aboutsummaryrefslogtreecommitdiffhomepage
path: root/tools/run_tests/run_performance_tests.py
blob: 92149115fe2387ac6e2d79cef5a54d4c67ec1339 (plain)
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
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
#!/usr/bin/env python2.7
# Copyright 2016, Google Inc.
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are
# met:
#
#     * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
#     * Redistributions in binary form must reproduce the above
# copyright notice, this list of conditions and the following disclaimer
# in the documentation and/or other materials provided with the
# distribution.
#     * Neither the name of Google Inc. nor the names of its
# contributors may be used to endorse or promote products derived from
# this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

"""Run performance tests locally or remotely."""

from __future__ import print_function

import argparse
import collections
import itertools
import jobset
import json
import multiprocessing
import os
import performance.scenario_config as scenario_config
import pipes
import re
import subprocess
import sys
import tempfile
import time
import traceback
import uuid


_ROOT = os.path.abspath(os.path.join(os.path.dirname(sys.argv[0]), '../..'))
os.chdir(_ROOT)


_REMOTE_HOST_USERNAME = 'jenkins'


class QpsWorkerJob:
  """Encapsulates a qps worker server job."""

  def __init__(self, spec, language, host_and_port):
    self._spec = spec
    self.language = language
    self.host_and_port = host_and_port
    self._job = None

  def start(self):
    self._job = jobset.Job(self._spec, newline_on_success=True, travis=True, add_env={})

  def is_running(self):
    """Polls a job and returns True if given job is still running."""
    return self._job and self._job.state() == jobset._RUNNING

  def kill(self):
    if self._job:
      self._job.kill()
      self._job = None


def create_qpsworker_job(language, shortname=None,
                         port=10000, remote_host=None):
  cmdline = language.worker_cmdline() + ['--driver_port=%s' % port]
  if remote_host:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, remote_host)
    cmdline = ['ssh',
               str(user_at_host),
               'cd ~/performance_workspace/grpc/ && %s' % ' '.join(cmdline)]
    host_and_port='%s:%s' % (remote_host, port)
  else:
    host_and_port='localhost:%s' % port

  # TODO(jtattermusch): with some care, we can calculate the right timeout
  # of a worker from the sum of warmup + benchmark times for all the scenarios
  jobspec = jobset.JobSpec(
      cmdline=cmdline,
      shortname=shortname,
      timeout_seconds=2*60*60)
  return QpsWorkerJob(jobspec, language, host_and_port)


def create_scenario_jobspec(scenario_json, workers, remote_host=None,
                            bq_result_table=None):
  """Runs one scenario using QPS driver."""
  # setting QPS_WORKERS env variable here makes sure it works with SSH too.
  cmd = 'QPS_WORKERS="%s" ' % ','.join(workers)
  if bq_result_table:
    cmd += 'BQ_RESULT_TABLE="%s" ' % bq_result_table
  cmd += 'tools/run_tests/performance/run_qps_driver.sh '
  cmd += '--scenarios_json=%s ' % pipes.quote(json.dumps({'scenarios': [scenario_json]}))
  cmd += '--scenario_result_file=scenario_result.json'
  if remote_host:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, remote_host)
    cmd = 'ssh %s "cd ~/performance_workspace/grpc/ && "%s' % (user_at_host, pipes.quote(cmd))

  return jobset.JobSpec(
      cmdline=[cmd],
      shortname='qps_json_driver.%s' % scenario_json['name'],
      timeout_seconds=3*60,
      shell=True,
      verbose_success=True)


def create_quit_jobspec(workers, remote_host=None):
  """Runs quit using QPS driver."""
  # setting QPS_WORKERS env variable here makes sure it works with SSH too.
  cmd = 'QPS_WORKERS="%s" bins/opt/qps_json_driver --quit' % ','.join(w.host_and_port for w in workers)
  if remote_host:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, remote_host)
    cmd = 'ssh %s "cd ~/performance_workspace/grpc/ && "%s' % (user_at_host, pipes.quote(cmd))

  return jobset.JobSpec(
      cmdline=[cmd],
      shortname='qps_json_driver.quit',
      timeout_seconds=3*60,
      shell=True,
      verbose_success=True)


