-
Notifications
You must be signed in to change notification settings - Fork 8
/
autopprof.go
296 lines (257 loc) · 7.12 KB
/
autopprof.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
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
//go:build linux
// +build linux
package autopprof
import (
"bytes"
"context"
"fmt"
"log"
"time"
"github.com/daangn/autopprof/report"
)
const (
reportTimeout = 5 * time.Second
)
type autoPprof struct {
// watchInterval is the interval to watch the resource usages.
// Default: 5s.
watchInterval time.Duration
// cpuThreshold is the cpu usage threshold to trigger profile.
// If the cpu usage is over the threshold, the autopprof will
// report the cpu profile.
// Default: 0.75. (mean 75%)
cpuThreshold float64
// memThreshold is the memory usage threshold to trigger profile.
// If the memory usage is over the threshold, the autopprof will
// report the heap profile.
// Default: 0.75. (mean 75%)
memThreshold float64
// minConsecutiveOverThreshold is the minimum consecutive
// number of over a threshold for reporting profile again.
// Default: 12.
minConsecutiveOverThreshold int
// queryer is used to query the quota and the cgroup stat.
queryer queryer
// profiler is used to profile the cpu and the heap memory.
profiler profiler
// reporter is the reporter to send the profiling reports.
reporter report.Reporter
// reportBoth sets whether to trigger reports for both CPU and memory when either threshold is exceeded.
// If some profiling is disabled, exclude it.
reportBoth bool
// Flags to disable the profiling.
disableCPUProf bool
disableMemProf bool
// stopC is the signal channel to stop the watch processes.
stopC chan struct{}
}
// globalAp is the global autopprof instance.
var globalAp *autoPprof
// Start configures and runs the autopprof process.
func Start(opt Option) error {
qryer, err := newQueryer()
if err != nil {
return err
}
if err := opt.validate(); err != nil {
return err
}
profr := newDefaultProfiler(defaultCPUProfilingDuration)
ap := &autoPprof{
watchInterval: defaultWatchInterval,
cpuThreshold: defaultCPUThreshold,
memThreshold: defaultMemThreshold,
minConsecutiveOverThreshold: defaultMinConsecutiveOverThreshold,
queryer: qryer,
profiler: profr,
reporter: opt.Reporter,
reportBoth: opt.ReportBoth,
disableCPUProf: opt.DisableCPUProf,
disableMemProf: opt.DisableMemProf,
stopC: make(chan struct{}),
}
if opt.CPUThreshold != 0 {
ap.cpuThreshold = opt.CPUThreshold
}
if opt.MemThreshold != 0 {
ap.memThreshold = opt.MemThreshold
}
if !ap.disableCPUProf {
if err := ap.loadCPUQuota(); err != nil {
return err
}
}
go ap.watch()
globalAp = ap
return nil
}
// Stop stops the global autopprof process.
func Stop() {
if globalAp != nil {
globalAp.stop()
}
}
func (ap *autoPprof) loadCPUQuota() error {
err := ap.queryer.setCPUQuota()
if err == nil {
return nil
}
// If memory profiling is disabled and CPU quota isn't set,
// returns an error immediately.
if ap.disableMemProf {
return err
}
// If memory profiling is enabled, just logs the error and
// disables the cpu profiling.
log.Println(
"autopprof: disable the cpu profiling due to the CPU quota isn't set",
)
ap.disableCPUProf = true
return nil
}
func (ap *autoPprof) watch() {
go ap.watchCPUUsage()
go ap.watchMemUsage()
<-ap.stopC
}
func (ap *autoPprof) watchCPUUsage() {
if ap.disableCPUProf {
return
}
ticker := time.NewTicker(ap.watchInterval)
defer ticker.Stop()
var consecutiveOverThresholdCnt int
for {
select {
case <-ticker.C:
usage, err := ap.queryer.cpuUsage()
if err != nil {
log.Println(err)
return
}
if usage < ap.cpuThreshold {
// Reset the count if the cpu usage goes under the threshold.
consecutiveOverThresholdCnt = 0
continue
}
// If cpu utilization remains high for a short period of time, no
// duplicate reports are sent.
