mu               sync.Mutex
+	httpRequests     map[string]int64
+	httpErrors       map[string]int64
+	httpDuration     map[string]*histogram
+	scraperJobs      map[string]int64
+	scraperDuration  map[string]*histogram
+}
+
+type histogram struct {
+	buckets []float64
+	counts  []int64
+	sum     float64
+	count   int64
+}
+
+type statusRecorder struct {
+	http.ResponseWriter
+	status int
+}
+
+func newMetricsState() *metricsStateData {
+	return &metricsStateData{
+		startedAt:       time.Now(),
+		httpRequests:    map[string]int64{},
+		httpErrors:      map[string]int64{},
+		httpDuration:    map[string]*histogram{},
+		scraperJobs:     map[string]int64{},
+		scraperDuration: map[string]*histogram{},
+	}
+}
+
+func newHistogram() *histogram {
+	buckets := []float64{0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 30}
+	return &histogram{buckets: buckets, counts: make([]int64, len(buckets))}
+}
+
+func (r *statusRecorder) WriteHeader(status int) {
+	r.status = status
+	r.ResponseWriter.WriteHeader(status)
+}
+
+func instrumentHandler(path string, next http.Handler) http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		start := time.Now()
+		metricsState.inFlight.Add(1)
+		defer metricsState.inFlight.Add(-1)
+
+		recorder := &statusRecorder{ResponseWriter: w, status: http.StatusOK}
+		next.ServeHTTP(recorder, r)
+		metricsState.recordHTTP(r.Method, path, recorder.status, time.Since(start).Seconds())
+	})
+}
+
+func (m *metricsStateData) recordHTTP(method string, path string, status int, seconds float64) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	statusCode := fmt.Sprintf("%d", status)
+	key := method + "|" + path + "|" + statusCode
+	m.httpRequests[key]++
+	if status >= 500 {
+		m.httpErrors[key]++
+	}
+	histogramKey := method + "|" + path
+	m.histogram(m.httpDuration, histogramKey).observe(seconds)
+}
+
+func (m *metricsStateData) recordScrape(result string, seconds float64) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	m.scraperJobs[result]++
+	m.histogram(m.scraperDuration, result).observe(seconds)
+}
+
+func (m *metricsStateData) histogram(histograms map[string]*histogram, key string) *histogram {
+	h, ok := histograms[key]
+	if !ok {
+		h = newHistogram()
+		histograms[key] = h
+	}
+	return h
+}
+
+func (h *histogram) observe(seconds float64) {
+	h.count++
+	h.sum += seconds
+	for i, bucket := range h.buckets {
+		if seconds <= bucket {
+			h.counts[i]++
+		}
+	}
+}
+
 func handleHealth(w http.ResponseWriter, r *http.Request) {
 	w.WriteHeader(http.StatusOK)
 	fmt.Fprint(w, "ok")
 }
 
