hsbench/hsbench.go

1066 lines
28 KiB
Go

// hsbench.go
// Copyright (c) 2017 Wasabi Technology, Inc.
// Copyright (c) 2019 Red Hat Inc.
package main
import (
"bytes"
"code.cloudfoundry.org/bytefmt"
"crypto/hmac"
"crypto/md5"
"crypto/sha1"
"crypto/tls"
"encoding/base64"
"encoding/csv"
"encoding/json"
"flag"
"fmt"
"github.com/aws/aws-sdk-go/aws"
"github.com/aws/aws-sdk-go/aws/credentials"
"github.com/aws/aws-sdk-go/aws/session"
"github.com/aws/aws-sdk-go/service/s3"
"io"
"log"
"math"
"math/rand"
"net"
"net/http"
"os"
"sort"
"strconv"
"strings"
"sync"
"sync/atomic"
"time"
)
// Global variables
var access_key, secret_key, url_host, bucket_prefix, bucket_list, object_prefix, region, modes, output, json_output, sizeArg string
var buckets []string
var duration_secs, threads, loops int
var object_data []byte
var object_data_md5 string
var max_keys, running_threads, bucket_count, first_object, object_count, object_size, op_counter int64
var object_count_flag bool
var endtime time.Time
var interval float64
// Our HTTP transport used for the roundtripper below
var HTTPTransport http.RoundTripper = &http.Transport{
Proxy: http.ProxyFromEnvironment,
Dial: (&net.Dialer{
Timeout: 30 * time.Second,
KeepAlive: 30 * time.Second,
}).Dial,
TLSHandshakeTimeout: 10 * time.Second,
ExpectContinueTimeout: 0,
// Set the number of idle connections to 2X the number of threads
MaxIdleConnsPerHost: 2*threads,
MaxIdleConns: 2*threads,
// But limit their idle time to 1 minute
IdleConnTimeout: time.Minute,
// Ignore TLS errors
TLSClientConfig: &tls.Config{InsecureSkipVerify: true},
}
var httpClient = &http.Client{Transport: HTTPTransport}
func getS3Client() *s3.S3 {
// Build our config
creds := credentials.NewStaticCredentials(access_key, secret_key, "")
loglevel := aws.LogOff
// Build the rest of the configuration
awsConfig := &aws.Config{
Region: aws.String(region),
Endpoint: aws.String(url_host),
Credentials: creds,
LogLevel: &loglevel,
S3ForcePathStyle: aws.Bool(true),
S3Disable100Continue: aws.Bool(true),
// Comment following to use default transport
HTTPClient: &http.Client{Transport: HTTPTransport},
}
session := session.New(awsConfig)
client := s3.New(session)
if client == nil {
log.Fatalf("FATAL: Unable to create new client.")
}
// Return success
return client
}
// canonicalAmzHeaders -- return the x-amz headers canonicalized
func canonicalAmzHeaders(req *http.Request) string {
// Parse out all x-amz headers
var headers []string
for header := range req.Header {
norm := strings.ToLower(strings.TrimSpace(header))
if strings.HasPrefix(norm, "x-amz") {
headers = append(headers, norm)
}
}
// Put them in sorted order
sort.Strings(headers)
// Now add back the values
for n, header := range headers {
headers[n] = header + ":" + strings.Replace(req.Header.Get(header), "\n", " ", -1)
}
// Finally, put them back together
if len(headers) > 0 {
return strings.Join(headers, "\n") + "\n"
} else {
return ""
}
}
func hmacSHA1(key []byte, content string) []byte {
mac := hmac.New(sha1.New, key)
mac.Write([]byte(content))
return mac.Sum(nil)
}
func setSignature(req *http.Request) {
// Setup default parameters