// This is to prevent the autopprof from sending too many reports.
if consecutiveOverThresholdCnt == 0 {
if err := ap.reportCPUProfile(usage); err != nil {
log.Println(fmt.Errorf(
"autopprof: failed to report the cpu profile: %w", err,
))
}
if ap.reportBoth && !ap.disableMemProf {
memUsage, err := ap.queryer.memUsage()
if err != nil {
log.Println(err)
return
}
if err := ap.reportHeapProfile(memUsage); err != nil {
log.Println(fmt.Errorf(
"autopprof: failed to report the heap profile: %w", err,
))
}
}
}
consecutiveOverThresholdCnt++
if consecutiveOverThresholdCnt >= ap.minConsecutiveOverThreshold {
// Reset the count and ready to report the cpu profile again.
consecutiveOverThresholdCnt = 0
}
case <-ap.stopC:
return
}
}
}
func (ap *autoPprof) reportCPUProfile(cpuUsage float64) error {
b, err := ap.profiler.profileCPU()
if err != nil {
return fmt.Errorf("autopprof: failed to profile the cpu: %w", err)
}
ctx, cancel := context.WithTimeout(context.Background(), reportTimeout)
defer cancel()
ci := report.CPUInfo{
ThresholdPercentage: ap.cpuThreshold * 100,
UsagePercentage: cpuUsage * 100,
}
bReader := bytes.NewReader(b)
if err := ap.reporter.ReportCPUProfile(ctx, bReader, ci); err != nil {
return err
}
return nil
}
func (ap *autoPprof) watchMemUsage() {
if ap.disableMemProf {
return
}
ticker := time.NewTicker(ap.watchInterval)
defer ticker.Stop()
var consecutiveOverThresholdCnt int
for {
select {
case <-ticker.C:
usage, err := ap.queryer.memUsage()
if err != nil {
log.Println(err)
return
}
if usage < ap.memThreshold {
// Reset the count if the memory usage goes under the threshold.
consecutiveOverThresholdCnt = 0
continue
}
// If memory utilization remains high for a short period of time,
// no duplicate reports are sent.
// This is to prevent the autopprof from sending too many reports.
if consecutiveOverThresholdCnt == 0 {
if err := ap.reportHeapProfile(usage); err != nil {
log.Println(fmt.Errorf(
"autopprof: failed to report the heap profile: %w", err,
))
}
if ap.reportBoth && !ap.disableCPUProf {
cpuUsage, err := ap.queryer.cpuUsage()
if err != nil {
log.Println(err)
return
}
if err := ap.reportCPUProfile(cpuUsage); err != nil {
log.Println(fmt.Errorf(
"autopprof: failed to report the cpu profile: %w", err,
))
}
}
}
consecutiveOverThresholdCnt++
if consecutiveOverThresholdCnt >= ap.minConsecutiveOverThreshold {
// Reset the count and ready to report the heap profile again.
consecutiveOverThresholdCnt = 0
}
case <-ap.stopC:
return
}
}
}
func (ap *autoPprof) reportHeapProfile(memUsage float64) error {
b, err := ap.profiler.profileHeap()
if err != nil {
return fmt.Errorf("autopprof: failed to profile the heap: %w", err)
}
ctx, cancel := context.WithTimeout(context.Background(), reportTimeout)
defer cancel()
mi := report.MemInfo{
ThresholdPercentage: ap.memThreshold * 100,
UsagePercentage: memUsage * 100,
}
bReader := bytes.NewReader(b)
if err := ap.reporter.ReportHeapProfile(ctx, bReader, mi); err != nil {
return err
}
return nil
}
func (ap *autoPprof) stop() {
close(ap.stopC)
}