blob: 0c7747567d84495d595828cf2b2aea62aa0c4013 [file] [log] [blame]
mbligh6203ace2007-10-04 21:54:24 +00001#!/usr/bin/python -u
mbligh1ffd5dc2008-11-25 13:24:05 +00002# Copyright 2007-2008 Martin J. Bligh <mbligh@google.com>, Google Inc.
mbligh82648e52008-11-20 16:54:25 +00003# Released under the GPL v2
mblighdcd57a82007-07-11 23:06:47 +00004
mblighc8949b82007-07-23 16:33:58 +00005"""
Aviv Keshetde6bb192013-01-30 16:17:22 -08006Run a control file through the server side engine
mblighdcd57a82007-07-11 23:06:47 +00007"""
mbligh1ffd5dc2008-11-25 13:24:05 +00008
Fang Deng042c1472014-10-23 13:56:41 -07009import ast
10import datetime
11import getpass
12import logging
13import os
14import re
15import signal
Dan Shicf4d2032015-03-12 15:04:21 -070016import socket
Fang Deng042c1472014-10-23 13:56:41 -070017import sys
18import traceback
19import time
20import urllib2
mbligh1ffd5dc2008-11-25 13:24:05 +000021
mblighf5427bb2008-04-09 15:55:57 +000022import common
mbligh9ff89cd2009-09-03 20:28:17 +000023
Dan Shia1ecd5c2013-06-06 11:21:31 -070024from autotest_lib.client.common_lib import control_data
Dan Shi32649b82015-08-29 20:53:36 -070025from autotest_lib.client.common_lib import error
Dan Shia1ecd5c2013-06-06 11:21:31 -070026from autotest_lib.client.common_lib import global_config
Dan Shi5ddf6a32015-05-02 00:22:01 -070027from autotest_lib.client.common_lib import utils
Dan Shi37bee222015-04-13 15:46:47 -070028from autotest_lib.client.common_lib.cros.graphite import autotest_es
29from autotest_lib.client.common_lib.cros.graphite import autotest_stats
Prashanth Balasubramanianf8b83712014-11-06 15:58:21 -080030try:
31 from autotest_lib.puppylab import results_mocker
32except ImportError:
33 results_mocker = None
34
Dan Shia06f3e22015-09-03 16:15:15 -070035_CONFIG = global_config.global_config
36
37require_atfork = _CONFIG.get_config_value(
mblighcb8cb332009-09-03 21:08:56 +000038 'AUTOSERV', 'require_atfork_module', type=bool, default=True)
39
Dan Shia1ecd5c2013-06-06 11:21:31 -070040
Jakob Jueliche497b552014-09-23 19:11:59 -070041# Number of seconds to wait before returning if testing mode is enabled
Prashanth B6285f6a2014-05-08 18:01:27 -070042TESTING_MODE_SLEEP_SECS = 1
Jakob Jueliche497b552014-09-23 19:11:59 -070043
mblighcb8cb332009-09-03 21:08:56 +000044try:
45 import atfork
46 atfork.monkeypatch_os_fork_functions()
47 import atfork.stdlib_fixer
48 # Fix the Python standard library for threading+fork safety with its
49 # internal locks. http://code.google.com/p/python-atfork/
50 import warnings
51 warnings.filterwarnings('ignore', 'logging module already imported')
52 atfork.stdlib_fixer.fix_logging_module()
53except ImportError, e:
54 from autotest_lib.client.common_lib import global_config
Dan Shia06f3e22015-09-03 16:15:15 -070055 if _CONFIG.get_config_value(
mblighcb8cb332009-09-03 21:08:56 +000056 'AUTOSERV', 'require_atfork_module', type=bool, default=False):
57 print >>sys.stderr, 'Please run utils/build_externals.py'
58 print e
59 sys.exit(1)
mbligh9ff89cd2009-09-03 20:28:17 +000060
Dan Shia1ecd5c2013-06-06 11:21:31 -070061from autotest_lib.server import frontend
showard75cdfee2009-06-10 17:40:41 +000062from autotest_lib.server import server_logging_config
showard043c62a2009-06-10 19:48:57 +000063from autotest_lib.server import server_job, utils, autoserv_parser, autotest
Dan Shia1ecd5c2013-06-06 11:21:31 -070064from autotest_lib.server import utils as server_utils
Dan Shicf4d2032015-03-12 15:04:21 -070065from autotest_lib.site_utils import job_directories
Fang Deng042c1472014-10-23 13:56:41 -070066from autotest_lib.site_utils import job_overhead
Dan Shicf4d2032015-03-12 15:04:21 -070067from autotest_lib.site_utils import lxc
Dan Shi7836d252015-04-27 15:33:58 -070068from autotest_lib.site_utils import lxc_utils
showard75cdfee2009-06-10 17:40:41 +000069from autotest_lib.client.common_lib import pidfile, logging_manager
Gabe Black1e1c41b2015-02-04 23:55:15 -080070from autotest_lib.client.common_lib.cros.graphite import autotest_stats
mbligh92c0fc22008-11-20 16:52:23 +000071
Dan Shicf4d2032015-03-12 15:04:21 -070072# Control segment to stage server-side package.