dateHdr := time.Now().UTC().Format("20060102T150405Z")
req.Header.Set("X-Amz-Date", dateHdr)
// Get the canonical resource and header
canonicalResource := req.URL.EscapedPath()
canonicalHeaders := canonicalAmzHeaders(req)
stringToSign := req.Method + "\n" + req.Header.Get("Content-MD5") + "\n" + req.Header.Get("Content-Type") + "\n\n" +
canonicalHeaders + canonicalResource
hash := hmacSHA1([]byte(secret_key), stringToSign)
signature := base64.StdEncoding.EncodeToString(hash)
req.Header.Set("Authorization", fmt.Sprintf("AWS %s:%s", access_key, signature))
}
type IntervalStats struct {
loop int
name string
mode string
bytes int64
slowdowns int64
intervalNano int64
latNano []int64
}
func (is *IntervalStats) makeOutputStats() OutputStats {
// Compute and log the stats
ops := len(is.latNano)
totalLat := int64(0)
minLat := float64(0)
maxLat := float64(0)
NinetyNineLat := float64(0)
avgLat := float64(0)
if ops > 0 {
minLat = float64(is.latNano[0]) / 1000000
maxLat = float64(is.latNano[ops-1]) / 1000000
for i := range is.latNano {
totalLat += is.latNano[i]
}
avgLat = float64(totalLat) / float64(ops) / 1000000
NintyNineLatNano := is.latNano[int64(math.Round(0.99*float64(ops)))-1]
NinetyNineLat = float64(NintyNineLatNano) / 1000000
}
seconds := float64(is.intervalNano) / 1000000000
mbps := float64(is.bytes) / seconds / bytefmt.MEGABYTE
iops := float64(ops) / seconds
return OutputStats{
is.loop,
is.name,
seconds,
is.mode,
ops,
mbps,
iops,
minLat,
avgLat,
NinetyNineLat,
maxLat,
is.slowdowns}
}
type OutputStats struct {
Loop int
IntervalName string
Seconds float64
Mode string
Ops int
Mbps float64
Iops float64
MinLat float64
AvgLat float64
NinetyNineLat float64
MaxLat float64
Slowdowns int64
}
func (o *OutputStats) log() {
log.Printf(
"Loop: %d, Int: %s, Dur(s): %.1f, Mode: %s, Ops: %d, MB/s: %.2f, IO/s: %.0f, Lat(ms): [ min: %.1f, avg: %.1f, 99%%: %.1f, max: %.1f ], Slowdowns: %d",
o.Loop,
o.IntervalName,
o.Seconds,
o.Mode,
o.Ops,
o.Mbps,
o.Iops,
o.MinLat,
o.AvgLat,
o.NinetyNineLat,
o.MaxLat,
o.Slowdowns)
}
func (o *OutputStats) csv_header(w *csv.Writer) {
if w == nil {
log.Fatal("OutputStats passed nil CSV writer")
}
s := []string{
"Loop",
"Inteval",
"Duration(s)",
"Mode", "Ops",
"MB/s",
"IO/s",
"Min Latency (ms)",
"Avg Latency(ms)",
"99% Latency(ms)",
"Max Latency(ms)",
"Slowdowns"}
if err := w.Write(s); err != nil {
log.Fatal("Error writing to CSV writer: ", err)
}
}
func (o *OutputStats) csv(w *csv.Writer) {
if w == nil {
log.Fatal("OutputStats Passed nil csv writer")
}
s := []string{
strconv.Itoa(o.Loop),
o.IntervalName,
strconv.FormatFloat(o.Seconds, 'f', 2, 64),
o.Mode,
strconv.Itoa(o.Ops),
strconv.FormatFloat(o.Mbps, 'f', 2, 64),
strconv.FormatFloat(o.Iops, 'f', 2, 64),
strconv.FormatFloat(o.MinLat, 'f', 2, 64),
strconv.FormatFloat(o.AvgLat, 'f', 2, 64),
strconv.FormatFloat(o.NinetyNineLat, 'f', 2, 64),
strconv.FormatFloat(o.MaxLat, 'f', 2, 64),
strconv.FormatInt(o.Slowdowns, 10)}
if err := w.Write(s); err != nil {
log.Fatal("Error writing to CSV writer: ", err)
}
}
func (o *OutputStats) json(jfile *os.File) {
if jfile == nil {
log.Fatal("OutputStats passed nil JSON file")
}
