source: trunk/bin/sampler @ 11

Last change on this file since 11 was 9, checked in by tim, 9 years ago

.

  • Property svn:executable set to *
File size: 6.6 KB
Line 
1#!/usr/bin/env python3
2#-*- mode: Python;-*-
3
4import sys
5import os
6import time
7import random
8import tempfile
9import argparse
10import socket
11import json
12try:
13    import requests
14except:
15    sys.stderr.write('ERROR: Could not import requests module.  Ensure it is installed.\n')
16    sys.stderr.write('       Under Debian, the package name is "python3-requests"\n.')
17    sys.exit(1)
18
19VERSION = "{DEVELOPMENT}"
20if VERSION == "{DEVELOPMENT}":
21    script_dir = '.'
22    try:
23        script_dir = os.path.dirname(os.path.realpath(__file__))
24    except:
25        try:
26            script_dir = os.path.dirname(os.path.abspath(sys.argv[0]))
27        except:
28            pass
29    sys.path.append("%s/../lib" % script_dir)
30
31from nanownlib import *
32import nanownlib.storage
33
34parser = argparse.ArgumentParser(
35    description="")
36parser.add_argument('-c', dest='cases', type=str, default='{"short":10000,"long":13000}',
37                    help='JSON representation of echo timing cases.')
38parser.add_argument('--no-tcpts', action='store_true', help='Disbale TCP timestamp profiling')
39parser.add_argument('--no-control', action='store_true', help='Do not collect separate control data.  Instead, synthesize it from test and train data.')
40parser.add_argument('session_name', default=None,
41                    help='Name for the sampler session (used in output filenames)')
42parser.add_argument('sample_count', type=int, default=None,
43                    help='Number of samples to collect')
44parser.add_argument('host', default=None,
45                    help='IP address or host name of server')
46parser.add_argument('port', nargs='?', type=int, default=80,
47                    help='TCP port number of HTTP service (default: 80)')
48options = parser.parse_args()
49
50
51session_name = options.session_name
52num_samples=options.sample_count
53hostname = options.host
54port = options.port
55protocol = 'http'
56
57cases = json.loads(options.cases)
58
59def extractReportedRuntime(headers, body):
60    if body.startswith("waited: "):
61        return int(body.split("waited: ")[1], 10)
62
63
64def sendRequest(data=None):
65    method = 'GET'
66    path = '/data/clamav-audit-results.txt?t=' + data
67    url = "%s://%s:%d%s" % (protocol,hostname,port,path)
68    headers = {}
69    body = (b'')
70    req = requests.Request('GET', url).prepare()
71
72    retry = True
73    while retry:
74        try:
75            session = requests.Session()
76            response = session.send(req)
77            #print(repr(response.raw._original_response.local_address))
78            reported = extractReportedRuntime(response.headers, response.text)
79            retry = False
80        except Exception as e:
81            sys.stderr.write("ERROR: HTTP request problem: %s\n" % repr(e))
82            time.sleep(1.0)
83            sys.stderr.write("ERROR: retrying...\n")
84   
85    return {'userspace_rtt':response.elapsed.microseconds*1000,
86            'reported':reported,
87            'local_port':response.raw._original_response.local_address[1]}
88
89
90def fetch(probedata, data):
91    #   http://docs.python-requests.org/en/latest/api/#requests.Response
92    result = sendRequest("%d" % data)
93    result.update(probedata)
94   
95    return result
96
97
98setCPUAffinity()
99setTCPTimestamps()
100host_ip = socket.gethostbyname(hostname) #XXX: what about multiple A records?
101
102meta = {'tcpts_mean':None,'tcpts_stddev':None,'tcpts_slopes':None}
103if not options.no_tcpts:
