Add prometheus metrics

integration_tests/metrics_test.go

@@ -10,4 +10,26 @@ var _ = Describe("metrics API endpoint", func() {
 		resp := doRequest(newRequest("GET", routerAPIURL("/metrics")))
 		Expect(resp.StatusCode).To(Equal(200))
 	})
+
+	Context("when looking at metrics returned in response", func() {
+		var (
+			responseBody string
+		)
+
+		BeforeEach(func() {
+			resp := doRequest(newRequest("GET", routerAPIURL("/metrics")))
+			Expect(resp.StatusCode).To(Equal(200))
+			responseBody = readBody(resp)
+		})
+
+		It("should contain router internal server error metrics", func() {
+			Skip("Metric is a vector so it is not registered by default")
+			Expect(responseBody).To(ContainSubstring("router_internal_server_error_count"))
+		})
+
+		It("should contain routing table metrics", func() {
+			Expect(responseBody).To(ContainSubstring("router_route_reload_count"))
+			Expect(responseBody).To(ContainSubstring("router_route_reload_error_count"))
+		})
+	})
 })

main.go

@@ -88,6 +88,8 @@ func main() {
 		os.Exit(0)
 	}
 
+	initMetrics()
+
 	logInfo(fmt.Sprintf("router: using GOMAXPROCS value of %d", runtime.GOMAXPROCS(0)))
 
 	if tlsSkipVerify {

metrics.go

@@ -0,0 +1,36 @@
+package main
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+var (
+	internalServerErrorCountMetric = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "router_internal_server_error_count",
+			Help: "Number of internal server errors encountered by router",
+		},
+		[]string{"host"},
+	)
+
+	routeReloadCountMetric = prometheus.NewCounter(
+		prometheus.CounterOpts{
+			Name: "router_route_reload_count",
+			Help: "Number of times routing table has been reloaded",
+		},
+	)
+
+	routeReloadErrorCountMetric = prometheus.NewCounter(
+		prometheus.CounterOpts{
+			Name: "router_route_reload_error_count",
+			Help: "Number of errors encountered by reloading routing table",
+		},
+	)
+)
+
+func initMetrics() {
+	prometheus.MustRegister(internalServerErrorCountMetric)
+
+	prometheus.MustRegister(routeReloadCountMetric)
+	prometheus.MustRegister(routeReloadErrorCountMetric)
+}

router.go

@@ -7,6 +7,8 @@ import (
 	"sync"
 	"time"
 
+	"github.com/prometheus/client_golang/prometheus"
+
 	"github.com/alphagov/router/handlers"
 	"github.com/alphagov/router/logger"
 	"github.com/alphagov/router/triemux"
@@ -78,11 +80,16 @@ func (rt *Router) ServeHTTP(w http.ResponseWriter, req *http.Request) {
 	defer func() {
 		if r := recover(); r != nil {
 			logWarn("router: recovered from panic in ServeHTTP:", r)
+
 			errorMessage := fmt.Sprintf("panic: %v", r)
 			err := logger.RecoveredError{ErrorMessage: errorMessage}
+
 			logger.NotifySentry(logger.ReportableError{Error: err, Request: req})
 			rt.logger.LogFromClientRequest(map[string]interface{}{"error": errorMessage, "status": 500}, req)
+
 			w.WriteHeader(http.StatusInternalServerError)
+
+			internalServerErrorCountMetric.With(prometheus.Labels{"host": req.Host}).Inc()
 		}
 	}()
 	rt.lock.RLock()
@@ -96,13 +103,17 @@ func (rt *Router) ServeHTTP(w http.ResponseWriter, req *http.Request) {
 // create a new proxy mux, load applications (backends) and routes into it, and
 // then flip the "mux" pointer in the Router.
 func (rt *Router) ReloadRoutes() {
+	routeReloadCountMetric.Inc()
+
 	defer func() {
 		if r := recover(); r != nil {
 			logWarn("router: recovered from panic in ReloadRoutes:", r)
 			logInfo("router: original routes have not been modified")
 			errorMessage := fmt.Sprintf("panic: %v", r)
 			err := logger.RecoveredError{ErrorMessage: errorMessage}
 			logger.NotifySentry(logger.ReportableError{Error: err})
+
+			routeReloadErrorCountMetric.Inc()
 		}
 	}()