def create_netperf_jobspec(server_host='localhost', client_host=None,
                           bq_result_table=None):
  """Runs netperf benchmark."""
  cmd = 'NETPERF_SERVER_HOST="%s" ' % server_host
  if bq_result_table:
    cmd += 'BQ_RESULT_TABLE="%s" ' % bq_result_table
  if client_host:
    # If netperf is running remotely, the env variables populated by Jenkins
    # won't be available on the client, but we need them for uploading results
    # to BigQuery.
    jenkins_job_name = os.getenv('JOB_NAME')
    if jenkins_job_name:
      cmd += 'JOB_NAME="%s" ' % jenkins_job_name
    jenkins_build_number = os.getenv('BUILD_NUMBER')
    if jenkins_build_number:
      cmd += 'BUILD_NUMBER="%s" ' % jenkins_build_number

  cmd += 'tools/run_tests/performance/run_netperf.sh'
  if client_host:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, client_host)
    cmd = 'ssh %s "cd ~/performance_workspace/grpc/ && "%s' % (user_at_host, pipes.quote(cmd))

  return jobset.JobSpec(
      cmdline=[cmd],
      shortname='netperf',
      timeout_seconds=60,
      shell=True,
      verbose_success=True)


def archive_repo(languages):
  """Archives local version of repo including submodules."""
  cmdline=['tar', '-cf', '../grpc.tar', '../grpc/']
  if 'java' in languages:
    cmdline.append('../grpc-java')
  if 'go' in languages:
    cmdline.append('../grpc-go')

  archive_job = jobset.JobSpec(
      cmdline=cmdline,
      shortname='archive_repo',
      timeout_seconds=3*60)

  jobset.message('START', 'Archiving local repository.', do_newline=True)
  num_failures, _ = jobset.run(
      [archive_job], newline_on_success=True, maxjobs=1)
  if num_failures == 0:
    jobset.message('SUCCESS',
                   'Archive with local repository created successfully.',
                   do_newline=True)
  else:
    jobset.message('FAILED', 'Failed to archive local repository.',
                   do_newline=True)
    sys.exit(1)


def prepare_remote_hosts(hosts, prepare_local=False):
  """Prepares remote hosts (and maybe prepare localhost as well)."""
  prepare_timeout = 5*60
  prepare_jobs = []
  for host in hosts:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, host)
    prepare_jobs.append(
        jobset.JobSpec(
            cmdline=['tools/run_tests/performance/remote_host_prepare.sh'],
            shortname='remote_host_prepare.%s' % host,
            environ = {'USER_AT_HOST': user_at_host},
            timeout_seconds=prepare_timeout))
  if prepare_local:
    # Prepare localhost as well
    prepare_jobs.append(
        jobset.JobSpec(
            cmdline=['tools/run_tests/performance/kill_workers.sh'],
            shortname='local_prepare',
            timeout_seconds=prepare_timeout))
  jobset.message('START', 'Preparing hosts.', do_newline=True)
  num_failures, _ = jobset.run(
      prepare_jobs, newline_on_success=True, maxjobs=10)
  if num_failures == 0:
    jobset.message('SUCCESS',
                   'Prepare step completed successfully.',
                   do_newline=True)
  else:
    jobset.message('FAILED', 'Failed to prepare remote hosts.',
                   do_newline=True)
    sys.exit(1)


def build_on_remote_hosts(hosts, languages=scenario_config.LANGUAGES.keys(), build_local=False):
  """Builds performance worker on remote hosts (and maybe also locally)."""
  build_timeout = 15*60
  build_jobs = []
  for host in hosts:
    user_at_host = '%s@%s' % (_REMOTE_HOST_USERNAME, host)
    build_jobs.append(
        jobset.JobSpec(
            cmdline=['tools/run_tests/performance/remote_host_build.sh'] + languages,
            shortname='remote_host_build.%s' % host,
            environ = {'USER_AT_HOST': user_at_host, 'CONFIG': 'opt'},
            timeout_seconds=build_timeout))
  if build_local:
    # Build locally as well
    build_jobs.append(
        jobset.JobSpec(
            cmdline=['tools/run_tests/performance/build_performance.sh'] + languages,
            shortname='local_build',
            environ = {'CONFIG': 'opt'},
            timeout_seconds=build_timeout))
  jobset.message('START', 'Building.', do_newline=True)
  num_failures, _ = jobset.run(
      build_jobs, newline_on_success=True, maxjobs=10)
  if num_failures == 0:
    jobset.message('SUCCESS',
                   'Built successfully.',
                   do_newline=True)
  else:
    jobset.message('FAILED', 'Build failed.',
                   do_newline=True)
    sys.exit(1)


