Remove "_ms" suffix from request latency metric name

In practice, the precision of the recorded timing values
depends on the metrics backend and the client library.
Since we're using time.Duration based API, we cannot claim
the unit of measure in the metric name.

rpcmetrics/metrics.go

@@ -43,8 +43,8 @@ type Metrics struct {
 	// Failures is a counter of the number of times any failure has been observed.
 	Failures metrics.Counter `metric:"failures"`
 
-	// RequestLatencyMs is a histogram of the latency of requests in milliseconds.
-	RequestLatencyMs metrics.Timer `metric:"request_latency_ms"`
+	// RequestLatency is a histogram of the latency of requests.
+	RequestLatency metrics.Timer `metric:"request_latency"`
 
 	// HTTPStatusCode2xx is a counter of the total number of requests with HTTP status code 200-299
 	HTTPStatusCode2xx metrics.Counter `metric:"http_status_code_2xx"`

rpcmetrics/observer.go

@@ -157,7 +157,7 @@ func (so *SpanObserver) OnFinish(options opentracing.FinishOptions) {
 	} else {
 		mets.Success.Inc(1)
 	}
-	mets.RequestLatencyMs.Record(options.FinishTime.Sub(so.startTime))
+	mets.RequestLatency.Record(options.FinishTime.Sub(so.startTime))
 	mets.recordHTTPStatusCode(so.httpStatusCode)
 }
 

rpcmetrics/observer_test.go

@@ -127,7 +127,7 @@ func TestObserver(t *testing.T) {
 		// TODO something wrong with string generation, .P99 should not be appended to the tag
 		// as a result we cannot use u.AssertGaugeMetrics
 		_, g := testTracer.metrics.Snapshot()
-		assert.EqualValues(t, 51, g["request_latency_ms|endpoint=get-user.P99"])
+		assert.EqualValues(t, 51, g["request_latency|endpoint=get-user.P99"])
 	})
 }