+func handleMetrics(w http.ResponseWriter, r *http.Request) {
+	w.Header().Set("Content-Type", "text/plain; version=0.0.4; charset=utf-8")
+	metricsState.writePrometheus(w)
+}
+
+func (m *metricsStateData) writePrometheus(w io.Writer) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	var runtimeStats runtime.MemStats
+	runtime.ReadMemStats(&runtimeStats)
+
+	fmt.Fprintf(w, "# HELP http_requests_total Total HTTP requests by method, path, and status.\n")
+	fmt.Fprintf(w, "# TYPE http_requests_total counter\n")
+	for key, count := range m.httpRequests {
+		method, path, status := splitMetricKey(key)
+		fmt.Fprintf(w, "http_requests_total{service=\"scraper\",method=\"%s\",path=\"%s\",status=\"%s\"} %d\n", method, path, status, count)
+	}
+
+	fmt.Fprintf(w, "# HELP http_errors_total Total HTTP 5xx responses by method, path, and status.\n")
+	fmt.Fprintf(w, "# TYPE http_errors_total counter\n")
+	for key, count := range m.httpErrors {
+		method, path, status := splitMetricKey(key)
+		fmt.Fprintf(w, "http_errors_total{service=\"scraper\",method=\"%s\",path=\"%s\",status=\"%s\"} %d\n", method, path, status, count)
+	}
+
+	fmt.Fprintf(w, "# HELP http_requests_in_flight Current in-flight HTTP requests.\n")
+	fmt.Fprintf(w, "# TYPE http_requests_in_flight gauge\n")
+	fmt.Fprintf(w, "http_requests_in_flight{service=\"scraper\"} %d\n", m.inFlight.Load())
+
+	fmt.Fprintf(w, "# HELP http_request_duration_seconds HTTP request duration by method and path.\n")
+	fmt.Fprintf(w, "# TYPE http_request_duration_seconds histogram\n")
+	for key, histogram := range m.httpDuration {
+		method, path, _ := splitMetricKey(key)
+		writeHistogram(w, "http_request_duration_seconds", fmt.Sprintf("service=\"scraper\",method=\"%s\",path=\"%s\"", method, path), histogram)
+	}
+
+	fmt.Fprintf(w, "# HELP scraper_jobs_total Total scrape jobs by result.\n")
+	fmt.Fprintf(w, "# TYPE scraper_jobs_total counter\n")
+	for result, count := range m.scraperJobs {
+		fmt.Fprintf(w, "scraper_jobs_total{result=\"%s\"} %d\n", result, count)
+	}
+
+	fmt.Fprintf(w, "# HELP scraper_job_duration_seconds Scrape job duration by result.\n")
+	fmt.Fprintf(w, "# TYPE scraper_job_duration_seconds histogram\n")
+	for result, histogram := range m.scraperDuration {
+		writeHistogram(w, "scraper_job_duration_seconds", fmt.Sprintf("result=\"%s\"", result), histogram)
+	}
+
+	fmt.Fprintf(w, "# HELP scraper_proxy_errors_total Total Webshare proxy selection/configuration errors.\n")
+	fmt.Fprintf(w, "# TYPE scraper_proxy_errors_total counter\n")
+	fmt.Fprintf(w, "scraper_proxy_errors_total %d\n", m.proxyErrorsTotal.Load())
+	fmt.Fprintf(w, "# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.\n")
+	fmt.Fprintf(w, "# TYPE process_start_time_seconds gauge\n")
+	fmt.Fprintf(w, "process_start_time_seconds %.0f\n", float64(m.startedAt.Unix()))
+	fmt.Fprintf(w, "# HELP go_goroutines Number of goroutines that currently exist.\n")
+	fmt.Fprintf(w, "# TYPE go_goroutines gauge\n")
+	fmt.Fprintf(w, "go_goroutines %d\n", runtime.NumGoroutine())
+	fmt.Fprintf(w, "# HELP go_memstats_alloc_bytes Number of bytes allocated and still in use.\n")
+	fmt.Fprintf(w, "# TYPE go_memstats_alloc_bytes gauge\n")
+	fmt.Fprintf(w, "go_memstats_alloc_bytes %d\n", runtimeStats.Alloc)
+}
+
+func writeHistogram(w io.Writer, name string, labels string, histogram *histogram) {
+	for i, bucket := range histogram.buckets {
+		fmt.Fprintf(w, "%s_bucket{%s,le=\"%g\"} %d\n", name, labels, bucket, histogram.counts[i])
+	}
+	fmt.Fprintf(w, "%s_bucket{%s,le=\"+Inf\"} %d\n", name, labels, histogram.count)
+	fmt.Fprintf(w, "%s_sum{%s} %.6f\n", name, labels, histogram.sum)
+	fmt.Fprintf(w, "%s_count{%s} %d\n", name, labels, histogram.count)
+}
+
+func splitMetricKey(key string) (string, string, string) {
+	parts := strings.Split(key, "|")
+	for len(parts) < 3 {
+		parts = append(parts, "")
+	}
+	return parts[0], parts[1], parts[2]
+}
+
 func handleScrape(w http.ResponseWriter, r *http.Request) {
+	start := time.Now()
+	result := "success"
+	defer func() {
+		metricsState.recordScrape(result, time.Since(start).Seconds())
+	}()
+
 	if r.Method != http.MethodPost {
+		result = "error"
 		respondWithError(w, http.StatusMethodNotAllowed, "Method not allowed", "Only POST requests are supported")
 		return
 	}
 
