94 lines
2.8 KiB
Go
94 lines
2.8 KiB
Go
package vegeta
|
|
|
|
import (
|
|
"strconv"
|
|
"time"
|
|
"github.com/bmizerany/perks/quantile"
|
|
"fmt"
|
|
"strings"
|
|
)
|
|
|
|
// Metrics holds the stats computed out of a slice of Results
|
|
// that is used for some of the Reporters
|
|
type Metrics struct {
|
|
Latencies struct {
|
|
Mean time.Duration `json:"mean"`
|
|
P95 time.Duration `json:"95th"` // P95 is the 95th percentile upper value
|
|
P99 time.Duration `json:"99th"` // P99 is the 99th percentile upper value
|
|
Max time.Duration `json:"max"`
|
|
} `json:"latencies"`
|
|
|
|
BytesIn struct {
|
|
Total uint64 `json:"total"`
|
|
Mean float64 `json:"mean"`
|
|
} `json:"bytes_in"`
|
|
|
|
BytesOut struct {
|
|
Total uint64 `json:"total"`
|
|
Mean float64 `json:"mean"`
|
|
} `json:"bytes_out"`
|
|
|
|
Duration time.Duration `json:"duration"`
|
|
Requests uint64 `json:"requests"`
|
|
Success float64 `json:"success"`
|
|
StatusCodes map[string]int `json:"status_codes"`
|
|
Errors []string `json:"errors"`
|
|
}
|
|
|
|
// NewMetrics computes and returns a Metrics struct out of a slice of Results
|
|
func NewMetrics(results []Result) *Metrics {
|
|
m := &Metrics{
|
|
Requests: uint64(len(results)),
|
|
StatusCodes: map[string]int{},
|
|
}
|
|
errorSet := map[string]struct{}{}
|
|
quants := quantile.NewTargeted(0.95, 0.99)
|
|
totalSuccess, totalLatencies := 0, time.Duration(0)
|
|
|
|
for _, result := range results {
|
|
quants.Insert(float64(result.Latency))
|
|
m.StatusCodes[strconv.Itoa(int(result.Code))]++
|
|
totalLatencies += result.Latency
|
|
m.BytesOut.Total += result.BytesOut
|
|
m.BytesIn.Total += result.BytesIn
|
|
if result.Latency > m.Latencies.Max {
|
|
m.Latencies.Max = result.Latency
|
|
}
|
|
if result.Code >= 200 && result.Code < 300 {
|
|
totalSuccess++
|
|
}
|
|
if result.Error != "" {
|
|
errorSet[result.Error] = struct{}{}
|
|
}
|
|
}
|
|
|
|
m.Duration = results[len(results)-1].Timestamp.Sub(results[0].Timestamp)
|
|
m.Latencies.Mean = time.Duration(float64(totalLatencies) / float64(m.Requests))
|
|
m.Latencies.P95 = time.Duration(quants.Query(0.95))
|
|
m.Latencies.P99 = time.Duration(quants.Query(0.99))
|
|
m.BytesIn.Mean = float64(m.BytesIn.Total) / float64(m.Requests)
|
|
m.BytesOut.Mean = float64(m.BytesOut.Total) / float64(m.Requests)
|
|
m.Success = float64(totalSuccess) / float64(m.Requests)
|
|
|
|
m.Errors = make([]string, 0, len(errorSet))
|
|
for err := range errorSet {
|
|
m.Errors = append(m.Errors, err)
|
|
}
|
|
|
|
return m
|
|
}
|
|
|
|
|
|
func csvString(d time.Duration) string {
|
|
var result float64 = float64(d.Nanoseconds() / 1000.0 / 1000.0) // in miliseconds
|
|
return strconv.FormatFloat(result, 'f', -1, 64)
|
|
}
|
|
|
|
func(m *Metrics) Csv(rate uint64) []string {
|
|
result := fmt.Sprintf("%d req/s,%s,%s,%s,%s,%f,%f,%.2f",rate,
|
|
csvString(m.Latencies.Mean), csvString(m.Latencies.P95), csvString(m.Latencies.P99), csvString(m.Latencies.Max),
|
|
m.BytesIn.Mean, m.BytesOut.Mean, m.Success * 100)
|
|
return strings.Split(result, ",")
|
|
}
|
|
|