jdata, err := json.Marshal(o)
if err != nil {
log.Fatal("Error marshaling JSON: ", err)
}
log.Println(string(jdata))
_, err = jfile.WriteString(string(jdata) + "\n")
if err != nil {
log.Fatal("Error writing to JSON file: ", err)
}
}
type ThreadStats struct {
start int64
curInterval int64
intervals []IntervalStats
}
func makeThreadStats(s int64, loop int, mode string, intervalNano int64) ThreadStats {
ts := ThreadStats{s, 0, []IntervalStats{}}
ts.intervals = append(ts.intervals, IntervalStats{loop, "0", mode, 0, 0, intervalNano, []int64{}})
return ts
}
func (ts *ThreadStats) updateIntervals(loop int, mode string, intervalNano int64) int64 {
// Interval statistics disabled, so just return the current interval
if intervalNano < 0 {
return ts.curInterval
}
for ts.start+intervalNano*(ts.curInterval+1) < time.Now().UnixNano() {
ts.curInterval++
ts.intervals = append(
ts.intervals,
IntervalStats{
loop,
strconv.FormatInt(ts.curInterval, 10),
mode,
0,
0,
intervalNano,
[]int64{}})
}
return ts.curInterval
}
func (ts *ThreadStats) finish() {
ts.curInterval = -1
}
type Stats struct {
// threads
threads int
// The loop we are in
loop int
// Test mode being run
mode string
// start time in nanoseconds
startNano int64
// end time in nanoseconds
endNano int64
// Duration in nanoseconds for each interval
intervalNano int64
// Per-thread statistics
threadStats []ThreadStats
// a map of per-interval thread completion counters
intervalCompletions sync.Map
// a counter of how many threads have finished updating stats entirely
completions int32
}
func makeStats(loop int, mode string, threads int, intervalNano int64) Stats {
start := time.Now().UnixNano()
s := Stats{threads, loop, mode, start, 0, intervalNano, []ThreadStats{}, sync.Map{}, 0}
for i := 0; i < threads; i++ {
s.threadStats = append(s.threadStats, makeThreadStats(start, s.loop, s.mode, s.intervalNano))
s.updateIntervals(i)
}
return s
}
func (stats *Stats) makeOutputStats(i int64) (OutputStats, bool) {
// Check bounds first
if stats.intervalNano < 0 || i < 0 {
return OutputStats{}, false
}
// Not safe to log if not all writers have completed.
value, ok := stats.intervalCompletions.Load(i)
if !ok {
return OutputStats{}, false
}
cp, ok := value.(*int32)
if !ok {
return OutputStats{}, false
}
count := atomic.LoadInt32(cp)
if count < int32(stats.threads) {
return OutputStats{}, false
}
bytes := int64(0)
ops := int64(0)
slowdowns := int64(0)
for t := 0; t < stats.threads; t++ {
bytes += stats.threadStats[t].intervals[i].bytes
ops += int64(len(stats.threadStats[t].intervals[i].latNano))
slowdowns += stats.threadStats[t].intervals[i].slowdowns
}
// Aggregate the per-thread Latency slice
tmpLat := make([]int64, ops)
var c int
for t := 0; t < stats.threads; t++ {
c += copy(tmpLat[c:], stats.threadStats[t].intervals[i].latNano)
}
sort.Slice(tmpLat, func(i, j int) bool { return tmpLat[i] < tmpLat[j] })
is := IntervalStats{stats.loop, strconv.FormatInt(i, 10), stats.mode, bytes, slowdowns, stats.intervalNano, tmpLat}
return is.makeOutputStats(), true
}
func (stats *Stats) makeTotalStats() (OutputStats, bool) {
// Not safe to log if not all writers have completed.
