forked from grafana/cortex-tools
/
loadgen.go
248 lines (216 loc) · 7.15 KB
/
loadgen.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
package commands
import (
"context"
"fmt"
"log"
"math/rand"
"net/http"
"net/url"
"sync"
"time"
"github.com/gogo/protobuf/proto"
"github.com/golang/snappy"
"github.com/pkg/errors"
"github.com/prometheus/client_golang/api"
v1 "github.com/prometheus/client_golang/api/prometheus/v1"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promauto"
"github.com/prometheus/client_golang/prometheus/promhttp"
"github.com/prometheus/common/config"
"github.com/prometheus/common/model"
"github.com/prometheus/prometheus/prompb"
"github.com/prometheus/prometheus/storage/remote"
"github.com/sirupsen/logrus"
"gopkg.in/alecthomas/kingpin.v2"
)
var (
// 15 seconds is a common send interval. To provide useful metrics at high latencies we will
// add 15 to the default Prometheus buckets
defBuckets = append(prometheus.DefBuckets, 15)
)
var writeRequestDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
Namespace: "loadgen",
Name: "write_request_duration_seconds",
Buckets: defBuckets,
}, []string{"success"})
var queryRequestDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
Namespace: "loadgen",
Name: "query_request_duration_seconds",
Buckets: defBuckets,
}, []string{"success"})
type LoadgenCommand struct {
writeURL string
activeSeries int
scrapeInterval time.Duration
parallelism int
batchSize int
writeTimeout time.Duration
metricName string
queryURL string
query string
queryParallelism int
queryTimeout time.Duration
queryDuration time.Duration
metricsListenAddress string
// Runtime stuff.
wg sync.WaitGroup
writeClient remote.WriteClient
queryClient v1.API
}
func (c *LoadgenCommand) Register(app *kingpin.Application) {
loadgenCommand := &LoadgenCommand{}
cmd := app.Command("loadgen", "Simple load generator for Cortex.").Action(loadgenCommand.run)
cmd.Flag("write-url", "").
Default("").StringVar(&loadgenCommand.writeURL)
cmd.Flag("series-name", "name of the metric that will be generated").
Default("node_cpu_seconds_total").StringVar(&loadgenCommand.metricName)
cmd.Flag("active-series", "number of active series to send").
Default("1000").IntVar(&loadgenCommand.activeSeries)
cmd.Flag("scrape-interval", "period to send metrics").
Default("15s").DurationVar(&loadgenCommand.scrapeInterval)
cmd.Flag("parallelism", "how many metrics to send simultaneously").
Default("10").IntVar(&loadgenCommand.parallelism)
cmd.Flag("batch-size", "how big a batch to send").
Default("100").IntVar(&loadgenCommand.batchSize)
cmd.Flag("write-timeout", "timeout for write requests").
Default("500ms").DurationVar(&loadgenCommand.writeTimeout)
cmd.Flag("query-url", "").
Default("").StringVar(&loadgenCommand.queryURL)
cmd.Flag("query", "query to run").
Default("sum(node_cpu_seconds_total)").StringVar(&loadgenCommand.query)
cmd.Flag("query-parallelism", "number of queries to run in parallel").
Default("10").IntVar(&loadgenCommand.queryParallelism)
cmd.Flag("query-timeout", "").
Default("20s").DurationVar(&loadgenCommand.queryTimeout)
cmd.Flag("query-duration", "length of query").
Default("1h").DurationVar(&loadgenCommand.queryDuration)
cmd.Flag("metrics-listen-address", "address to serve metrics on").
