Merge pull request #25044 from karalabe/rpc-histograms

rpc: swap out timer metrics to histograms

rpc/handler.go

@@ -346,7 +346,7 @@ func (h *handler) handleCall(cp *callProc, msg *jsonrpcMessage) *jsonrpcMessage
 			successfulRequestGauge.Inc(1)
 		}
 		rpcServingTimer.UpdateSince(start)
-		newRPCServingTimer(msg.Method, answer.Error == nil).UpdateSince(start)
+		updateServeTimeHistogram(msg.Method, answer.Error == nil, time.Since(start))
 	}
 	return answer
 }

rpc/metrics.go

@@ -18,6 +18,7 @@ package rpc
 
 import (
 	"fmt"
+	"time"
 
 	"github.com/ethereum/go-ethereum/metrics"
 )
@@ -26,14 +27,24 @@ var (
 	rpcRequestGauge        = metrics.NewRegisteredGauge("rpc/requests", nil)
 	successfulRequestGauge = metrics.NewRegisteredGauge("rpc/success", nil)
 	failedRequestGauge     = metrics.NewRegisteredGauge("rpc/failure", nil)
-	rpcServingTimer        = metrics.NewRegisteredTimer("rpc/duration/all", nil)
+
+	// serveTimeHistName is the prefix of the per-request serving time histograms.
+	serveTimeHistName = "rpc/duration"
+
+	rpcServingTimer = metrics.NewRegisteredTimer("rpc/duration/all", nil)
 )
 
-func newRPCServingTimer(method string, valid bool) metrics.Timer {
+// updateServeTimeHistogram tracks the serving time of a remote RPC call.
-	flag := "success"
+func updateServeTimeHistogram(method string, success bool, elapsed time.Duration) {
-	if !valid {
+	note := "success"
-		flag = "failure"
+	if !success {
+		note = "failure"
+	}
+	h := fmt.Sprintf("%s/%s/%s", serveTimeHistName, method, note)
+	sampler := func() metrics.Sample {
+		return metrics.ResettingSample(
+			metrics.NewExpDecaySample(1028, 0.015),
+		)
 	}
-	m := fmt.Sprintf("rpc/duration/%s/%s", method, flag)
+	metrics.GetOrRegisterHistogramLazy(h, nil, sampler).Update(elapsed.Microseconds())
-	return metrics.GetOrRegisterTimer(m, nil)
 }