 	var req ScrapeRequest
 	if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
+		result = "error"
 		respondWithError(w, http.StatusBadRequest, "Invalid JSON", err.Error())
 		return
 	}
 
 	if req.URL == "" {
+		result = "error"
 		respondWithError(w, http.StatusBadRequest, "Missing URL", "url field is required")
 		return
 	}
@@ -109,29 +300,32 @@ func handleScrape(w http.ResponseWriter, r *http.Request) {
 		format = "json"
 	}
 	if format != "md" && format != "json" {
+		result = "error"
 		respondWithError(w, http.StatusBadRequest, "Invalid format", "format must be 'md' or 'json'")
 		return
 	}
 
 	country := normalizeCountryCode(req.Country)
 	if err := validateCountry(country); err != nil {
+		result = "error"
 		respondWithError(w, http.StatusBadRequest, "Invalid country", err.Error())
 		return
 	}
 
 	req.URL = normalizeURL(req.URL)
-	result, err := scrapeURL(req.URL, format, country)
+	scrapeResult, err := scrapeURL(req.URL, format, country)
 	if err != nil {
+		result = "error"
 		respondWithError(w, http.StatusInternalServerError, "Scraping failed", err.Error())
 		return
 	}
 
 	if format == "md" {
 		w.Header().Set("Content-Type", "text/plain; charset=utf-8")
-		fmt.Fprint(w, result.Content)
+		fmt.Fprint(w, scrapeResult.Content)
 	} else {
 		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(result)
+		json.NewEncoder(w).Encode(scrapeResult)
 	}
 }
 
@@ -139,10 +333,12 @@ func scrapeURL(url, format string, country string) (*ScrapeResponse, error) {
 	if country != "" {
 		proxy, err := selectWebshareProxy(country)
 		if err != nil {
+			metricsState.proxyErrorsTotal.Add(1)
 			return nil, err
 		}
 		transport, err := webshareProxyTransport(proxy)
 		if err != nil {
+			metricsState.proxyErrorsTotal.Add(1)
 			return nil, err
 		}
 		return scrapeURLWithTransport(url, format, country, transport)

services/site-analyzer/main.go

@@ -12,9 +12,11 @@ import (
 	"net/url"
 	"os"
 	"os/signal"
+	"runtime"
 	"sort"
 	"strconv"
 	"strings"
+	"sync"
 	"sync/atomic"
 	"syscall"
 	"time"
@@ -114,6 +116,7 @@ type app struct {
 }
 
 type metrics struct {
+	startedAt            time.Time
 	requestsTotal        atomic.Int64
 	analyzeRequestsTotal atomic.Int64
 	analyzeRejectedTotal atomic.Int64
@@ -121,6 +124,23 @@ type metrics struct {
 	analyzeDurationCount atomic.Int64
 	analyzeDurationNanos atomic.Int64
 	activeAnalyses       atomic.Int64
+	mu                   sync.Mutex
+	httpRequests         map[string]int64
+	httpErrors           map[string]int64
+	httpDuration         map[string]*histogram
+	analyzeDuration      *histogram
+}
+
+type histogram struct {
+	buckets []float64
+	counts  []int64
+	sum     float64
+	count   int64
+}
+
+type statusRecorder struct {
+	http.ResponseWriter
+	status int
 }
 
 var errUnsafeTarget = errors.New("target resolves to a private or otherwise unsafe address")
@@ -130,9 +150,9 @@ func main() {
 	application := newApp(cfg)
 
