GL/cmd/dashboard-loadtest/main.go

// Command dashboard-loadtest runs a concurrent read-only HTTP load test against
// a deployed GL dashboard.
package main

import (
	"context"
	"errors"
	"flag"
	"fmt"
	"io"
	"net/http"
	"net/url"
	"os"
	"os/signal"
	"sort"
	"strings"
	"sync"
	"sync/atomic"
	"syscall"
	"time"
)

type result struct {
	Duration time.Duration
	Requests int64
	Errors   int64
	Non2xx   int64
	Latency  []time.Duration
}

func main() {
	baseURL := flag.String("url", "http://127.0.0.1:8080", "dashboard base URL")
	paths := flag.String("paths", "/", "comma-separated request paths")
	concurrency := flag.Int("concurrency", 20, "number of concurrent workers")
	duration := flag.Duration("duration", 15*time.Second, "test duration")
	requestTimeout := flag.Duration("request-timeout", 5*time.Second, "timeout for each request")
	flag.Parse()

	if *concurrency < 1 || *duration <= 0 || *requestTimeout <= 0 {
		fmt.Fprintln(os.Stderr, "concurrency and durations must be positive")
		os.Exit(2)
	}
	targets, err := targets(*baseURL, *paths)
	if err != nil {
		fmt.Fprintln(os.Stderr, err)
		os.Exit(2)
	}

	transport := http.DefaultTransport.(*http.Transport).Clone()
	transport.MaxIdleConns = *concurrency
	transport.MaxIdleConnsPerHost = *concurrency
	client := &http.Client{Transport: transport, Timeout: *requestTimeout}
	defer transport.CloseIdleConnections()

	signalCtx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
	defer stop()
	ctx, cancel := context.WithTimeout(signalCtx, *duration)
	defer cancel()

	fmt.Printf("Dashboard load test: %d workers for %s against %s\n", *concurrency, *duration, *baseURL)
	started := time.Now()
	result := run(ctx, client, targets, *concurrency)
	result.Duration = time.Since(started)
	printResult(result)
	if result.Errors > 0 || result.Non2xx > 0 {
		os.Exit(1)
	}
}

func targets(baseURL, pathList string) ([]string, error) {
	base, err := url.Parse(baseURL)
	if err != nil || base.Scheme == "" || base.Host == "" {
		return nil, fmt.Errorf("invalid dashboard URL %q", baseURL)
	}
	paths := strings.Split(pathList, ",")
	result := make([]string, 0, len(paths))
	for _, value := range paths {
		value = strings.TrimSpace(value)
		if value == "" {
			continue
		}
		reference, parseErr := url.Parse(value)
		if parseErr != nil || reference.IsAbs() || !strings.HasPrefix(reference.Path, "/") {
			return nil, fmt.Errorf("invalid request path %q", value)
		}
		result = append(result, base.ResolveReference(reference).String())
	}
	if len(result) == 0 {
		return nil, errors.New("at least one request path is required")
	}
	return result, nil
}

func run(ctx context.Context, client *http.Client, targets []string, concurrency int) result {
	var (
		requests int64
		failures int64
		non2xx   int64
		sequence uint64
		latency  = make([]time.Duration, 0, concurrency*100)
		lock     sync.Mutex
		workers  sync.WaitGroup
	)
	workers.Add(concurrency)
	for range concurrency {
		go func() {
			defer workers.Done()
			for ctx.Err() == nil {
				target := targets[(atomic.AddUint64(&sequence, 1)-1)%uint64(len(targets))]
				request, err := http.NewRequestWithContext(context.WithoutCancel(ctx), http.MethodGet, target, nil)
				if err != nil {
					atomic.AddInt64(&failures, 1)
					continue
				}
				started := time.Now()
				response, err := client.Do(request)
				elapsed := time.Since(started)
				if err != nil {
					atomic.AddInt64(&requests, 1)
					atomic.AddInt64(&failures, 1)
					continue
				}
				_, copyErr := io.Copy(io.Discard, io.LimitReader(response.Body, 2<<20))
				closeErr := response.Body.Close()
				atomic.AddInt64(&requests, 1)
				if response.StatusCode < 200 || response.StatusCode >= 300 {
					atomic.AddInt64(&non2xx, 1)
				}
				if copyErr != nil || closeErr != nil {
					atomic.AddInt64(&failures, 1)
				}
				lock.Lock()
				latency = append(latency, elapsed)
				lock.Unlock()
			}
		}()
	}
	workers.Wait()
	return result{Requests: requests, Errors: failures, Non2xx: non2xx, Latency: latency}
}

func printResult(value result) {
	seconds := value.Duration.Seconds()
	requestsPerSecond := float64(value.Requests)
	if seconds > 0 {
		requestsPerSecond /= seconds
	}
	fmt.Printf("requests: %d\n", value.Requests)
	fmt.Printf("throughput: %.1f req/s\n", requestsPerSecond)
	fmt.Printf("latency: p50=%s p95=%s p99=%s\n", percentile(value.Latency, 50), percentile(value.Latency, 95), percentile(value.Latency, 99))
	fmt.Printf("errors: %d transport, %d non-2xx\n", value.Errors, value.Non2xx)
}

func percentile(values []time.Duration, percent int) time.Duration {
	if len(values) == 0 {
		return 0
	}
	ordered := append([]time.Duration(nil), values...)
	sort.Slice(ordered, func(left, right int) bool { return ordered[left] < ordered[right] })
	index := (len(ordered)*percent + 99) / 100
	if index < 1 {
		index = 1
	}
	return ordered[index-1]
}