completions := atomic.LoadInt32(&stats.completions)
if completions < int32(threads) {
log.Printf("log, completions: %d", completions)
return OutputStats{}, false
}
bytes := int64(0)
ops := int64(0)
slowdowns := int64(0)
for t := 0; t < stats.threads; t++ {
for i := 0; i < len(stats.threadStats[t].intervals); i++ {
bytes += stats.threadStats[t].intervals[i].bytes
ops += int64(len(stats.threadStats[t].intervals[i].latNano))
slowdowns += stats.threadStats[t].intervals[i].slowdowns
}
}
// Aggregate the per-thread Latency slice
tmpLat := make([]int64, ops)
var c int
for t := 0; t < stats.threads; t++ {
for i := 0; i < len(stats.threadStats[t].intervals); i++ {
c += copy(tmpLat[c:], stats.threadStats[t].intervals[i].latNano)
}
}
sort.Slice(tmpLat, func(i, j int) bool { return tmpLat[i] < tmpLat[j] })
is := IntervalStats{stats.loop, "TOTAL", stats.mode, bytes, slowdowns, stats.endNano - stats.startNano, tmpLat}
return is.makeOutputStats(), true
}
// Only safe to call from the calling thread
func (stats *Stats) updateIntervals(thread_num int) int64 {
curInterval := stats.threadStats[thread_num].curInterval
newInterval := stats.threadStats[thread_num].updateIntervals(stats.loop, stats.mode, stats.intervalNano)
// Finish has already been called
if curInterval < 0 {
return -1
}
for i := curInterval; i < newInterval; i++ {
// load or store the current value
value, _ := stats.intervalCompletions.LoadOrStore(i, new(int32))
cp, ok := value.(*int32)
if !ok {
log.Printf("updateIntervals: got data of type %T but wanted *int32", value)
continue
}
count := atomic.AddInt32(cp, 1)
if count == int32(stats.threads) {
if is, ok := stats.makeOutputStats(i); ok {
is.log()
}
}
}
return newInterval
}
func (stats *Stats) addOp(thread_num int, bytes int64, latNano int64) {
// Interval statistics
cur := stats.threadStats[thread_num].curInterval
if cur < 0 {
return
}
stats.threadStats[thread_num].intervals[cur].bytes += bytes
stats.threadStats[thread_num].intervals[cur].latNano =
append(stats.threadStats[thread_num].intervals[cur].latNano, latNano)
}
func (stats *Stats) addSlowDown(thread_num int) {
cur := stats.threadStats[thread_num].curInterval
stats.threadStats[thread_num].intervals[cur].slowdowns++
}
func (stats *Stats) finish(thread_num int) {
stats.updateIntervals(thread_num)
stats.threadStats[thread_num].finish()
count := atomic.AddInt32(&stats.completions, 1)
if count == int32(stats.threads) {
stats.endNano = time.Now().UnixNano()
}
}
func runUpload(thread_num int, fendtime time.Time, stats *Stats) {
errcnt := 0
svc := s3.New(session.New(), cfg)
for {
if duration_secs > -1 && time.Now().After(endtime) {
break
}
objnum := atomic.AddInt64(&op_counter, 1)
bucket_num := objnum % int64(bucket_count)
if object_count > -1 && objnum >= object_count {
objnum = atomic.AddInt64(&op_counter, -1)
break
}
fileobj := bytes.NewReader(object_data)
key := fmt.Sprintf("%s%012d", object_prefix, objnum)
r := &s3.PutObjectInput{
Bucket: &buckets[bucket_num],
Key: &key,
Body: fileobj,
}
start := time.Now().UnixNano()
req, _ := svc.PutObjectRequest(r)
// Disable payload checksum calculation (very expensive)
req.HTTPRequest.Header.Add("X-Amz-Content-Sha256", "UNSIGNED-PAYLOAD")
err := req.Send()
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
errcnt++
stats.addSlowDown(thread_num)
atomic.AddInt64(&op_counter, -1)
log.Printf("upload err", err)
} else {
// Update the stats
stats.addOp(thread_num, object_size, end-start)
}
if errcnt > 2 {
break
}
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
func readBody(r io.Reader) (int64, error) {