 	mux := http.NewServeMux()
-	mux.HandleFunc("/analyze", application.handleAnalyze)
+	mux.Handle("/analyze", application.instrumentHandler("/analyze", http.HandlerFunc(application.handleAnalyze)))
-	mux.HandleFunc("/health", application.handleHealth)
+	mux.Handle("/health", application.instrumentHandler("/health", http.HandlerFunc(application.handleHealth)))
-	mux.HandleFunc("/ready", application.handleReady)
+	mux.Handle("/ready", application.instrumentHandler("/ready", http.HandlerFunc(application.handleReady)))
 	mux.HandleFunc("/metrics", application.handleMetrics)
 
 	server := &http.Server{
@@ -176,7 +196,7 @@ func newApp(cfg config) *app {
 	application := &app{
 		config:    cfg,
 		semaphore: make(chan struct{}, cfg.maxConcurrentAnalyses),
-		metrics:   &metrics{},
+		metrics:   newMetrics(),
 	}
 
 	transport := &http.Transport{
@@ -203,6 +223,73 @@ func newApp(cfg config) *app {
 	return application
 }
 
+func newMetrics() *metrics {
+	return &metrics{
+		startedAt:       time.Now(),
+		httpRequests:    map[string]int64{},
+		httpErrors:      map[string]int64{},
+		httpDuration:    map[string]*histogram{},
+		analyzeDuration: newHistogram(),
+	}
+}
+
+func newHistogram() *histogram {
+	buckets := []float64{0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 30}
+	return &histogram{buckets: buckets, counts: make([]int64, len(buckets))}
+}
+
+func (r *statusRecorder) WriteHeader(status int) {
+	r.status = status
+	r.ResponseWriter.WriteHeader(status)
+}
+
+func (a *app) instrumentHandler(path string, next http.Handler) http.Handler {
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		start := time.Now()
+		recorder := &statusRecorder{ResponseWriter: w, status: http.StatusOK}
+		next.ServeHTTP(recorder, r)
+		a.metrics.recordHTTP(r.Method, path, recorder.status, time.Since(start).Seconds())
+	})
+}
+
+func (m *metrics) recordHTTP(method string, path string, status int, seconds float64) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	statusCode := fmt.Sprintf("%d", status)
+	key := method + "|" + path + "|" + statusCode
+	m.httpRequests[key]++
+	if status >= 500 {
+		m.httpErrors[key]++
+	}
+	histogramKey := method + "|" + path
+	m.histogram(m.httpDuration, histogramKey).observe(seconds)
+}
+
+func (m *metrics) recordAnalyzeDuration(seconds float64) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	m.analyzeDuration.observe(seconds)
+}
+
+func (m *metrics) histogram(histograms map[string]*histogram, key string) *histogram {
+	h, ok := histograms[key]
+	if !ok {
+		h = newHistogram()
+		histograms[key] = h
+	}
+	return h
+}
+
+func (h *histogram) observe(seconds float64) {
+	h.count++
+	h.sum += seconds
+	for i, bucket := range h.buckets {
+		if seconds <= bucket {
+			h.counts[i]++
+		}
+	}
+}
+
 func loadConfig() config {
 	analysisTimeout := durationEnv("ANALYSIS_TIMEOUT", 15*time.Second)
 	fetchTimeout := durationEnv("FETCH_TIMEOUT", 10*time.Second)
@@ -285,32 +372,100 @@ func (a *app) handleReady(w http.ResponseWriter, r *http.Request) {
 }
 
 func (a *app) handleMetrics(w http.ResponseWriter, r *http.Request) {
-	a.metrics.requestsTotal.Add(1)
 	w.Header().Set("Content-Type", "text/plain; version=0.0.4; charset=utf-8")
+	a.metrics.writePrometheus(w, a.config)
+}
 