73STAGE_SERVER_SIDE_PACKAGE_CONTROL_FILE = server_job._control_segment_path(
74 'stage_server_side_package')
75
Dan Shia06f3e22015-09-03 16:15:15 -070076# Command line to start servod in a moblab.
77START_SERVOD_CMD = 'sudo start servod BOARD=%s PORT=%s'
78STOP_SERVOD_CMD = 'sudo stop servod'
79
Alex Millerf1af17e2013-01-09 22:50:32 -080080def log_alarm(signum, frame):
81 logging.error("Received SIGALARM. Ignoring and continuing on.")
Alex Miller0528d6f2013-01-11 10:49:48 -080082 sys.exit(1)
Alex Millerf1af17e2013-01-09 22:50:32 -080083
Dan Shicf4d2032015-03-12 15:04:21 -070084
85def _get_machines(parser):
86 """Get a list of machine names from command line arg -m or a file.
87
88 @param parser: Parser for the command line arguments.
89
90 @return: A list of machine names from command line arg -m or the
91 machines file specified in the command line arg -M.
92 """
93 if parser.options.machines:
94 machines = parser.options.machines.replace(',', ' ').strip().split()
95 else:
96 machines = []
97 machines_file = parser.options.machines_file
98 if machines_file:
99 machines = []
100 for m in open(machines_file, 'r').readlines():
101 # remove comments, spaces
102 m = re.sub('#.*', '', m).strip()
103 if m:
104 machines.append(m)
105 logging.debug('Read list of machines from file: %s', machines_file)
106 logging.debug('Machines: %s', ','.join(machines))
107
108 if machines:
109 for machine in machines:
110 if not machine or re.search('\s', machine):
111 parser.parser.error("Invalid machine: %s" % str(machine))
112 machines = list(set(machines))
113 machines.sort()
114 return machines
115
116
117def _stage_ssp(parser):
118 """Stage server-side package.
119
120 This function calls a control segment to stage server-side package based on
121 the job and autoserv command line option. The detail implementation could
122 be different for each host type. Currently, only CrosHost has
123 stage_server_side_package function defined.
124 The script returns None if no server-side package is available. However,
125 it may raise exception if it failed for reasons other than artifact (the
126 server-side package) not found.
127
128 @param parser: Command line arguments parser passed in the autoserv process.
129
130 @return: url of the staged server-side package. Return None if server-
131 side package is not found for the build.
132 """
Dan Shi36cfd832014-10-10 13:38:51 -0700133 # If test_source_build is not specified, default to use server-side test
134 # code from build specified in --image.
Dan Shicf4d2032015-03-12 15:04:21 -0700135 namespace = {'machines': _get_machines(parser),
Dan Shi36cfd832014-10-10 13:38:51 -0700136 'image': (parser.options.test_source_build or
137 parser.options.image),}
Dan Shicf4d2032015-03-12 15:04:21 -0700138 script_locals = {}
139 execfile(STAGE_SERVER_SIDE_PACKAGE_CONTROL_FILE, namespace, script_locals)
140 return script_locals['ssp_url']
141
142
143def _run_with_ssp(container_name, job_id, results, parser, ssp_url):
144 """Run the server job with server-side packaging.
145
146 @param container_name: Name of the container to run the test.
147 @param job_id: ID of the test job.
148 @param results: Folder to store results. This could be different from
149 parser.options.results:
150 parser.options.results can be set to None for results to be
151 stored in a temp folder.
152 results can be None for autoserv run requires no logging.
153 @param parser: Command line parser that contains the options.
154 @param ssp_url: url of the staged server-side package.
155 """
156 bucket = lxc.ContainerBucket()
157 control = (parser.args[0] if len(parser.args) > 0 and parser.args[0] != ''
158 else None)
159 test_container = bucket.setup_test(container_name, job_id, ssp_url, results,
160 control=control)
161 args = sys.argv[:]
162 args.remove('--require-ssp')
Dan Shi77b79a62015-07-29 16:22:05 -0700163 # --parent_job_id is only useful in autoserv running in host, not in
164 # container. Include this argument will cause test to fail for builds before
165 # CL 286265 was merged.
166 if '--parent_job_id' in args:
167 index = args.index('--parent_job_id')
168 args.remove('--parent_job_id')
169 # Remove the actual parent job id in command line arg.
170 del args[index]
Dan Shicf4d2032015-03-12 15:04:21 -0700171
172 # A dictionary of paths to replace in the command line. Key is the path to
173 # be replaced with the one in value.
174 paths_to_replace = {}
175 # Replace the control file path with the one in container.
176 if control:
177 container_control_filename = os.path.join(
178 lxc.CONTROL_TEMP_PATH, os.path.basename(control))
179 paths_to_replace[control] = container_control_filename
180 # Update result directory with the one in container.
