-
Notifications
You must be signed in to change notification settings - Fork 672
/
usage.go
320 lines (264 loc) · 8.03 KB
/
usage.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
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
// Copyright (C) 2019-2024, Ava Labs, Inc. All rights reserved.
// See the file LICENSE for licensing terms.
package resource
import (
"math"
"strconv"
"sync"
"time"
"github.com/prometheus/client_golang/prometheus"
"github.com/shirou/gopsutil/cpu"
"github.com/shirou/gopsutil/process"
"go.uber.org/zap"
"github.com/ava-labs/avalanchego/utils/logging"
"github.com/ava-labs/avalanchego/utils/storage"
)
var (
lnHalf = math.Log(.5)
_ Manager = (*manager)(nil)
)
type CPUUser interface {
// CPUUsage returns the number of CPU cores of usage this user has attributed
// to it.
//
// For example, if this user is reporting a process's CPU utilization and
// that process is currently using 150% CPU (i.e. one and a half cores of
// compute) then the return value will be 1.5.
CPUUsage() float64
}
type DiskUser interface {
// DiskUsage returns the number of bytes per second read from/written to
// disk recently.
DiskUsage() (read float64, write float64)
// returns number of bytes available in the db volume
AvailableDiskBytes() uint64
}
type User interface {
CPUUser
DiskUser
}
type ProcessTracker interface {
// TrackProcess adds [pid] to the list of processes that this tracker is
// currently managing. Duplicate requests are dropped.
TrackProcess(pid int)
// UntrackProcess removes [pid] from the list of processes that this tracker
// is currently managing. Untracking a currently untracked [pid] is a noop.
UntrackProcess(pid int)
}
type Manager interface {
User
ProcessTracker
// Shutdown allocated resources and stop tracking all processes.
Shutdown()
}
type manager struct {
log logging.Logger
processMetrics *metrics
processesLock sync.Mutex
processes map[int]*proc
usageLock sync.RWMutex
cpuUsage float64
// [readUsage] is the number of bytes/second read from disk recently.
readUsage float64
// [writeUsage] is the number of bytes/second written to disk recently.
writeUsage float64
availableDiskBytes uint64
closeOnce sync.Once
onClose chan struct{}
}
func NewManager(
log logging.Logger,
diskPath string,
frequency,
cpuHalflife,
diskHalflife time.Duration,
metricsRegisterer prometheus.Registerer,
) (Manager, error) {
processMetrics, err := newMetrics(metricsRegisterer)
if err != nil {
return nil, err
}
m := &manager{
log: log,
processMetrics: processMetrics,
processes: make(map[int]*proc),
onClose: make(chan struct{}),
availableDiskBytes: math.MaxUint64,
}
go m.update(diskPath, frequency, cpuHalflife, diskHalflife)
return m, nil
}
func (m *manager) CPUUsage() float64 {
m.usageLock.RLock()
defer m.usageLock.RUnlock()
return m.cpuUsage
}
func (m *manager) DiskUsage() (float64, float64) {
m.usageLock.RLock()
defer m.usageLock.RUnlock()
return m.readUsage, m.writeUsage
}
func (m *manager) AvailableDiskBytes() uint64 {
m.usageLock.RLock()
defer m.usageLock.RUnlock()
return m.availableDiskBytes
}
func (m *manager) TrackProcess(pid int) {
p, err := process.NewProcess(int32(pid))
if err != nil {
return
}
process := &proc{
p: p,
log: m.log,
}
m.processesLock.Lock()
m.processes[pid] = process
m.processesLock.Unlock()
}
func (m *manager) UntrackProcess(pid int) {
m.processesLock.Lock()
delete(m.processes, pid)
m.processesLock.Unlock()
}
func (m *manager) Shutdown() {
m.closeOnce.Do(func() {
close(m.onClose)
})
}
func (m *manager) update(diskPath string, frequency, cpuHalflife, diskHalflife time.Duration) {
ticker := time.NewTicker(frequency)
defer ticker.Stop()
newCPUWeight, oldCPUWeight := getSampleWeights(frequency, cpuHalflife)
newDiskWeight, oldDiskWeight := getSampleWeights(frequency, diskHalflife)
frequencyInSeconds := frequency.Seconds()
for {
currentCPUUsage, currentReadUsage, currentWriteUsage := m.getActiveUsage(frequencyInSeconds)
currentScaledCPUUsage := newCPUWeight * currentCPUUsage
currentScaledReadUsage := newDiskWeight * currentReadUsage
currentScaledWriteUsage := newDiskWeight * currentWriteUsage
availableBytes, getBytesErr := storage.AvailableBytes(diskPath)
if getBytesErr != nil {
m.log.Verbo("failed to lookup resource",
zap.String("resource", "system disk"),
zap.String("path", diskPath),
zap.Error(getBytesErr),
)
}
m.usageLock.Lock()
m.cpuUsage = oldCPUWeight*m.cpuUsage + currentScaledCPUUsage
m.readUsage = oldDiskWeight*m.readUsage + currentScaledReadUsage
m.writeUsage = oldDiskWeight*m.writeUsage + currentScaledWriteUsage
if getBytesErr == nil {
m.availableDiskBytes = availableBytes
}
m.usageLock.Unlock()
select {
case <-ticker.C:
case <-m.onClose:
return
}
}
}
// Returns:
// 1. Current CPU usage by all processes.