-	durationSeconds := float64(a.metrics.analyzeDurationNanos.Load()) / float64(time.Second)
+func (m *metrics) writePrometheus(w io.Writer, cfg config) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	var runtimeStats runtime.MemStats
+	runtime.ReadMemStats(&runtimeStats)
+
+	durationSeconds := float64(m.analyzeDurationNanos.Load()) / float64(time.Second)
 	fmt.Fprintf(w, "# HELP site_analyzer_requests_total Total HTTP requests received.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_requests_total counter\n")
-	fmt.Fprintf(w, "site_analyzer_requests_total %d\n", a.metrics.requestsTotal.Load())
+	fmt.Fprintf(w, "site_analyzer_requests_total %d\n", m.requestsTotal.Load())
 	fmt.Fprintf(w, "# HELP site_analyzer_analyze_requests_total Total analyze requests received.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_analyze_requests_total counter\n")
-	fmt.Fprintf(w, "site_analyzer_analyze_requests_total %d\n", a.metrics.analyzeRequestsTotal.Load())
+	fmt.Fprintf(w, "site_analyzer_analyze_requests_total %d\n", m.analyzeRequestsTotal.Load())
 	fmt.Fprintf(w, "# HELP site_analyzer_analyze_rejected_total Analyze requests rejected because the pod is saturated.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_analyze_rejected_total counter\n")
-	fmt.Fprintf(w, "site_analyzer_analyze_rejected_total %d\n", a.metrics.analyzeRejectedTotal.Load())
+	fmt.Fprintf(w, "site_analyzer_analyze_rejected_total %d\n", m.analyzeRejectedTotal.Load())
 	fmt.Fprintf(w, "# HELP site_analyzer_analyze_errors_total Analyze requests that failed before producing a response.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_analyze_errors_total counter\n")
-	fmt.Fprintf(w, "site_analyzer_analyze_errors_total %d\n", a.metrics.analyzeErrorsTotal.Load())
+	fmt.Fprintf(w, "site_analyzer_analyze_errors_total %d\n", m.analyzeErrorsTotal.Load())
 	fmt.Fprintf(w, "# HELP site_analyzer_active_analyses Current number of in-flight analyses.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_active_analyses gauge\n")
-	fmt.Fprintf(w, "site_analyzer_active_analyses %d\n", a.metrics.activeAnalyses.Load())
+	fmt.Fprintf(w, "site_analyzer_active_analyses %d\n", m.activeAnalyses.Load())
 	fmt.Fprintf(w, "# HELP site_analyzer_max_concurrent_analyses Configured maximum in-flight analyses per pod.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_max_concurrent_analyses gauge\n")
-	fmt.Fprintf(w, "site_analyzer_max_concurrent_analyses %d\n", a.config.maxConcurrentAnalyses)
+	fmt.Fprintf(w, "site_analyzer_max_concurrent_analyses %d\n", cfg.maxConcurrentAnalyses)
 	fmt.Fprintf(w, "# HELP site_analyzer_analyze_duration_seconds Total analyze request duration.\n")
 	fmt.Fprintf(w, "# TYPE site_analyzer_analyze_duration_seconds summary\n")
 	fmt.Fprintf(w, "site_analyzer_analyze_duration_seconds_sum %.6f\n", durationSeconds)
-	fmt.Fprintf(w, "site_analyzer_analyze_duration_seconds_count %d\n", a.metrics.analyzeDurationCount.Load())
+	fmt.Fprintf(w, "site_analyzer_analyze_duration_seconds_count %d\n", m.analyzeDurationCount.Load())
+
+	fmt.Fprintf(w, "# HELP http_requests_total Total HTTP requests by method, path, and status.\n")
+	fmt.Fprintf(w, "# TYPE http_requests_total counter\n")
+	for key, count := range m.httpRequests {
+		method, path, status := splitMetricKey(key)
+		fmt.Fprintf(w, "http_requests_total{service=\"site-analyzer\",method=\"%s\",path=\"%s\",status=\"%s\"} %d\n", method, path, status, count)