181 if parser.options.results:
182 container_result_dir = os.path.join(lxc.RESULT_DIR_FMT % job_id)
183 paths_to_replace[parser.options.results] = container_result_dir
184 # Update parse_job directory with the one in container. The assumption is
185 # that the result folder to be parsed is always the same as the results_dir.
186 if parser.options.parse_job:
187 container_parse_dir = os.path.join(lxc.RESULT_DIR_FMT % job_id)
188 paths_to_replace[parser.options.parse_job] = container_result_dir
189
190 args = [paths_to_replace.get(arg, arg) for arg in args]
191
192 # Apply --use-existing-results, results directory is aready created and
193 # mounted in container. Apply this arg to avoid exception being raised.
194 if not '--use-existing-results' in args:
195 args.append('--use-existing-results')
196
197 # Make sure autoserv running in container using a different pid file.
198 if not '--pidfile-label' in args:
199 args.extend(['--pidfile-label', 'container_autoserv'])
200
Dan Shid1f51232015-04-18 00:29:14 -0700201 cmd_line = ' '.join(["'%s'" % arg if ' ' in arg else arg for arg in args])
Dan Shicf4d2032015-03-12 15:04:21 -0700202 logging.info('Run command in container: %s', cmd_line)
Dan Shi37bee222015-04-13 15:46:47 -0700203 success = False
Dan Shicf4d2032015-03-12 15:04:21 -0700204 try:
205 test_container.attach_run(cmd_line)
Dan Shi37bee222015-04-13 15:46:47 -0700206 success = True
Dan Shicf4d2032015-03-12 15:04:21 -0700207 finally:
Dan Shi37bee222015-04-13 15:46:47 -0700208 counter_key = '%s.%s' % (lxc.STATS_KEY,
209 'success' if success else 'fail')
210 autotest_stats.Counter(counter_key).increment()
211 # metadata is uploaded separately so it can use http to upload.
212 metadata = {'drone': socket.gethostname(),
213 'job_id': job_id,
214 'success': success}
215 autotest_es.post(use_http=True,
216 type_str=lxc.CONTAINER_RUN_TEST_METADB_TYPE,
217 metadata=metadata)
Dan Shicf4d2032015-03-12 15:04:21 -0700218 test_container.destroy()
219
220
Dan Shi3f1b8a52015-04-21 11:11:06 -0700221def correct_results_folder_permission(results):
222 """Make sure the results folder has the right permission settings.
223
224 For tests running with server-side packaging, the results folder has the
225 owner of root. This must be changed to the user running the autoserv
226 process, so parsing job can access the results folder.
227 TODO(dshi): crbug.com/459344 Remove this function when test container can be
228 unprivileged container.
229
230 @param results: Path to the results folder.
231
232 """
233 if not results:
234 return
235
Dan Shi32649b82015-08-29 20:53:36 -0700236 try:
237 utils.run('sudo -n chown -R %s "%s"' % (os.getuid(), results))
238 utils.run('sudo -n chgrp -R %s "%s"' % (os.getgid(), results))
239 except error.CmdError as e:
240 metadata = {'error': str(e),
241 'result_folder': results,
242 'drone': socket.gethostname()}
243 autotest_es.post(use_http=True, type_str='correct_results_folder_failure',
244 metadata=metadata)
245 raise
Dan Shi3f1b8a52015-04-21 11:11:06 -0700246
247
Dan Shia06f3e22015-09-03 16:15:15 -0700248def _start_servod(machine):
249 """Try to start servod in moblab if it's not already running or running with
250 different board or port.
251
252 @param machine: Name of the dut used for test.
253 """
254 if not utils.is_moblab():
255 return
256
Dan Shi1cded882015-09-23 16:52:26 -0700257 logging.debug('Trying to start servod.')
Dan Shia06f3e22015-09-03 16:15:15 -0700258 try:
259 afe = frontend.AFE()
260 board = server_utils.get_board_from_afe(machine, afe)
261 hosts = afe.get_hosts(hostname=machine)
262 servo_host = hosts[0].attributes.get('servo_host', None)
263 servo_port = hosts[0].attributes.get('servo_port', 9999)
264 if not servo_host in ['localhost', '127.0.0.1']:
Dan Shi1cded882015-09-23 16:52:26 -0700265 logging.warn('Starting servod is aborted. The dut\'s servo_host '
266 'attribute is not set to localhost.')
Dan Shia06f3e22015-09-03 16:15:15 -0700267 return
268 except (urllib2.HTTPError, urllib2.URLError):
269 # Ignore error if RPC failed to get board
270 logging.error('Failed to get board name from AFE. Start servod is '
271 'aborted')
272 return
273
274 try:
275 pid = utils.run('pgrep servod').stdout
276 cmd_line = utils.run('ps -fp %s' % pid).stdout
277 if ('--board %s' % board in cmd_line and
278 '--port %s' % servo_port in cmd_line):
279 logging.debug('Servod is already running with given board and port.'
280 ' There is no need to restart servod.')
281 return
282 logging.debug('Servod is running with different board or port. '
283 'Stopping existing servod.')