Default(":8080").StringVar(&loadgenCommand.metricsListenAddress)
}
func (c *LoadgenCommand) run(k *kingpin.ParseContext) error {
if c.writeURL == "" && c.queryURL == "" {
return errors.New("either a -write-url or -query-url flag must be provided to run the loadgen command")
}
http.Handle("/metrics", promhttp.Handler())
go func() {
err := http.ListenAndServe(c.metricsListenAddress, nil)
if err != nil {
logrus.WithError(err).Errorln("metrics listener failed")
}
}()
if c.writeURL != "" {
log.Printf("setting up write load gen:\n url=%s\n parallelism: %v\n active_series: %d\n interval: %v\n", c.writeURL, c.parallelism, c.activeSeries, c.scrapeInterval)
writeURL, err := url.Parse(c.writeURL)
if err != nil {
return err
}
writeClient, err := remote.NewWriteClient("loadgen", &remote.ClientConfig{
URL: &config.URL{URL: writeURL},
Timeout: model.Duration(c.writeTimeout),
})
if err != nil {
return err
}
c.writeClient = writeClient
c.wg.Add(c.parallelism)
metricsPerShard := c.activeSeries / c.parallelism
for i := 0; i < c.activeSeries; i += metricsPerShard {
go c.runWriteShard(i, i+metricsPerShard)
}
} else {
log.Println("write load generation is disabled, -write-url flag has not been set")
}
if c.queryURL != "" {
log.Printf("setting up query load gen:\n url=%s\n parallelism: %v\n query: %s", c.queryURL, c.queryParallelism, c.query)
queryClient, err := api.NewClient(api.Config{
Address: c.queryURL,
})
if err != nil {
return err
}
c.queryClient = v1.NewAPI(queryClient)
c.wg.Add(c.queryParallelism)
for i := 0; i < c.queryParallelism; i++ {
go c.runQueryShard()
}
} else {
log.Println("query load generation is disabled, -query-url flag has not been set")
}
c.wg.Wait()
return nil
}
func (c *LoadgenCommand) runWriteShard(from, to int) {
defer c.wg.Done()
ticker := time.NewTicker(c.scrapeInterval)
c.runScrape(from, to)
for range ticker.C {
c.runScrape(from, to)
}
}
func (c *LoadgenCommand) runScrape(from, to int) {
for i := from; i < to; i += c.batchSize {
if err := c.runBatch(i, i+c.batchSize); err != nil {
log.Printf("error sending batch: %v", err)
}
}
fmt.Printf("sent %d samples\n", to-from)
}
func (c *LoadgenCommand) runBatch(from, to int) error {
var (
req = prompb.WriteRequest{
Timeseries: make([]prompb.TimeSeries, 0, to-from),
}
now = time.Now().UnixNano() / int64(time.Millisecond)
)
for i := from; i < to; i++ {
timeseries := prompb.TimeSeries{
Labels: []prompb.Label{
{Name: "__name__", Value: c.metricName},
{Name: "job", Value: "node_exporter"},
{Name: "instance", Value: fmt.Sprintf("instance%000d", i)},
{Name: "cpu", Value: "0"},
{Name: "mode", Value: "idle"},
},
Samples: []prompb.Sample{{
Timestamp: now,
Value: rand.Float64(),
}},
}
req.Timeseries = append(req.Timeseries, timeseries)
}
data, err := proto.Marshal(&req)
if err != nil {
return err
}
compressed := snappy.Encode(nil, data)
start := time.Now()
if err := c.writeClient.Store(context.Background(), compressed); err != nil {
writeRequestDuration.WithLabelValues("error").Observe(time.Since(start).Seconds())
return err
}
writeRequestDuration.WithLabelValues("success").Observe(time.Since(start).Seconds())
return nil
}
func (c *LoadgenCommand) runQueryShard() {
defer c.wg.Done()
for {
c.runQuery()
}
}
func (c *LoadgenCommand) runQuery() {
ctx, cancel := context.WithTimeout(context.Background(), c.queryTimeout)
defer cancel()
r := v1.Range{
Start: time.Now().Add(-c.queryDuration),
End: time.Now(),
Step: time.Minute,
}
start := time.Now()
_, _, err := c.queryClient.QueryRange(ctx, c.query, r)
if err != nil {
queryRequestDuration.WithLabelValues("error").Observe(time.Since(start).Seconds())
log.Printf("error doing query: %v", err)
return
}
queryRequestDuration.WithLabelValues("success").Observe(time.Since(start).Seconds())
}