// 2. Current bytes/sec read from disk by all processes.
// 3. Current bytes/sec written to disk by all processes.
func (m *manager) getActiveUsage(secondsSinceLastUpdate float64) (float64, float64, float64) {
m.processesLock.Lock()
defer m.processesLock.Unlock()
var (
totalCPU float64
totalRead float64
totalWrite float64
)
for _, p := range m.processes {
cpu, read, write := p.getActiveUsage(secondsSinceLastUpdate)
totalCPU += cpu
totalRead += read
totalWrite += write
processIDStr := strconv.Itoa(int(p.p.Pid))
m.processMetrics.numCPUCycles.WithLabelValues(processIDStr).Set(p.lastTotalCPU)
m.processMetrics.numDiskReads.WithLabelValues(processIDStr).Set(float64(p.numReads))
m.processMetrics.numDiskReadBytes.WithLabelValues(processIDStr).Set(float64(p.lastReadBytes))
m.processMetrics.numDiskWrites.WithLabelValues(processIDStr).Set(float64(p.numWrites))
m.processMetrics.numDiskWritesBytes.WithLabelValues(processIDStr).Set(float64(p.lastWriteBytes))
}
return totalCPU, totalRead, totalWrite
}
type proc struct {
p *process.Process
log logging.Logger
initialized bool
// [lastTotalCPU] is the most recent measurement of total CPU usage.
lastTotalCPU float64
// [numReads] is the total number of disk reads performed.
numReads uint64
// [lastReadBytes] is the most recent measurement of total disk bytes read.
lastReadBytes uint64
// [numWrites] is the total number of disk writes performed.
numWrites uint64
// [lastWriteBytes] is the most recent measurement of total disk bytes
// written.
lastWriteBytes uint64
}
func (p *proc) getActiveUsage(secondsSinceLastUpdate float64) (float64, float64, float64) {
// If there is an error tracking the CPU/disk utilization of a process,
// assume that the utilization is 0.
times, err := p.p.Times()
if err != nil {
p.log.Verbo("failed to lookup resource",
zap.String("resource", "process CPU"),
zap.Int32("pid", p.p.Pid),
zap.Error(err),
)
times = &cpu.TimesStat{}
}
// Note: IOCounters is not implemented on macos and therefore always returns
// an error on macos.
io, err := p.p.IOCounters()
if err != nil {
p.log.Verbo("failed to lookup resource",
zap.String("resource", "process IO"),
zap.Int32("pid", p.p.Pid),
zap.Error(err),
)
io = &process.IOCountersStat{}
}
var (
cpu float64
read float64
write float64
)
totalCPU := times.Total()
if p.initialized {
if totalCPU > p.lastTotalCPU {
newCPU := totalCPU - p.lastTotalCPU
cpu = newCPU / secondsSinceLastUpdate
}
if io.ReadBytes > p.lastReadBytes {
newRead := io.ReadBytes - p.lastReadBytes
read = float64(newRead) / secondsSinceLastUpdate
}
if io.WriteBytes > p.lastWriteBytes {
newWrite := io.WriteBytes - p.lastWriteBytes
write = float64(newWrite) / secondsSinceLastUpdate
}
}
p.initialized = true
p.lastTotalCPU = totalCPU
p.numReads = io.ReadCount
p.lastReadBytes = io.ReadBytes
p.numWrites = io.WriteCount
p.lastWriteBytes = io.WriteBytes
return cpu, read, write
}
// getSampleWeights converts the frequency of CPU sampling and the halflife of
// the CPU sample's usefulness into weights to scale the newly sampled point and
// previously samples.
func getSampleWeights(frequency, halflife time.Duration) (float64, float64) {
halflifeInSamples := float64(halflife) / float64(frequency)
oldWeight := math.Exp(lnHalf / halflifeInSamples)
newWeight := 1 - oldWeight
return newWeight, oldWeight
}