284 utils.run('sudo stop servod')
285 except error.CmdError:
286 # servod is not running.
287 pass
288
289 try:
290 utils.run(START_SERVOD_CMD % (board, servo_port))
291 logging.debug('Servod is started')
292 except error.CmdError as e:
293 logging.error('Servod failed to be started, error: %s', e)
294
295
Dan Shic68fefb2015-04-07 10:10:52 -0700296def run_autoserv(pid_file_manager, results, parser, ssp_url, use_ssp):
Dan Shicf4d2032015-03-12 15:04:21 -0700297 """Run server job with given options.
298
299 @param pid_file_manager: PidFileManager used to monitor the autoserv process
300 @param results: Folder to store results.
301 @param parser: Parser for the command line arguments.
302 @param ssp_url: Url to server-side package.
Dan Shic68fefb2015-04-07 10:10:52 -0700303 @param use_ssp: Set to True to run with server-side packaging.
Dan Shicf4d2032015-03-12 15:04:21 -0700304 """
Dan Shiec1d47d2015-02-13 11:38:13 -0800305 if parser.options.warn_no_ssp:
Dan Shic68fefb2015-04-07 10:10:52 -0700306 # Post a warning in the log.
Dan Shiec1d47d2015-02-13 11:38:13 -0800307 logging.warn('Autoserv is required to run with server-side packaging. '
308 'However, no drone is found to support server-side '
309 'packaging. The test will be executed in a drone without '
310 'server-side packaging supported.')
311
jadmanski0afbb632008-06-06 21:10:57 +0000312 # send stdin to /dev/null
313 dev_null = os.open(os.devnull, os.O_RDONLY)
314 os.dup2(dev_null, sys.stdin.fileno())
315 os.close(dev_null)
mblighdbf37612007-11-24 19:38:11 +0000316
jadmanski0afbb632008-06-06 21:10:57 +0000317 # Create separate process group
318 os.setpgrp()
mbligh1d42d4e2007-11-05 22:42:00 +0000319
Dan Shicf4d2032015-03-12 15:04:21 -0700320 # Container name is predefined so the container can be destroyed in
321 # handle_sigterm.
322 job_or_task_id = job_directories.get_job_id_or_task_id(
323 parser.options.results)
324 container_name = (lxc.TEST_CONTAINER_NAME_FMT %
Dan Shid68d51c2015-04-21 17:00:42 -0700325 (job_or_task_id, time.time(), os.getpid()))
Dan Shicf4d2032015-03-12 15:04:21 -0700326
jadmanski0afbb632008-06-06 21:10:57 +0000327 # Implement SIGTERM handler
mblighc2299562009-07-02 19:00:36 +0000328 def handle_sigterm(signum, frame):
Simran Basi9d9b7292013-10-16 16:44:07 -0700329 logging.debug('Received SIGTERM')
mblighff7d61f2008-12-22 14:53:35 +0000330 if pid_file_manager:
331 pid_file_manager.close_file(1, signal.SIGTERM)
Simran Basi49e21e62013-10-17 12:40:33 -0700332 logging.debug('Finished writing to pid_file. Killing process.')
Dan Shi3f1b8a52015-04-21 11:11:06 -0700333
334 # Update results folder's file permission. This needs to be done ASAP
335 # before the parsing process tries to access the log.
336 if use_ssp and results:
337 correct_results_folder_permission(results)
338
Simran Basid6b83772014-01-06 16:31:30 -0800339 # TODO (sbasi) - remove the time.sleep when crbug.com/302815 is solved.
340 # This sleep allows the pending output to be logged before the kill
341 # signal is sent.
342 time.sleep(.1)
Dan Shic68fefb2015-04-07 10:10:52 -0700343 if use_ssp:
Dan Shicf4d2032015-03-12 15:04:21 -0700344 logging.debug('Destroy container %s before aborting the autoserv '
345 'process.', container_name)
Dan Shi3f1b8a52015-04-21 11:11:06 -0700346 metadata = {'drone': socket.gethostname(),
347 'job_id': job_or_task_id,
348 'container_name': container_name,
349 'action': 'abort',
350 'success': True}
Dan Shicf4d2032015-03-12 15:04:21 -0700351 try:
352 bucket = lxc.ContainerBucket()
353 container = bucket.get(container_name)
354 if container:
355 container.destroy()
356 else:
Dan Shi3f1b8a52015-04-21 11:11:06 -0700357 metadata['success'] = False
358 metadata['error'] = 'container not found'
Dan Shicf4d2032015-03-12 15:04:21 -0700359 logging.debug('Container %s is not found.', container_name)
360 except:
Dan Shi3f1b8a52015-04-21 11:11:06 -0700361 metadata['success'] = False
362 metadata['error'] = 'Exception: %s' % sys.exc_info()
Dan Shicf4d2032015-03-12 15:04:21 -0700363 # Handle any exception so the autoserv process can be aborted.