var bytesRead int64 = 0
buf := make([]byte, 65536)
for {
n, err := r.Read(buf)
if n > 0 {
bytesRead += int64(n)
}
if err != nil {
if err == io.EOF {
return bytesRead, nil
} else {
return bytesRead, err
}
}
}
}
func runDownload(thread_num int, fendtime time.Time, stats *Stats) {
errcnt := 0
svc := s3.New(session.New(), cfg)
for {
if duration_secs > -1 && time.Now().After(endtime) {
break
}
var objnum int64
if object_count > -1 {
// Run random download if the number of objects is known
objnum = rand.Int63() % object_count
} else {
objnum = atomic.AddInt64(&op_counter, 1)
if object_count > -1 && objnum >= object_count {
atomic.AddInt64(&op_counter, -1)
break
}
}
bucket_num := objnum % int64(bucket_count)
key := fmt.Sprintf("%s%012d", object_prefix, objnum)
r := &s3.GetObjectInput{
Bucket: &buckets[bucket_num],
Key: &key,
}
start := time.Now().UnixNano()
req, resp := svc.GetObjectRequest(r)
err := req.Send()
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
errcnt++
stats.addSlowDown(thread_num)
log.Printf("download err", err)
} else {
var bytesRead int64 = 0
defer resp.Body.Close()
bytesRead, err := readBody(resp.Body)
// Update the stats
stats.addOp(thread_num, bytesRead, end-start)
if err != nil {
errcnt++
stats.addSlowDown(thread_num)
log.Printf("download err", err)
}
}
if errcnt > 2 {
break
}
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
func runDelete(thread_num int, stats *Stats) {
errcnt := 0
svc := s3.New(session.New(), cfg)
for {
if duration_secs > -1 && time.Now().After(endtime) {
break
}
objnum := atomic.AddInt64(&op_counter, 1)
if object_count > -1 && objnum >= object_count {
atomic.AddInt64(&op_counter, -1)
break
}
bucket_num := objnum % int64(bucket_count)
key := fmt.Sprintf("%s%012d", object_prefix, objnum)
r := &s3.DeleteObjectInput{
Bucket: &buckets[bucket_num],
Key: &key,
}
start := time.Now().UnixNano()
req, out := svc.DeleteObjectRequest(r)
err := req.Send()
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
errcnt++
stats.addSlowDown(thread_num)
log.Printf("delete err", err, "out", out.String())
} else {
// Update the stats
stats.addOp(thread_num, object_size, end-start)
}
if errcnt > 2 {
break
}
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
func runBucketDelete(thread_num int, stats *Stats) {
svc := s3.New(session.New(), cfg)
for {
bucket_num := atomic.AddInt64(&op_counter, 1)
if bucket_num >= bucket_count {
atomic.AddInt64(&op_counter, -1)
break
}
r := &s3.DeleteBucketInput{
Bucket: &buckets[bucket_num],
}
start := time.Now().UnixNano()
_, err := svc.DeleteBucket(r)
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
break
}
stats.addOp(thread_num, 0, end-start)
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
func runBucketList(thread_num int, stats *Stats) {
svc := s3.New(session.New(), cfg)
marker := ""
bucket_num := rand.Int63() % bucket_count
for {
if duration_secs > -1 && time.Now().After(endtime) {
break
}
start := time.Now().UnixNano()
p, err := svc.ListObjects(&s3.ListObjectsInput{
Bucket: &buckets[bucket_num],
Marker: &marker,
MaxKeys: &max_keys,
})
end := time.Now().UnixNano()
if err != nil {
break
}
stats.addOp(thread_num, 0, end-start)
stats.updateIntervals(thread_num)
if *p.IsTruncated {
marker = *p.NextMarker
} else {
marker = ""
bucket_num = rand.Int63() % bucket_count
}
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
var cfg *aws.Config
func runBucketsInit(thread_num int, stats *Stats) {
svc := s3.New(session.New(), cfg)
for {
bucket_num := atomic.AddInt64(&op_counter, 1)
if bucket_num >= bucket_count {
atomic.AddInt64(&op_counter, -1)
break
}
