File: queue_manager_test.go

package info (click to toggle)
prometheus 2.7.1%2Bds-3
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 4,912 kB
  • sloc: lex: 171; sh: 155; makefile: 78
file content (330 lines) | stat: -rw-r--r-- 8,810 bytes parent folder | download
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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
// Copyright 2013 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

package remote

import (
	"context"
	"fmt"
	"reflect"
	"sync"
	"sync/atomic"
	"testing"
	"time"

	"github.com/prometheus/common/model"
	"github.com/prometheus/prometheus/config"
	"github.com/prometheus/prometheus/prompb"
)

const defaultFlushDeadline = 1 * time.Minute

type TestStorageClient struct {
	receivedSamples map[string][]prompb.Sample
	expectedSamples map[string][]prompb.Sample
	wg              sync.WaitGroup
	mtx             sync.Mutex
}

func NewTestStorageClient() *TestStorageClient {
	return &TestStorageClient{
		receivedSamples: map[string][]prompb.Sample{},
		expectedSamples: map[string][]prompb.Sample{},
	}
}

func (c *TestStorageClient) expectSamples(ss model.Samples) {
	c.mtx.Lock()
	defer c.mtx.Unlock()

	c.expectedSamples = map[string][]prompb.Sample{}
	c.receivedSamples = map[string][]prompb.Sample{}

	for _, s := range ss {
		ts := labelProtosToLabels(MetricToLabelProtos(s.Metric)).String()
		c.expectedSamples[ts] = append(c.expectedSamples[ts], prompb.Sample{
			Timestamp: int64(s.Timestamp),
			Value:     float64(s.Value),
		})
	}
	c.wg.Add(len(ss))
}

func (c *TestStorageClient) waitForExpectedSamples(t *testing.T) {
	c.wg.Wait()

	c.mtx.Lock()
	defer c.mtx.Unlock()
	for ts, expectedSamples := range c.expectedSamples {
		if !reflect.DeepEqual(expectedSamples, c.receivedSamples[ts]) {
			t.Fatalf("%s: Expected %v, got %v", ts, expectedSamples, c.receivedSamples[ts])
		}
	}
}

func (c *TestStorageClient) Store(_ context.Context, req *prompb.WriteRequest) error {
	c.mtx.Lock()
	defer c.mtx.Unlock()
	count := 0
	for _, ts := range req.Timeseries {
		labels := labelProtosToLabels(ts.Labels).String()
		for _, sample := range ts.Samples {
			count++
			c.receivedSamples[labels] = append(c.receivedSamples[labels], sample)
		}
	}
	c.wg.Add(-count)
	return nil
}

func (c *TestStorageClient) Name() string {
	return "teststorageclient"
}

func TestSampleDelivery(t *testing.T) {
	// Let's create an even number of send batches so we don't run into the
	// batch timeout case.
	n := config.DefaultQueueConfig.Capacity * 2

	samples := make(model.Samples, 0, n)
	for i := 0; i < n; i++ {
		name := model.LabelValue(fmt.Sprintf("test_metric_%d", i))
		samples = append(samples, &model.Sample{
			Metric: model.Metric{
				model.MetricNameLabel: name,
			},
			Value: model.SampleValue(i),
		})
	}

	c := NewTestStorageClient()
	c.expectSamples(samples[:len(samples)/2])

	cfg := config.DefaultQueueConfig
	cfg.MaxShards = 1
	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)

	// These should be received by the client.
	for _, s := range samples[:len(samples)/2] {
		m.Append(s)
	}
	// These will be dropped because the queue is full.
	for _, s := range samples[len(samples)/2:] {
		m.Append(s)
	}
	m.Start()
	defer m.Stop()

	c.waitForExpectedSamples(t)
}

func TestSampleDeliveryTimeout(t *testing.T) {
	// Let's send one less sample than batch size, and wait the timeout duration
	n := config.DefaultQueueConfig.Capacity - 1

	samples := make(model.Samples, 0, n)
	for i := 0; i < n; i++ {
		name := model.LabelValue(fmt.Sprintf("test_metric_%d", i))
		samples = append(samples, &model.Sample{
			Metric: model.Metric{
				model.MetricNameLabel: name,
			},
			Value: model.SampleValue(i),
		})
	}

	c := NewTestStorageClient()

	cfg := config.DefaultQueueConfig
	cfg.MaxShards = 1
	cfg.BatchSendDeadline = model.Duration(100 * time.Millisecond)
	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)
	m.Start()
	defer m.Stop()

	// Send the samples twice, waiting for the samples in the meantime.
	c.expectSamples(samples)
	for _, s := range samples {
		m.Append(s)
	}
	c.waitForExpectedSamples(t)

	c.expectSamples(samples)
	for _, s := range samples {
		m.Append(s)
	}
	c.waitForExpectedSamples(t)
}

func TestSampleDeliveryOrder(t *testing.T) {
	ts := 10
	n := config.DefaultQueueConfig.MaxSamplesPerSend * ts

	samples := make(model.Samples, 0, n)
	for i := 0; i < n; i++ {
		name := model.LabelValue(fmt.Sprintf("test_metric_%d", i%ts))
		samples = append(samples, &model.Sample{
			Metric: model.Metric{
				model.MetricNameLabel: name,
			},
			Value:     model.SampleValue(i),
			Timestamp: model.Time(i),
		})
	}

	c := NewTestStorageClient()
	c.expectSamples(samples)
	m := NewQueueManager(nil, config.DefaultQueueConfig, nil, nil, c, defaultFlushDeadline)