364 logging.error('Failed to destroy container %s. Error: %s',
365 container_name, sys.exc_info())
Dan Shi3f1b8a52015-04-21 11:11:06 -0700366 autotest_es.post(use_http=True,
367 type_str=lxc.CONTAINER_RUN_TEST_METADB_TYPE,
368 metadata=metadata)
Dan Shie4a4f9f2015-07-20 09:00:25 -0700369 # Try to correct the result file permission again after the
370 # container is destroyed, as the container might have created some
371 # new files in the result folder.
372 if results:
373 correct_results_folder_permission(results)
Dan Shicf4d2032015-03-12 15:04:21 -0700374
jadmanski0afbb632008-06-06 21:10:57 +0000375 os.killpg(os.getpgrp(), signal.SIGKILL)
mblighfaf0cd42007-11-19 16:00:24 +0000376
jadmanski0afbb632008-06-06 21:10:57 +0000377 # Set signal handler
mblighc2299562009-07-02 19:00:36 +0000378 signal.signal(signal.SIGTERM, handle_sigterm)
mbligha46678d2008-05-01 20:00:01 +0000379
Simran Basid6b83772014-01-06 16:31:30 -0800380 # faulthandler is only needed to debug in the Lab and is not avaliable to
381 # be imported in the chroot as part of VMTest, so Try-Except it.
382 try:
383 import faulthandler
384 faulthandler.register(signal.SIGTERM, all_threads=True, chain=True)
385 logging.debug('faulthandler registered on SIGTERM.')
386 except ImportError:
Christopher Grant4beca022015-06-16 15:14:47 -0400387 sys.exc_clear()
Simran Basid6b83772014-01-06 16:31:30 -0800388
David Rochberg8a60d1e2011-02-01 14:22:07 -0500389 # Ignore SIGTTOU's generated by output from forked children.
390 signal.signal(signal.SIGTTOU, signal.SIG_IGN)
391
Alex Millerf1af17e2013-01-09 22:50:32 -0800392 # If we received a SIGALARM, let's be loud about it.
393 signal.signal(signal.SIGALRM, log_alarm)
394
mbligha5f5e542009-12-30 16:57:49 +0000395 # Server side tests that call shell scripts often depend on $USER being set
396 # but depending on how you launch your autotest scheduler it may not be set.
397 os.environ['USER'] = getpass.getuser()
398
mblighb2bea302008-07-24 20:25:57 +0000399 label = parser.options.label
mbligh374f3412009-05-13 21:29:45 +0000400 group_name = parser.options.group_name
mblighb2bea302008-07-24 20:25:57 +0000401 user = parser.options.user
402 client = parser.options.client
403 server = parser.options.server
jadmanski0afbb632008-06-06 21:10:57 +0000404 install_before = parser.options.install_before
mblighb2bea302008-07-24 20:25:57 +0000405 install_after = parser.options.install_after
406 verify = parser.options.verify
407 repair = parser.options.repair
showard45ae8192008-11-05 19:32:53 +0000408 cleanup = parser.options.cleanup
Alex Millercb79ba72013-05-29 14:43:00 -0700409 provision = parser.options.provision
Dan Shi07e09af2013-04-12 09:31:29 -0700410 reset = parser.options.reset
Alex Miller667b5f22014-02-28 15:33:39 -0800411 job_labels = parser.options.job_labels
mblighb2bea302008-07-24 20:25:57 +0000412 no_tee = parser.options.no_tee
jadmanski0afbb632008-06-06 21:10:57 +0000413 parse_job = parser.options.parse_job
mblighe7d9c602009-07-02 19:02:33 +0000414 execution_tag = parser.options.execution_tag
415 if not execution_tag:
416 execution_tag = parse_job
jadmanski0afbb632008-06-06 21:10:57 +0000417 ssh_user = parser.options.ssh_user
418 ssh_port = parser.options.ssh_port
419 ssh_pass = parser.options.ssh_pass
jadmanskidef0c3c2009-03-25 20:07:10 +0000420 collect_crashinfo = parser.options.collect_crashinfo
mblighe0cbc912010-03-11 18:03:07 +0000421 control_filename = parser.options.control_filename
Scott Zawalski91493c82013-01-25 16:15:20 -0500422 test_retry = parser.options.test_retry
beepscb6f1e22013-06-28 19:14:10 -0700423 verify_job_repo_url = parser.options.verify_job_repo_url
Christopher Wileyf594c5e2013-07-03 18:25:30 -0700424 skip_crash_collection = parser.options.skip_crash_collection
Aviv Keshet18ee3142013-08-12 15:01:51 -0700425 ssh_verbosity = int(parser.options.ssh_verbosity)
Fang Deng6cc20de2013-09-06 15:47:32 -0700426 ssh_options = parser.options.ssh_options
Dan Shib669cbd2013-09-13 11:17:17 -0700427 no_use_packaging = parser.options.no_use_packaging
mbligha46678d2008-05-01 20:00:01 +0000428
mblighb2bea302008-07-24 20:25:57 +0000429 # can't be both a client and a server side test
430 if client and server:
Eric Li861b2d52011-02-04 14:50:35 -0800431 parser.parser.error("Can not specify a test as both server and client!")
mblighb2bea302008-07-24 20:25:57 +0000432
Alex Millercb79ba72013-05-29 14:43:00 -0700433 if provision and client:
434 parser.parser.error("Cannot specify provisioning and client!")