+	}
+	fmt.Fprintf(w, "# HELP http_errors_total Total HTTP 5xx responses by method, path, and status.\n")
+	fmt.Fprintf(w, "# TYPE http_errors_total counter\n")
+	for key, count := range m.httpErrors {
+		method, path, status := splitMetricKey(key)
+		fmt.Fprintf(w, "http_errors_total{service=\"site-analyzer\",method=\"%s\",path=\"%s\",status=\"%s\"} %d\n", method, path, status, count)
+	}
+	fmt.Fprintf(w, "# HELP http_requests_in_flight Current in-flight HTTP analyze requests.\n")
+	fmt.Fprintf(w, "# TYPE http_requests_in_flight gauge\n")
+	fmt.Fprintf(w, "http_requests_in_flight{service=\"site-analyzer\"} %d\n", m.activeAnalyses.Load())
+	fmt.Fprintf(w, "# HELP http_request_duration_seconds HTTP request duration by method and path.\n")
+	fmt.Fprintf(w, "# TYPE http_request_duration_seconds histogram\n")
+	for key, histogram := range m.httpDuration {
+		method, path, _ := splitMetricKey(key)
+		writeHistogram(w, "http_request_duration_seconds", fmt.Sprintf("service=\"site-analyzer\",method=\"%s\",path=\"%s\"", method, path), histogram)
+	}
+	fmt.Fprintf(w, "# HELP site_analyzer_duration_seconds Analyze request duration.\n")
+	fmt.Fprintf(w, "# TYPE site_analyzer_duration_seconds histogram\n")
+	writeHistogram(w, "site_analyzer_duration_seconds", "", m.analyzeDuration)
+	fmt.Fprintf(w, "# HELP process_start_time_seconds Start time of the process since unix epoch in seconds.\n")
+	fmt.Fprintf(w, "# TYPE process_start_time_seconds gauge\n")
+	fmt.Fprintf(w, "process_start_time_seconds %.0f\n", float64(m.startedAt.Unix()))
+	fmt.Fprintf(w, "# HELP go_goroutines Number of goroutines that currently exist.\n")
+	fmt.Fprintf(w, "# TYPE go_goroutines gauge\n")
+	fmt.Fprintf(w, "go_goroutines %d\n", runtime.NumGoroutine())
+	fmt.Fprintf(w, "# HELP go_memstats_alloc_bytes Number of bytes allocated and still in use.\n")
+	fmt.Fprintf(w, "# TYPE go_memstats_alloc_bytes gauge\n")
+	fmt.Fprintf(w, "go_memstats_alloc_bytes %d\n", runtimeStats.Alloc)
+}
+
+func writeHistogram(w io.Writer, name string, labels string, histogram *histogram) {
+	labelPrefix := ""
+	if labels != "" {
+		labelPrefix = labels + ","
+	}
+	for i, bucket := range histogram.buckets {
+		fmt.Fprintf(w, "%s_bucket{%sle=\"%g\"} %d\n", name, labelPrefix, bucket, histogram.counts[i])
+	}
+	fmt.Fprintf(w, "%s_bucket{%sle=\"+Inf\"} %d\n", name, labelPrefix, histogram.count)
+	if labels == "" {
+		fmt.Fprintf(w, "%s_sum %.6f\n", name, histogram.sum)
+		fmt.Fprintf(w, "%s_count %d\n", name, histogram.count)
+		return
+	}
+	fmt.Fprintf(w, "%s_sum{%s} %.6f\n", name, labels, histogram.sum)
+	fmt.Fprintf(w, "%s_count{%s} %d\n", name, labels, histogram.count)
+}
+
+func splitMetricKey(key string) (string, string, string) {
+	parts := strings.Split(key, "|")
+	for len(parts) < 3 {
+		parts = append(parts, "")
+	}
+	return parts[0], parts[1], parts[2]
 }
 
 func (a *app) handleAnalyze(w http.ResponseWriter, r *http.Request) {
@@ -318,8 +473,10 @@ func (a *app) handleAnalyze(w http.ResponseWriter, r *http.Request) {
 	a.metrics.requestsTotal.Add(1)
 	a.metrics.analyzeRequestsTotal.Add(1)
 	defer func() {
+		duration := time.Since(start)
 		a.metrics.analyzeDurationCount.Add(1)
-		a.metrics.analyzeDurationNanos.Add(time.Since(start).Nanoseconds())
+		a.metrics.analyzeDurationNanos.Add(duration.Nanoseconds())
+		a.metrics.recordAnalyzeDuration(duration.Seconds())
 	}()
 
 	if r.Method != http.MethodPost {