	// These should be received by the client.
	for _, s := range samples {
		m.Append(s)
	}
	m.Start()
	defer m.Stop()

	c.waitForExpectedSamples(t)
}

// TestBlockingStorageClient is a queue_manager StorageClient which will block
// on any calls to Store(), until the `block` channel is closed, at which point
// the `numCalls` property will contain a count of how many times Store() was
// called.
type TestBlockingStorageClient struct {
	numCalls uint64
	block    chan bool
}

func NewTestBlockedStorageClient() *TestBlockingStorageClient {
	return &TestBlockingStorageClient{
		block:    make(chan bool),
		numCalls: 0,
	}
}

func (c *TestBlockingStorageClient) Store(ctx context.Context, _ *prompb.WriteRequest) error {
	atomic.AddUint64(&c.numCalls, 1)
	select {
	case <-c.block:
	case <-ctx.Done():
	}
	return nil
}

func (c *TestBlockingStorageClient) NumCalls() uint64 {
	return atomic.LoadUint64(&c.numCalls)
}

func (c *TestBlockingStorageClient) unlock() {
	close(c.block)
}

func (c *TestBlockingStorageClient) Name() string {
	return "testblockingstorageclient"
}

func (t *QueueManager) queueLen() int {
	t.shardsMtx.Lock()
	defer t.shardsMtx.Unlock()
	queueLength := 0
	for _, shard := range t.shards.queues {
		queueLength += len(shard)
	}
	return queueLength
}

func TestSpawnNotMoreThanMaxConcurrentSendsGoroutines(t *testing.T) {
	// Our goal is to fully empty the queue:
	// `MaxSamplesPerSend*Shards` samples should be consumed by the
	// per-shard goroutines, and then another `MaxSamplesPerSend`
	// should be left on the queue.
	n := config.DefaultQueueConfig.MaxSamplesPerSend * 2

	samples := make(model.Samples, 0, n)
	for i := 0; i < n; i++ {
		name := model.LabelValue(fmt.Sprintf("test_metric_%d", i))
		samples = append(samples, &model.Sample{
			Metric: model.Metric{
				model.MetricNameLabel: name,
			},
			Value: model.SampleValue(i),
		})
	}

	c := NewTestBlockedStorageClient()
	cfg := config.DefaultQueueConfig
	cfg.MaxShards = 1
	cfg.Capacity = n
	m := NewQueueManager(nil, cfg, nil, nil, c, defaultFlushDeadline)

	m.Start()

	defer func() {
		c.unlock()
		m.Stop()
	}()

	for _, s := range samples {
		m.Append(s)
	}

	// Wait until the runShard() loops drain the queue.  If things went right, it
	// should then immediately block in sendSamples(), but, in case of error,
	// it would spawn too many goroutines, and thus we'd see more calls to
	// client.Store()
	//
	// The timed wait is maybe non-ideal, but, in order to verify that we're
	// not spawning too many concurrent goroutines, we have to wait on the
	// Run() loop to consume a specific number of elements from the
	// queue... and it doesn't signal that in any obvious way, except by
	// draining the queue.  We cap the waiting at 1 second -- that should give
	// plenty of time, and keeps the failure fairly quick if we're not draining
	// the queue properly.
	for i := 0; i < 100 && m.queueLen() > 0; i++ {
		time.Sleep(10 * time.Millisecond)
	}

	if m.queueLen() != config.DefaultQueueConfig.MaxSamplesPerSend {
		t.Fatalf("Failed to drain QueueManager queue, %d elements left",
			m.queueLen(),
		)
	}

	numCalls := c.NumCalls()
	if numCalls != uint64(1) {
		t.Errorf("Saw %d concurrent sends, expected 1", numCalls)
	}
}

func TestShutdown(t *testing.T) {
	deadline := 10 * time.Second
	c := NewTestBlockedStorageClient()
	m := NewQueueManager(nil, config.DefaultQueueConfig, nil, nil, c, deadline)
	for i := 0; i < config.DefaultQueueConfig.MaxSamplesPerSend; i++ {
		m.Append(&model.Sample{
			Metric: model.Metric{
				model.MetricNameLabel: model.LabelValue(fmt.Sprintf("test_metric_%d", i)),
			},
			Value:     model.SampleValue(i),
			Timestamp: model.Time(i),
		})
	}
	m.Start()

	start := time.Now()
	m.Stop()
	duration := time.Since(start)
	if duration > deadline+(deadline/10) {
		t.Errorf("Took too long to shutdown: %s > %s", duration, deadline)
	}
}