435
436 is_special_task = (verify or repair or cleanup or collect_crashinfo or
Dan Shi07e09af2013-04-12 09:31:29 -0700437 provision or reset)
Alex Millercb79ba72013-05-29 14:43:00 -0700438 if len(parser.args) < 1 and not is_special_task:
Eric Li861b2d52011-02-04 14:50:35 -0800439 parser.parser.error("Missing argument: control file")
mbligha46678d2008-05-01 20:00:01 +0000440
Aviv Keshet18ee3142013-08-12 15:01:51 -0700441 if ssh_verbosity > 0:
442 # ssh_verbosity is an integer between 0 and 3, inclusive
443 ssh_verbosity_flag = '-' + 'v' * ssh_verbosity
Fang Dengd1c2b732013-08-20 12:59:46 -0700444 else:
445 ssh_verbosity_flag = ''
Aviv Keshet18ee3142013-08-12 15:01:51 -0700446
showard45ae8192008-11-05 19:32:53 +0000447 # We have a control file unless it's just a verify/repair/cleanup job
jadmanski0afbb632008-06-06 21:10:57 +0000448 if len(parser.args) > 0:
449 control = parser.args[0]
450 else:
451 control = None
mbligha46678d2008-05-01 20:00:01 +0000452
Dan Shicf4d2032015-03-12 15:04:21 -0700453 machines = _get_machines(parser)
mbligh374f3412009-05-13 21:29:45 +0000454 if group_name and len(machines) < 2:
Dan Shicf4d2032015-03-12 15:04:21 -0700455 parser.parser.error('-G %r may only be supplied with more than one '
456 'machine.' % group_name)
mbligh374f3412009-05-13 21:29:45 +0000457
Christopher Wiley8a91f232013-07-09 11:02:27 -0700458 kwargs = {'group_name': group_name, 'tag': execution_tag,
Dan Shicf4d2032015-03-12 15:04:21 -0700459 'disable_sysinfo': parser.options.disable_sysinfo}
Dan Shi70647ca2015-07-16 22:52:35 -0700460 if parser.options.parent_job_id:
461 kwargs['parent_job_id'] = int(parser.options.parent_job_id)
mblighe0cbc912010-03-11 18:03:07 +0000462 if control_filename:
463 kwargs['control_filename'] = control_filename
jadmanski0afbb632008-06-06 21:10:57 +0000464 job = server_job.server_job(control, parser.args[1:], results, label,
465 user, machines, client, parse_job,
Fang Dengd1c2b732013-08-20 12:59:46 -0700466 ssh_user, ssh_port, ssh_pass,
Aviv Keshetc5947fa2013-09-04 14:06:29 -0700467 ssh_verbosity_flag, ssh_options,
468 test_retry, **kwargs)
Dan Shicf4d2032015-03-12 15:04:21 -0700469
showard75cdfee2009-06-10 17:40:41 +0000470 job.logging.start_logging()
mbligh4608b002010-01-05 18:22:35 +0000471 job.init_parser()
mbligha46678d2008-05-01 20:00:01 +0000472
mbligh161fe6f2008-06-19 16:26:04 +0000473 # perform checks
474 job.precheck()
475
jadmanski0afbb632008-06-06 21:10:57 +0000476 # run the job
477 exit_code = 0
Dan Shic1b8bdd2015-09-14 23:11:24 -0700478 auto_start_servod = _CONFIG.get_config_value(
479 'AUTOSERV', 'auto_start_servod', type=bool, default=False)
jadmanski0afbb632008-06-06 21:10:57 +0000480 try:
mbligh332000a2009-06-08 16:47:28 +0000481 try:
482 if repair:
Dan Shic1b8bdd2015-09-14 23:11:24 -0700483 if auto_start_servod and len(machines) == 1:
484 _start_servod(machines[0])
J. Richard Barnettec2d99cf2015-11-18 12:46:15 -0800485 job.repair(job_labels)
mbligh332000a2009-06-08 16:47:28 +0000486 elif verify:
Alex Miller667b5f22014-02-28 15:33:39 -0800487 job.verify(job_labels)
Alex Millercb79ba72013-05-29 14:43:00 -0700488 elif provision:
Alex Miller667b5f22014-02-28 15:33:39 -0800489 job.provision(job_labels)
Dan Shi07e09af2013-04-12 09:31:29 -0700490 elif reset:
Alex Miller667b5f22014-02-28 15:33:39 -0800491 job.reset(job_labels)
Fang Dengad78aca2014-10-02 18:15:46 -0700492 elif cleanup:
493 job.cleanup(job_labels)
mbligh332000a2009-06-08 16:47:28 +0000494 else:
Dan Shia06f3e22015-09-03 16:15:15 -0700495 if auto_start_servod and len(machines) == 1:
496 _start_servod(machines[0])
Dan Shic68fefb2015-04-07 10:10:52 -0700497 if use_ssp:
Dan Shicf4d2032015-03-12 15:04:21 -0700498 try:
499 _run_with_ssp(container_name, job_or_task_id, results,
500 parser, ssp_url)
501 finally:
502 # Update the ownership of files in result folder.