start := time.Now().UnixNano()
in := &s3.CreateBucketInput{Bucket: aws.String(buckets[bucket_num])}
_, err := svc.CreateBucket(in)
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
if !strings.Contains(err.Error(), s3.ErrCodeBucketAlreadyOwnedByYou) &&
!strings.Contains(err.Error(), "BucketAlreadyExists") {
log.Fatalf("FATAL: Unable to create bucket %s (is your access and secret correct?): %v", buckets[bucket_num], err)
}
}
stats.addOp(thread_num, 0, end-start)
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
type pagedObject struct {
bucket_num int64
key string
size int64
}
func runPagedList(wg *sync.WaitGroup, bucket_num int64, list chan<- pagedObject) {
svc := s3.New(session.New(), cfg)
svc.ListObjectsPages(
&s3.ListObjectsInput{
Bucket: &buckets[bucket_num],
MaxKeys: &max_keys,
},
func(page *s3.ListObjectsOutput, last bool) bool {
for _, v := range page.Contents {
list <- pagedObject{
bucket_num: bucket_num,
key: *v.Key,
size: *v.Size,
}
}
return true
})
wg.Done()
}
func runBucketsClear(list <-chan pagedObject, thread_num int, stats *Stats) {
svc := s3.New(session.New(), cfg)
for {
v := <-list
start := time.Now().UnixNano()
_, err := svc.DeleteObject(&s3.DeleteObjectInput{
Bucket: &buckets[v.bucket_num],
Key: &v.key,
})
end := time.Now().UnixNano()
stats.updateIntervals(thread_num)
if err != nil {
break
}
stats.addOp(thread_num, v.size, end-start)
}
stats.finish(thread_num)
atomic.AddInt64(&running_threads, -1)
}
func runWrapper(loop int, r rune) []OutputStats {
op_counter = first_object-1
running_threads = int64(threads)
intervalNano := int64(interval * 1000000000)
endtime = time.Now().Add(time.Second * time.Duration(duration_secs))
var stats Stats
// If we perviously set the object count after running a put
// test, set the object count back to -1 for the new put test.
if r == 'p' && object_count_flag {
object_count = -1
object_count_flag = false
}
switch r {
case 'c':
log.Printf("Running Loop %d BUCKET CLEAR TEST", loop)
stats = makeStats(loop, "BCLR", threads, intervalNano)
list := make(chan pagedObject, threads*2)
var wg = sync.WaitGroup{}
for b := int64(0); b < bucket_count; b++ {
wg.Add(1)
go runPagedList(&wg, b, list)
}
for n := 0; n < threads; n++ {
go runBucketsClear(list, n, &stats)
}
wg.Wait()
close(list)
case 'x':
log.Printf("Running Loop %d BUCKET DELETE TEST", loop)
stats = makeStats(loop, "BDEL", threads, intervalNano)
for n := 0; n < threads; n++ {
go runBucketDelete(n, &stats)
}
case 'i':
log.Printf("Running Loop %d BUCKET INIT TEST", loop)
stats = makeStats(loop, "BINIT", threads, intervalNano)
for n := 0; n < threads; n++ {
go runBucketsInit(n, &stats)
}
case 'p':
log.Printf("Running Loop %d OBJECT PUT TEST", loop)
stats = makeStats(loop, "PUT", threads, intervalNano)
for n := 0; n < threads; n++ {
go runUpload(n, endtime, &stats)
}
case 'l':
log.Printf("Running Loop %d BUCKET LIST TEST", loop)
stats = makeStats(loop, "LIST", threads, intervalNano)
for n := 0; n < threads; n++ {
go runBucketList(n, &stats)
}
case 'g':
log.Printf("Running Loop %d OBJECT GET TEST", loop)
stats = makeStats(loop, "GET", threads, intervalNano)
for n := 0; n < threads; n++ {
go runDownload(n, endtime, &stats)
}
case 'd':
log.Printf("Running Loop %d OBJECT DELETE TEST", loop)
stats = makeStats(loop, "DEL", threads, intervalNano)
for n := 0; n < threads; n++ {
go runDelete(n, &stats)
}
}
// Wait for it to finish
for atomic.LoadInt64(&running_threads) > 0 {
time.Sleep(time.Millisecond)
}
// If the user didn't set the object_count, we can set it here
// to limit subsequent get/del tests to valid objects only.