def create_qpsworkers(languages, worker_hosts):
  """Creates QPS workers (but does not start them)."""
  if not worker_hosts:
    # run two workers locally (for each language)
    workers=[(None, 10000), (None, 10010)]
  elif len(worker_hosts) == 1:
    # run two workers on the remote host (for each language)
    workers=[(worker_hosts[0], 10000), (worker_hosts[0], 10010)]
  else:
    # run one worker per each remote host (for each language)
    workers=[(worker_host, 10000) for worker_host in worker_hosts]

  return [create_qpsworker_job(language,
                               shortname= 'qps_worker_%s_%s' % (language,
                                                                worker_idx),
                               port=worker[1] + language.worker_port_offset(),
                               remote_host=worker[0])
          for language in languages
          for worker_idx, worker in enumerate(workers)]


Scenario = collections.namedtuple('Scenario', 'jobspec workers name')


def create_scenarios(languages, workers_by_lang, remote_host=None, regex='.*',
                     category='all', bq_result_table=None,
                     netperf=False, netperf_hosts=[]):
  """Create jobspecs for scenarios to run."""
  all_workers = [worker
                 for workers in workers_by_lang.values()
                 for worker in workers]
  scenarios = []
  _NO_WORKERS = []

  if netperf:
    if not netperf_hosts:
      netperf_server='localhost'
      netperf_client=None
    elif len(netperf_hosts) == 1:
      netperf_server=netperf_hosts[0]
      netperf_client=netperf_hosts[0]
    else:
      netperf_server=netperf_hosts[0]
      netperf_client=netperf_hosts[1]
    scenarios.append(Scenario(
        create_netperf_jobspec(server_host=netperf_server,
                               client_host=netperf_client,
                               bq_result_table=bq_result_table),
        _NO_WORKERS, 'netperf'))

  for language in languages:
    for scenario_json in language.scenarios():
      if re.search(args.regex, scenario_json['name']):
        categories = scenario_json.get('CATEGORIES', ['scalable', 'smoketest'])
        if category in categories or category == 'all':
          workers = workers_by_lang[str(language)][:]
          # 'SERVER_LANGUAGE' is an indicator for this script to pick
          # a server in different language.
          custom_server_lang = scenario_json.get('SERVER_LANGUAGE', None)
          custom_client_lang = scenario_json.get('CLIENT_LANGUAGE', None)
          scenario_json = scenario_config.remove_nonproto_fields(scenario_json)
          if custom_server_lang and custom_client_lang:
            raise Exception('Cannot set both custom CLIENT_LANGUAGE and SERVER_LANGUAGE'
                            'in the same scenario')
          if custom_server_lang:
            if not workers_by_lang.get(custom_server_lang, []):
              print('Warning: Skipping scenario %s as' % scenario_json['name'])
              print('SERVER_LANGUAGE is set to %s yet the language has '
                    'not been selected with -l' % custom_server_lang)
              continue
            for idx in range(0, scenario_json['num_servers']):
              # replace first X workers by workers of a different language
              workers[idx] = workers_by_lang[custom_server_lang][idx]
          if custom_client_lang:
            if not workers_by_lang.get(custom_client_lang, []):
              print('Warning: Skipping scenario %s as' % scenario_json['name'])
              print('CLIENT_LANGUAGE is set to %s yet the language has '
                    'not been selected with -l' % custom_client_lang)
              continue
            for idx in range(scenario_json['num_servers'], len(workers)):
              # replace all client workers by workers of a different language,
              # leave num_server workers as they are server workers.
              workers[idx] = workers_by_lang[custom_client_lang][idx]
          scenario = Scenario(
              create_scenario_jobspec(scenario_json,
                                      [w.host_and_port for w in workers],
                                      remote_host=remote_host,
                                      bq_result_table=bq_result_table),
              workers,
              scenario_json['name'])
          scenarios.append(scenario)