Dan Shi3f1b8a52015-04-21 11:11:06 -0700503 correct_results_folder_permission(results)
Dan Shicf4d2032015-03-12 15:04:21 -0700504 else:
505 job.run(install_before, install_after,
506 verify_job_repo_url=verify_job_repo_url,
507 only_collect_crashinfo=collect_crashinfo,
508 skip_crash_collection=skip_crash_collection,
Dan Shib669cbd2013-09-13 11:17:17 -0700509 job_labels=job_labels,
510 use_packaging=(not no_use_packaging))
mbligh332000a2009-06-08 16:47:28 +0000511 finally:
512 while job.hosts:
513 host = job.hosts.pop()
514 host.close()
jadmanski0afbb632008-06-06 21:10:57 +0000515 except:
jadmanski27b37ea2008-10-29 23:54:31 +0000516 exit_code = 1
jadmanski0afbb632008-06-06 21:10:57 +0000517 traceback.print_exc()
mbligha46678d2008-05-01 20:00:01 +0000518
mblighff7d61f2008-12-22 14:53:35 +0000519 if pid_file_manager:
520 pid_file_manager.num_tests_failed = job.num_tests_failed
521 pid_file_manager.close_file(exit_code)
jadmanskie0dffc32008-12-15 17:30:30 +0000522 job.cleanup_parser()
showard21baa452008-10-21 00:08:39 +0000523
jadmanski27b37ea2008-10-29 23:54:31 +0000524 sys.exit(exit_code)
mbligha46678d2008-05-01 20:00:01 +0000525
526
Fang Deng042c1472014-10-23 13:56:41 -0700527def record_autoserv(options, duration_secs):
528 """Record autoserv end-to-end time in metadata db.
529
530 @param options: parser options.
531 @param duration_secs: How long autoserv has taken, in secs.
532 """
533 # Get machine hostname
534 machines = options.machines.replace(
535 ',', ' ').strip().split() if options.machines else []
536 num_machines = len(machines)
537 if num_machines > 1:
538 # Skip the case where atomic group is used.
539 return
540 elif num_machines == 0:
541 machines.append('hostless')
542
543 # Determine the status that will be reported.
544 s = job_overhead.STATUS
545 task_mapping = {
546 'reset': s.RESETTING, 'verify': s.VERIFYING,
547 'provision': s.PROVISIONING, 'repair': s.REPAIRING,
548 'cleanup': s.CLEANING, 'collect_crashinfo': s.GATHERING}
Dan Shi888cfca2015-07-31 15:49:00 -0700549 match = filter(lambda task: getattr(options, task, False) == True,
550 task_mapping)
Fang Deng042c1472014-10-23 13:56:41 -0700551 status = task_mapping[match[0]] if match else s.RUNNING
552 is_special_task = status not in [s.RUNNING, s.GATHERING]
Dan Shicf4d2032015-03-12 15:04:21 -0700553 job_or_task_id = job_directories.get_job_id_or_task_id(options.results)
Fang Deng042c1472014-10-23 13:56:41 -0700554 job_overhead.record_state_duration(
555 job_or_task_id, machines[0], status, duration_secs,
556 is_special_task=is_special_task)
557
558
mbligha46678d2008-05-01 20:00:01 +0000559def main():
Fang Deng042c1472014-10-23 13:56:41 -0700560 start_time = datetime.datetime.now()
Dan Shia1ecd5c2013-06-06 11:21:31 -0700561 # White list of tests with run time measurement enabled.
Dan Shia06f3e22015-09-03 16:15:15 -0700562 measure_run_time_tests_names = _CONFIG.get_config_value(
563 'AUTOSERV', 'measure_run_time_tests', type=str)
Dan Shia1ecd5c2013-06-06 11:21:31 -0700564 if measure_run_time_tests_names:
565 measure_run_time_tests = [t.strip() for t in
566 measure_run_time_tests_names.split(',')]
567 else:
568 measure_run_time_tests = []
jadmanski0afbb632008-06-06 21:10:57 +0000569 # grab the parser
570 parser = autoserv_parser.autoserv_parser
mbligha5cb4062009-02-17 15:53:39 +0000571 parser.parse_args()
mbligha46678d2008-05-01 20:00:01 +0000572
jadmanski0afbb632008-06-06 21:10:57 +0000573 if len(sys.argv) == 1:
574 parser.parser.print_help()
575 sys.exit(1)
mbligha6f13082008-06-05 23:53:46 +0000576
Dan Shicf4d2032015-03-12 15:04:21 -0700577 # If the job requires to run with server-side package, try to stage server-
578 # side package first. If that fails with error that autotest server package
Dan Shic68fefb2015-04-07 10:10:52 -0700579 # does not exist, fall back to run the job without using server-side
580 # packaging. If option warn_no_ssp is specified, that means autoserv is
581 # running in a drone does not support SSP, thus no need to stage server-side
582 # package.
