FaceServer/report.py at master · zasper32171/FaceServer · GitHub

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
import threading
import heapq
import time
import os
import json
import requests
import psutil


class Reporter(threading.Thread):

    def __init__(self):

        threading.Thread.__init__(
            self, daemon=True
        )

        self._manager = None

        self._collect_period = 0.5
        self._report_period  = 5

        self._schedule    = []

        self._sys_info    = []
        self._proc_info   = []
        self._packet_info = []

        self._target_addr = None
        self._target_port = None

        self._terminating = threading.Event()

    def set_manager(self, manager):

        self._manager = manager

    def add_collector(self, addr, port):

        self._target_addr = addr
        self._target_port = port

    def set_collect_period(self, period):

        self._collect_period = period

    def set_report_period(self, period):

        self._report_period = period

    def collect_info(self):

        self.collect_sys_info()
        self.collect_proc_info()
        self.collect_packet_info()

        schedule_time = time.time() + self._collect_period
        heapq.heappush(self._schedule, (schedule_time, self.collect_info))

    def collect_sys_info(self):

        report = {}

        report['timestamp'] = time.time()

        report['cpu_util']  = psutil.cpu_percent()

        mem_info = psutil.virtual_memory()
        report['mem_total'] = mem_info[0] >> 20
        report['mem_util']  = mem_info[2]

        report['gpu_info']  = [
            {
                'gpu_id':        int(info[0]) if info[0].isdigit() else None,
                'gpu_power':     int(info[1]) if info[1].isdigit() else None,
                'gpu_util':      int(info[4]) if info[4].isdigit() else None,
                'gpu_mem_util':  int(info[5]) if info[5].isdigit() else None,
                'gpu_clock':     int(info[9]) if info[9].isdigit() else None
            } for info in [line.split() for line in os.popen(
                'nvidia-smi dmon -c 1'
            ).read().strip().split('\n')[2:] if not len(line) == 0]
        ]

        self._sys_info.append(report)

    def collect_proc_info(self):

        assert self._manager is not None

        gpu_info = [
            {
                'pid':          int(info[1]) if info[1].isdigit() else None,
                'gpu_id':       int(info[0]) if info[0].isdigit() else None,
                'gpu_util':     int(info[3]) if info[3].isdigit() else None,
                'gpu_mem_util': int(info[4]) if info[4].isdigit() else None
            } for info in [line.split() for line in os.popen(
                'nvidia-smi pmon -c 1'
            ).read().strip().split('\n')[2:] if not len(line) == 0]
        ]

        for s in self._manager.get_sessions():

            for h in s.get_handlers():

                report = {}

                report['timestamp'] = time.time()

                report['session_id'] = s.id
                report['handler_id'] = h.id

                p = psutil.Process(h.pid)

                # TODO: handle handlers with subprocesses
                assert not p.children()

                report['cpu_util'] = p.cpu_percent()
                report['mem_util'] = p.memory_percent()

                report['gpu_info'] = [
                    {
                        'gpu_id':       info['gpu_id'],
                        'gpu_util':     info['gpu_util'],
                        'gpu_mem_util': info['gpu_mem_util']
                    } for info in gpu_info if info['pid'] == h.pid
                ]

                self._proc_info.append(report)

    def collect_packet_info(self):

        for s in self._manager.get_sessions():
            sink = s.get_report_sink()

            while not sink.empty():
                report = sink.get()
                report['session_id'] = s.id

                self._packet_info.append(report)

    def send_report(self):

        assert self._target_addr is not None
        assert self._target_port is not None

        url = 'http://' + self._target_addr + ':' + str(self._target_port)

        payload = {'sys_info':    self._sys_info,
                   'proc_info':   self._proc_info,
                   'packet_info': self._packet_info}
        try:
            response = requests.post(
                url, data=json.dumps(payload), timeout=0.5
            ).json()

            if not response['result'] == 'Succeeded':

                print('Warning: report post failed (%s)' % response['reason'])

        except requests.exceptions.RequestException:

            print('Warning: timeout sending report. skipped.')

        else:
            self._sys_info, self._proc_info, self._packet_info = [], [], []

        schedule_time = time.time() + self._report_period
        heapq.heappush(self._schedule, (schedule_time, self.send_report))

    def run(self):

        schedule_time = time.time() + self._collect_period
        heapq.heappush(self._schedule, (schedule_time, self.collect_info))

        schedule_time = time.time() + self._report_period
        heapq.heappush(self._schedule, (schedule_time, self.send_report))

        while not self._terminating.is_set():

            schedule_time, task = self._schedule[0]

            if time.time() >= schedule_time:
                task()
                heapq.heappop(self._schedule)

            time.sleep(0.01)

    def stop(self):

        self._terminating.set()