if r == 'p' && object_count < 0 {
object_count = op_counter + 1
object_count_flag = true
}
// Create the Output Stats
os := make([]OutputStats, 0)
for i := int64(0); i >= 0; i++ {
if o, ok := stats.makeOutputStats(i); ok {
os = append(os, o)
} else {
break
}
}
if o, ok := stats.makeTotalStats(); ok {
o.log()
os = append(os, o)
}
return os
}
func init() {
// Parse command line
myflag := flag.NewFlagSet("myflag", flag.ExitOnError)
myflag.StringVar(&access_key, "a", os.Getenv("AWS_ACCESS_KEY_ID"), "Access key")
myflag.StringVar(&secret_key, "s", os.Getenv("AWS_SECRET_ACCESS_KEY"), "Secret key")
myflag.StringVar(&url_host, "u", os.Getenv("AWS_HOST"), "URL for host with method prefix")
myflag.StringVar(&object_prefix, "op", "", "Prefix for objects")
myflag.StringVar(&bucket_prefix, "bp", "hotsauce-bench", "Prefix for buckets")
myflag.StringVar(&bucket_list, "bl", "", "Use space-separated list of buckets for testing, not <prefix>000000000000")
myflag.StringVar(&region, "r", "us-east-1", "Region for testing")
myflag.StringVar(&modes, "m", "cxiplgdcx", "Run modes in order. See NOTES for more info")
myflag.StringVar(&output, "o", "", "Write CSV output to this file")
myflag.StringVar(&json_output, "j", "", "Write JSON output to this file")
myflag.Int64Var(&max_keys, "mk", 1000, "Maximum number of keys to retreive at once for bucket listings")
myflag.Int64Var(&object_count, "n", -1, "Maximum number of objects <-1 for unlimited>")
myflag.Int64Var(&first_object, "f", 0, "Object number to start with")
myflag.Int64Var(&bucket_count, "b", 1, "Number of buckets to distribute IOs across")
myflag.IntVar(&duration_secs, "d", 60, "Maximum test duration in seconds <-1 for unlimited>")
myflag.IntVar(&threads, "t", 1, "Number of threads to run")
myflag.IntVar(&loops, "l", 1, "Number of times to repeat test")
myflag.StringVar(&sizeArg, "z", "1M", "Size of objects in bytes with postfix K, M, and G")
myflag.Float64Var(&interval, "ri", 1.0, "Number of seconds between report intervals")
// define custom usage output with notes
notes :=
`
NOTES:
- Valid mode types for the -m mode string are:
c: clear all existing objects from buckets (requires lookups)
x: delete buckets
i: initialize buckets
p: put objects in buckets
l: list objects in buckets
g: get objects from buckets (randomly when object count is known, sequentially otherwise)
d: delete objects from buckets
These modes are processed in-order and can be repeated, ie "ippgd" will
initialize the buckets, put the objects, reput the objects, get the
objects, and then delete the objects. The repeat flag will repeat this
whole process the specified number of times.
- When performing bucket listings, many S3 storage systems limit the
maximum number of keys returned to 1000 even if MaxKeys is set higher.
hsbench will attempt to set MaxKeys to whatever value is passed via the
"mk" flag, but it's likely that any values above 1000 will be ignored.
`
myflag.Usage = func() {
fmt.Fprintf(flag.CommandLine.Output(), "\nUSAGE: %s [OPTIONS]\n\n", os.Args[0])
fmt.Fprintf(flag.CommandLine.Output(), "OPTIONS:\n")
myflag.PrintDefaults()
fmt.Fprintf(flag.CommandLine.Output(), notes)
}
if err := myflag.Parse(os.Args[1:]); err != nil {
os.Exit(1)
}
// Check the arguments
if object_count < 0 && duration_secs < 0 {
log.Fatal("The number of objects and duration can not both be unlimited")
}
if access_key == "" {
log.Fatal("Missing argument -a for access key.")