104    sys.stderr.write("INFO: Probing target for TCP timestamp precision...\n")
105    sniffer_fp = tempfile.NamedTemporaryFile('w+t')
106    sniffer = startSniffer(host_ip, port, sniffer_fp.name)
107    time.sleep(1.0)
108    ports = runTimestampProbes(host_ip, port, hostname, 12)
109    time.sleep(1.0)
110    stopSniffer(sniffer)
111    sniffer_fp.seek(0)
112    mean,stddev,slopes = computeTimestampPrecision(sniffer_fp, ports)
113    meta = {'tcpts_mean':mean,'tcpts_stddev':stddev,'tcpts_slopes':json.dumps(slopes)}
114   
115if meta['tcpts_mean'] == None:
116    sys.stderr.write("INFO: TCP timestamps not supported.\n")
117elif meta['tcpts_stddev'] == None:
118    sys.stderr.write("INFO: Not enough TCP timestamp samples.\n")
119elif meta['tcpts_stddev']/meta['tcpts_mean'] > 0.05:
120    sys.stderr.write("INFO: TCP timestamp precision is inconsistent.\n")
121    sys.stderr.write("INFO:     mean=%(tcpts_mean)f,stddev=%(tcpts_stddev)f,samples=%(tcpts_slopes)s\n"
122                     % meta)
123else:
124    sys.stderr.write("INFO: Estimated TCP timestamp precision: %f (stddev: %f, %f%%)\n"
125                     % (meta['tcpts_mean'], meta['tcpts_stddev'], 100*meta['tcpts_stddev']/meta['tcpts_mean']))
126
127
128sniffer_fp = tempfile.NamedTemporaryFile('w+t')
129db_file = "%s.db" % session_name
130
131sniffer = startSniffer(host_ip, port, sniffer_fp.name)
132db = nanownlib.storage.db(db_file)
133db.addMeta(meta)
134time.sleep(0.5) # ensure sniffer is fully ready and our process is migrated
135
136if options.no_control:
137    num_control = 0
138else:
139    num_control = int(num_samples*2/5)
140
141num_train = int((num_samples-num_control)/3)
142num_test = num_samples-num_train-num_control
143
144sample_types = [('train',num_train),
145                ('train_null',num_control),
146                ('test',num_test)]
147
148sid = 0
149report_interval = 20
150start = time.time()
151next_report = start+report_interval
152for st,count in sample_types:
153    if sniffer.poll() != None:
154        sys.stderr.write('ERROR: Sniffer quit, sender exiting...\n')
155        break
156       
157    for k in range(0,count):
158        sample_order = list(cases.items())
159        random.shuffle(sample_order)
160        if st.endswith('null'):
161            for i in range(1,len(sample_order)):
162                sample_order[i] = (sample_order[i][0],sample_order[0][1])
163            random.shuffle(sample_order)
164            #print('after', sample_order)
165           
166        results = []
167        now = int(time.time()*1000000000)
168        for i in range(len(sample_order)):
169            results.append(fetch({'sample':sid, 'test_case':sample_order[i][0],
170                                  'type':st, 'tc_order':i, 'time_of_day':now},
171                                 sample_order[i][1]))
172
173        #print(results)
174        db.addProbes(results)
175        db.conn.commit()
176        sid += 1
177
178        if (time.time() > next_report):
179            #s = time.time()
180            reportProgress(db, sample_types, start)
181            #print("reportProgress time:", time.time()-s)
182            next_report += report_interval
183
184print("probes complete in %f" % (time.time()-start))
185time.sleep(2.0) # Give sniffer a chance to collect remaining packets
186stopSniffer(sniffer)
187
188start = time.time()
189associatePackets(sniffer_fp, db)
190sniffer_fp.close()
191end = time.time()
192print("associate time:", end-start)
193
194if options.no_control:
195    print("TODO: implement control synthesizing!")
196
197start = time.time()
198num_probes = analyzeProbes(db)
199end = time.time()
200print("analyzed %d probes' packets in: %f" % (num_probes, end-start))
Note: See TracBrowser for help on using the repository browser.