main.go 6.63 KB
Newer Older
liming6's avatar
liming6 committed
1
package backend
liming6's avatar
liming6 committed
2
3

import (
4
	"fmt"
5
	"get-container/docker"
liming6's avatar
liming6 committed
6
	"get-container/gpu"
7
8
	"get-container/utils"
	"maps"
9
	"math"
10
11
	"os"
	"strconv"
12
	"strings"
liming6's avatar
liming6 committed
13
	"sync"
14
15
16
	"time"

	"github.com/shirou/gopsutil/v4/process"
liming6's avatar
liming6 committed
17
18
)

liming6's avatar
liming6 committed
19
20
21
22
23
/*
	backend包为tui包提供数据
*/

var (
24
25
26
27
	MapIdDCU                        = sync.Map{} // 记录dcu信息
	DockerPidInfo *DockerProcessMap = nil
	User                            = ""
	HostName                        = ""
liming6's avatar
liming6 committed
28
29
)

30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
func init() {
	i, err := docker.ContainerInfo.GetProcessIdInDocker(false)
	if err != nil || i == nil {
		DockerPidInfo = &DockerProcessMap{lock: sync.RWMutex{}, pids: make(map[int32]bool)}
		return
	}
	DockerPidInfo = &DockerProcessMap{lock: sync.RWMutex{}, pids: maps.Clone(i)}

	HostName, _ = os.Hostname()
	uid := os.Getuid()
	u, err := utils.GetSysUserById(uid)
	if err == nil && u != nil {
		User = u.Name
	} else {
		User = strconv.Itoa(uid)
	}
}

liming6's avatar
liming6 committed
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
type DCUInfo struct {
	Id               int
	Name             string // full
	PerformanceLevel string // full
	Fan              string // full
	Temp             float32
	PwrAvg           float32
	PwrCap           float32 // full
	BusId            string  // full
	MemTotal         int
	MemUsed          int
	MemUsedPerent    float32
	Mig              bool // full
	DCUUTil          float32
	Ecc              bool // full
	PwrMode          string
}

// UpdateDCUInfo 更新dcu信息,full表示是否全部更新
func UpdateDCUInfo(full bool) {
	wg := sync.WaitGroup{}
	var smiAll map[int]*gpu.SMIAllOutput
	var eccInfo, migInfo map[int]bool
	var memInfo map[int]gpu.DCUMemInfo
	var runInfo map[int]gpu.DCURunningInfo
	var errSmiAll, errEcc, errMem, errRun error
	if full {
		wg.Add(5)
		go func() {
			smiAll, errSmiAll = gpu.GetSMIAllOutput()
			wg.Done()
		}()
		go func() {
			eccInfo, errEcc = gpu.GetEccInfo()
			wg.Done()
		}()
	} else {
		wg.Add(3)
	}
	go func() {
		migInfo = gpu.GetMigInfo()
		wg.Done()
	}()
	go func() {
		memInfo, errMem = gpu.GetDCUMemInfo()
		wg.Done()
	}()
	go func() {
		runInfo, errRun = gpu.GetRunningInfo()
		wg.Done()
	}()
	wg.Wait()
	cache := make(map[int]DCUInfo)
	if errSmiAll == nil && smiAll != nil {
		for k, v := range smiAll {
			i, have := cache[k]
			if !have {
				i = DCUInfo{}
				i.Id = k
			}
			i.Name = v.CardSeries
			i.PerformanceLevel = v.PerLevel
			i.Fan = gpu.NA
			i.PwrCap = v.MaxPwr
			i.PwrAvg = v.AvgPwr
			i.BusId = v.PCIBus
			i.MemUsedPerent = v.HCUMemUsage
			i.DCUUTil = v.HCUUsage
			cache[k] = i
		}
	}
	if errEcc == nil && eccInfo != nil {
		for k, v := range eccInfo {
			i, have := cache[k]
			if !have {
				i = DCUInfo{}
				i.Id = k
			}
			i.Ecc = v
			cache[k] = i
		}
	}
	for k, v := range migInfo {
		i, have := cache[k]
		if !have {
			i = DCUInfo{}
			i.Id = k
		}
		i.Mig = v
		cache[k] = i
	}
	if memInfo != nil && errMem == nil {
		for k, v := range memInfo {
			i, have := cache[k]
			if !have {
				i = DCUInfo{}
				i.Id = k
			}
			i.MemTotal = int(v.Total.Num)
			i.MemUsed = int(v.Used.Num)
			cache[k] = i
		}
	}
	if errRun == nil && runInfo != nil {
		for k, v := range runInfo {
			i, have := cache[k]
			if !have {
				i = DCUInfo{}
				i.Id = k
			}
			i.Temp = v.Temp
			i.PwrAvg = v.AvgPower
			i.PerformanceLevel = v.PerformanceLevel
			i.MemUsedPerent = v.MemPerc
			i.DCUUTil = v.HCUPerc
			i.PwrMode = v.Mode
			cache[k] = i
		}
	}
	for k, v := range cache {
		old, have := MapIdDCU.LoadOrStore(k, &v)
		if !have {
			continue
		}
		d := old.(*DCUInfo)
		if full {
			d.Name = v.Name
			d.PerformanceLevel = v.PerformanceLevel
			d.Fan = v.Fan
			d.PwrCap = v.PwrCap
			d.BusId = v.BusId
			d.Mig = v.Mig
			d.Ecc = v.Ecc
		}
		d.Temp = v.Temp
		d.PwrAvg = v.PwrAvg
		d.MemTotal = v.MemTotal
		d.MemUsed = v.MemUsed
		d.MemUsedPerent = v.MemUsedPerent
		d.DCUUTil = v.DCUUTil
		d.PwrMode = v.PwrMode
liming6's avatar
liming6 committed
189
190
	}
}
liming6's avatar
liming6 committed
191
192
193
194
195
196
197
198
199
200
201