}
if secret_key == "" {
log.Fatal("Missing argument -s for secret key.")
}
if url_host == "" {
log.Fatal("Missing argument -u for host endpoint.")
}
invalid_mode := false
for _, r := range modes {
if r != 'i' &&
r != 'c' &&
r != 'p' &&
r != 'g' &&
r != 'l' &&
r != 'd' &&
r != 'x' {
s := fmt.Sprintf("Invalid mode '%s' passed to -m", string(r))
log.Printf(s)
invalid_mode = true
}
}
if invalid_mode {
log.Fatal("Invalid modes passed to -m, see help for details.")
}
var err error
var size uint64
if size, err = bytefmt.ToBytes(sizeArg); err != nil {
log.Fatalf("Invalid -z argument for object size: %v", err)
}
object_size = int64(size)
}
func initData() {
// Initialize data for the bucket
object_data = make([]byte, object_size)
rand.Read(object_data)
hasher := md5.New()
hasher.Write(object_data)
object_data_md5 = base64.StdEncoding.EncodeToString(hasher.Sum(nil))
}
func main() {
// Hello
log.Printf("Hotsauce S3 Benchmark Version 0.1")
cfg = &aws.Config{
Endpoint: aws.String(url_host),
Credentials: credentials.NewStaticCredentials(access_key, secret_key, ""),
Region: aws.String(region),
// DisableParamValidation: aws.Bool(true),
DisableComputeChecksums: aws.Bool(true),
S3ForcePathStyle: aws.Bool(true),
}
// Echo the parameters
log.Printf("Parameters:")
log.Printf("url=%s", url_host)
log.Printf("object_prefix=%s", object_prefix)
if bucket_list != "" {
log.Printf("bucket_list=%s", bucket_list)
} else {
log.Printf("bucket_prefix=%s", bucket_prefix)
}
log.Printf("region=%s", region)
log.Printf("modes=%s", modes)
log.Printf("output=%s", output)
log.Printf("json_output=%s", json_output)
log.Printf("max_keys=%d", max_keys)
log.Printf("object_count=%d", object_count)
log.Printf("first_object=%d", first_object)
log.Printf("bucket_count=%d", bucket_count)
log.Printf("duration=%d", duration_secs)
log.Printf("threads=%d", threads)
log.Printf("loops=%d", loops)
log.Printf("size=%s", sizeArg)
log.Printf("interval=%f", interval)
// Init Data
initData()
// Setup the slice of buckets
if bucket_list == "" {
for i := int64(0); i < bucket_count; i++ {
buckets = append(buckets, fmt.Sprintf("%s%012d", bucket_prefix, i))
}
} else {
buckets = strings.Split(bucket_list, " ")
}
// Loop running the tests
oStats := make([]OutputStats, 0)
for loop := 0; loop < loops; loop++ {
for _, r := range modes {
oStats = append(oStats, runWrapper(loop, r)...)
}
}
// Write CSV Output
if output != "" {
file, err := os.OpenFile(output, os.O_CREATE|os.O_WRONLY, 0777)
defer file.Close()
if err != nil {
log.Fatal("Could not open CSV file for writing.")
} else {
csvWriter := csv.NewWriter(file)
for i, o := range oStats {
if i == 0 {
o.csv_header(csvWriter)
}
o.csv(csvWriter)
}
csvWriter.Flush()
}
}
// Write JSON output
if json_output != "" {
file, err := os.OpenFile(json_output, os.O_CREATE|os.O_WRONLY, 0777)
defer file.Close()
if err != nil {
log.Fatal("Could not open JSON file for writing.")
}
data, err := json.Marshal(oStats)
if err != nil {
log.Fatal("Error marshaling JSON: ", err)
}
_, err = file.Write(data)
if err != nil {
log.Fatal("Error writing to JSON file: ", err)
}
file.Sync()
}
}