  return scenarios


def finish_qps_workers(jobs):
  """Waits for given jobs to finish and eventually kills them."""
  retries = 0
  while any(job.is_running() for job in jobs):
    for job in qpsworker_jobs:
      if job.is_running():
        print('QPS worker "%s" is still running.' % job.host_and_port)
    if retries > 10:
      print('Killing all QPS workers.')
      for job in jobs:
        job.kill()
    retries += 1
    time.sleep(3)
  print('All QPS workers finished.')


argp = argparse.ArgumentParser(description='Run performance tests.')
argp.add_argument('-l', '--language',
                  choices=['all'] + sorted(scenario_config.LANGUAGES.keys()),
                  nargs='+',
                  required=True,
                  help='Languages to benchmark.')
argp.add_argument('--remote_driver_host',
                  default=None,
                  help='Run QPS driver on given host. By default, QPS driver is run locally.')
argp.add_argument('--remote_worker_host',
                  nargs='+',
                  default=[],
                  help='Worker hosts where to start QPS workers.')
argp.add_argument('--dry_run',
                  default=False,
                  action='store_const',
                  const=True,
                  help='Just list scenarios to be run, but don\'t run them.')
argp.add_argument('-r', '--regex', default='.*', type=str,
                  help='Regex to select scenarios to run.')
argp.add_argument('--bq_result_table', default=None, type=str,
                  help='Bigquery "dataset.table" to upload results to.')
argp.add_argument('--category',
                  choices=['smoketest','all','scalable','sweep'],
                  default='all',
                  help='Select a category of tests to run.')
argp.add_argument('--netperf',
                  default=False,
                  action='store_const',
                  const=True,
                  help='Run netperf benchmark as one of the scenarios.')

args = argp.parse_args()

languages = set(scenario_config.LANGUAGES[l]
                for l in itertools.chain.from_iterable(
                      scenario_config.LANGUAGES.iterkeys() if x == 'all' else [x]
                      for x in args.language))


# Put together set of remote hosts where to run and build
remote_hosts = set()
if args.remote_worker_host:
  for host in args.remote_worker_host:
    remote_hosts.add(host)
if args.remote_driver_host:
  remote_hosts.add(args.remote_driver_host)

if not args.dry_run:
  if remote_hosts:
    archive_repo(languages=[str(l) for l in languages])
    prepare_remote_hosts(remote_hosts, prepare_local=True)
  else:
    prepare_remote_hosts([], prepare_local=True)

build_local = False
if not args.remote_driver_host:
  build_local = True
if not args.dry_run:
  build_on_remote_hosts(remote_hosts, languages=[str(l) for l in languages], build_local=build_local)

qpsworker_jobs = create_qpsworkers(languages, args.remote_worker_host)

# get list of worker addresses for each language.
workers_by_lang = dict([(str(language), []) for language in languages])
for job in qpsworker_jobs:
  workers_by_lang[str(job.language)].append(job)

scenarios = create_scenarios(languages,
                           workers_by_lang=workers_by_lang,
                           remote_host=args.remote_driver_host,
                           regex=args.regex,
                           category=args.category,
                           bq_result_table=args.bq_result_table,
                           netperf=args.netperf,
                           netperf_hosts=args.remote_worker_host)

if not scenarios:
  raise Exception('No scenarios to run')

for scenario in scenarios:
  if args.dry_run:
    print(scenario.name)
  else:
    try:
      for worker in scenario.workers:
        worker.start()
      jobset.run([scenario.jobspec,
                  create_quit_jobspec(scenario.workers, remote_host=args.remote_driver_host)],
                 newline_on_success=True, maxjobs=1)
    finally:
      finish_qps_workers(scenario.workers)