forked from grafana/pyroscope-go
-
Notifications
You must be signed in to change notification settings - Fork 0
/
collector.go
238 lines (210 loc) · 5.25 KB
/
collector.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
package pyroscope
import (
"bytes"
"fmt"
"io"
"time"
internal "github.com/grafana/pyroscope-go/internal/pprof"
"github.com/grafana/pyroscope-go/upstream"
)
type cpuProfileCollector struct {
name string
dur time.Duration
upstream upstream.Upstream
collector internal.Collector
logger Logger
buf *bytes.Buffer
timeStarted time.Time
// started indicates whether the collector
// is interrupted with StartCPUProfile.
started bool
events chan event
halt chan struct{}
done chan struct{}
}
type event struct {
typ eventType
done chan error
w io.Writer
}
type eventType int
const (
startEvent eventType = iota
stopEvent
flushEvent
)
func newEvent(typ eventType) event {
return event{typ: typ, done: make(chan error, 1)}
}
func (e event) send(c chan<- event) error {
c <- e
return <-e.done
}
func newStartEvent(w io.Writer) event {
e := newEvent(startEvent)
e.w = w
return e
}
func newCPUProfileCollector(
name string,
upstream upstream.Upstream,
logger Logger,
period time.Duration,
) *cpuProfileCollector {
buf := bytes.NewBuffer(make([]byte, 0, 1<<10))
return &cpuProfileCollector{
name: name,
dur: period,
upstream: upstream,
logger: logger,
collector: internal.DefaultCollector(),
buf: buf,
events: make(chan event),
halt: make(chan struct{}),
done: make(chan struct{}),
}
}
func (c *cpuProfileCollector) Start() {
c.logger.Debugf("starting cpu profile collector")
// From now on, internal pprof.StartCPUProfile
// is handled by this collector.
internal.SetCollector(c)
t := time.NewTicker(c.dur)
// Force pprof.StartCPUProfile: if CPU profiling is already
// in progress (pprof.StartCPUProfile called outside the
// package), profiling will start once it finishes.
_ = c.reset(nil)
for {
select {
case n := <-t.C:
// Skip and adjust the timer, if the actual
// profile duration is less than the desired,
// which may happen if the collector has been
// interrupted and then resumed, or flushed.
if d := n.Sub(c.timeStarted); d < c.dur {
if d < 0 {
// Ticker fired after the StartCPUProfile
// call, that interrupted background
// profiling.
d = c.dur
}
t.Reset(d)
continue
}
t.Reset(c.dur)
if !c.started {
// Collector can't start collecting profiles
// in background while profiling started with
// StartCPUProfile (foreground).
_ = c.reset(nil)
}
case <-c.halt:
t.Stop()
if c.started {
// Collector can't be stopped in-between
// StartCPUProfile and StopCPUProfile calls.
continue
}
c.collector.StopCPUProfile()
c.upload()
close(c.done)
return
case e := <-c.events:
c.handleEvent(e)
}
}
}
func (c *cpuProfileCollector) handleEvent(e event) {
var err error
defer func() {
e.done <- err
close(e.done)
}()
switch e.typ {
case startEvent:
if c.started { // Misuse.
// Just to avoid interruption of the background
// profiling that will fail immediately.
err = fmt.Errorf("cpu profiling already started")
} else {
err = c.reset(e.w)
c.started = err == nil
}
case stopEvent:
if c.started {
err = c.reset(nil)
c.started = false
}
case flushEvent:
if c.started {
// Flush can't be done if StartCPUProfile is called,
// as we'd need stopping the foreground collector first.
err = fmt.Errorf("flush rejected: cpu profiling is in progress")
} else {
err = c.reset(nil)
}
}
}
func (c *cpuProfileCollector) Stop() {
c.logger.Debugf("stopping cpu profile collector")
// Switches back to the standard pprof collector.
// If internal pprof.StartCPUProfile is called,
// the function blocks until StopCPUProfile.
internal.SetCollector(c.collector)
// Note that "halt" is not an event, but rather state
// of the collector: the channel can be read multiple
// times before the collector stops.
close(c.halt)
<-c.done
c.logger.Debugf("stopping cpu profile collector stopped")
}
func (c *cpuProfileCollector) StartCPUProfile(w io.Writer) error {
c.logger.Debugf("cpu profile collector interrupted with StartCPUProfile")
return newStartEvent(w).send(c.events)
}
func (c *cpuProfileCollector) StopCPUProfile() {
c.logger.Debugf("cpu profile collector restored")
_ = newEvent(stopEvent).send(c.events)
}
func (c *cpuProfileCollector) Flush() error {
return newEvent(flushEvent).send(c.events)
}
func (c *cpuProfileCollector) reset(w io.Writer) error {
c.collector.StopCPUProfile()
c.upload()
var d io.Writer = c.buf
if w != nil {
// pprof.StopCPUProfile dumps gzipped
// profile ignoring any writer failure.
d = io.MultiWriter(d, w)
}
c.timeStarted = time.Now()
if err := c.collector.StartCPUProfile(d); err != nil {
c.logger.Errorf("failed to start CPU profiling: %v", err)
c.timeStarted = time.Time{}
c.buf.Reset()
return err
}
return nil
}
func (c *cpuProfileCollector) upload() {
if c.timeStarted.IsZero() {
return
}
buf := c.buf.Bytes()
if len(buf) == 0 {
return
}
c.upstream.Upload(&upstream.UploadJob{
Name: c.name,
StartTime: c.timeStarted,
EndTime: time.Now(),
SpyName: "gospy",
SampleRate: DefaultSampleRate,
Units: "samples",
AggregationType: "sum",
Format: upstream.FormatPprof,
Profile: copyBuf(buf),
})
c.buf.Reset()
}