func GetDCUInfo() map[int]DCUInfo {
	result := make(map[int]DCUInfo)
	MapIdDCU.Range(func(key, value any) bool {
		id := key.(int)
		val := value.(*DCUInfo)
		result[id] = *val
		return true
	})
	return result
}
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

type DockerProcessMap struct {
	lock sync.RWMutex
	pids map[int32]bool
}

func (dpm *DockerProcessMap) GetPidInfo() map[int32]bool {
	rl := dpm.lock.RLocker()
	rl.Lock()
	defer rl.Unlock()
	return maps.Clone(dpm.pids)
}

// Update 重新获取数据,这是一个耗时的操作
func (dpm *DockerProcessMap) Update() map[int32]bool {
	i, err := docker.ContainerInfo.GetProcessIdInDocker(true)
	if err != nil || i == nil {
		dpm.pids = make(map[int32]bool)
		return make(map[int32]bool)
	}
	dpm.pids = maps.Clone(i)
	return maps.Clone(i)
}

type DCUProcessInfo struct {
	DCU    int    // 使用的dcu号
	DCUMem string // 使用的dcu内存容量
	SDMA   int
	Info   ProcessInfo // 通用进程信息
}

type ProcessInfo struct {
	Pid      int32   // 进程号
	User     string  // 用户名或uid
	CPU      float64 // CPU使用率
	Mem      float32 // 内存使用率
	Time     string  // 占用的CPU时间
	Cmd      string  // 命令
	InDocker bool    // 是否在docker容器里
}

func getProcessInfo(pids []int32) map[int32]ProcessInfo {
	result := make(map[int32]ProcessInfo)
	if len(pids) == 0 {
		return result
	}
	dockerPids := DockerPidInfo.GetPidInfo()
	for _, pid := range pids {
		p, err := process.NewProcess(int32(pid))
		if err != nil {
			continue
		}
		item := ProcessInfo{Pid: p.Pid}
		item.User, _ = p.Username()
		item.CPU, _ = p.CPUPercent()
		item.Mem, _ = p.MemoryPercent()
		t, err := p.Times()
		if err == nil {
260
			item.Time = durationStr(time.Duration((t.System + t.User)) * time.Second)
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
		}
		item.Cmd, _ = p.Cmdline()
		a, b := dockerPids[item.Pid]
		item.InDocker = a && b
		result[p.Pid] = item
	}
	return result
}

// GetDCUProcessInfo 返回值的key为dcu index
func GetDCUProcessInfo() map[int][]DCUProcessInfo {
	result := make(map[int][]DCUProcessInfo)
	info, err := gpu.GetDCUPidInfo()
	if err != nil {
		return result
	}
	pids := make([]int32, 0)
	for _, v := range info {
		pids = append(pids, v.Pid)
	}
	pinfo := getProcessInfo(pids)
	for _, v := range info {
		index := make([]int, 0)
		for _, i := range v.HCUIndex {
			ii, err := strconv.Atoi(i)
			if err != nil {
				continue
			}
			index = append(index, ii)
		}
		for _, i := range index {
			l, have := result[i]
			if !have {
				result[i] = make([]DCUProcessInfo, 0)
				l = result[i]
			}
			item := DCUProcessInfo{DCU: i}
			item.Info = pinfo[v.Pid]
			item.DCUMem = v.VRamUsed.HumanReadStr(1)
			item.SDMA = v.SDMAUsed
			l = append(l, item)
			result[i] = l
		}
	}
	return result
}
307
308
309
310
311
312
313
314

// durationStr 将时间段格式化为 小时:分钟:秒s的格式
func durationStr(d time.Duration) string {
	h := int(math.Floor(d.Hours()))
	m := int(d.Minutes()) % 60
	s := int(math.Floor(d.Seconds())) % 60
	return strings.Replace(fmt.Sprintf("%d:%2d:%2d", h, m, s), " ", "0", -1)
}