Dan Shicf4d2032015-03-12 15:04:21 -0700583 ssp_url = None
Dan Shi0b754c52015-04-20 14:20:38 -0700584 ssp_url_warning = False
Dan Shic68fefb2015-04-07 10:10:52 -0700585 if (not parser.options.warn_no_ssp and parser.options.require_ssp):
Dan Shicf4d2032015-03-12 15:04:21 -0700586 ssp_url = _stage_ssp(parser)
Dan Shi0b754c52015-04-20 14:20:38 -0700587 # The build does not have autotest server package. Fall back to not
588 # to use server-side package. Logging is postponed until logging being
589 # set up.
590 ssp_url_warning = not ssp_url
Dan Shicf4d2032015-03-12 15:04:21 -0700591
showard75cdfee2009-06-10 17:40:41 +0000592 if parser.options.no_logging:
593 results = None
594 else:
595 results = parser.options.results
mbligh80e1eba2008-11-19 00:26:18 +0000596 if not results:
597 results = 'results.' + time.strftime('%Y-%m-%d-%H.%M.%S')
598 results = os.path.abspath(results)
showard566d3c02010-01-12 18:57:01 +0000599 resultdir_exists = False
600 for filename in ('control.srv', 'status.log', '.autoserv_execute'):
601 if os.path.exists(os.path.join(results, filename)):
602 resultdir_exists = True
mbligh4608b002010-01-05 18:22:35 +0000603 if not parser.options.use_existing_results and resultdir_exists:
mbligh80e1eba2008-11-19 00:26:18 +0000604 error = "Error: results directory already exists: %s\n" % results
605 sys.stderr.write(error)
606 sys.exit(1)
mbligha788dc42009-03-26 21:10:16 +0000607
608 # Now that we certified that there's no leftover results dir from
609 # previous jobs, lets create the result dir since the logging system
610 # needs to create the log file in there.
611 if not os.path.isdir(results):
612 os.makedirs(results)
showard75cdfee2009-06-10 17:40:41 +0000613
Dan Shic68fefb2015-04-07 10:10:52 -0700614 # Server-side packaging will only be used if it's required and the package
615 # is available. If warn_no_ssp is specified, it means that autoserv is
616 # running in a drone does not have SSP supported and a warning will be logs.
617 # Therefore, it should not run with SSP.
618 use_ssp = (not parser.options.warn_no_ssp and parser.options.require_ssp
619 and ssp_url)
620 if use_ssp:
Dan Shie28de552015-05-06 16:51:58 -0700621 log_dir = os.path.join(results, 'ssp_logs') if results else None
Dan Shicf4d2032015-03-12 15:04:21 -0700622 if log_dir and not os.path.exists(log_dir):
623 os.makedirs(log_dir)
624 else:
625 log_dir = results
Dan Shi3f1b8a52015-04-21 11:11:06 -0700626
showard75cdfee2009-06-10 17:40:41 +0000627 logging_manager.configure_logging(
Dan Shicf4d2032015-03-12 15:04:21 -0700628 server_logging_config.ServerLoggingConfig(),
629 results_dir=log_dir,
showard10d84172009-06-18 23:16:50 +0000630 use_console=not parser.options.no_tee,
631 verbose=parser.options.verbose,
632 no_console_prefix=parser.options.no_console_prefix)
Dan Shicf4d2032015-03-12 15:04:21 -0700633
Dan Shi0b754c52015-04-20 14:20:38 -0700634 if ssp_url_warning:
635 logging.warn(
636 'Autoserv is required to run with server-side packaging. '
637 'However, no server-side package can be found based on '
638 '`--image`, host attribute job_repo_url or host label of '
639 'cros-version. The test will be executed without '
640 'server-side packaging supported.')
641
showard75cdfee2009-06-10 17:40:41 +0000642 if results:
mbligha788dc42009-03-26 21:10:16 +0000643 logging.info("Results placed in %s" % results)
mbligh10717632008-11-19 00:21:57 +0000644
mbligh4608b002010-01-05 18:22:35 +0000645 # wait until now to perform this check, so it get properly logged
Dan Shicf4d2032015-03-12 15:04:21 -0700646 if (parser.options.use_existing_results and not resultdir_exists and
Dan Shiff78f112015-06-12 13:34:02 -0700647 not utils.is_in_container()):
mbligh4608b002010-01-05 18:22:35 +0000648 logging.error("No existing results directory found: %s", results)
649 sys.exit(1)
650
Dan Shicf4d2032015-03-12 15:04:21 -0700651 logging.debug('autoserv is running in drone %s.', socket.gethostname())
Aviv Keshet5c40ec62013-08-20 12:11:12 -0700652 logging.debug('autoserv command was: %s', ' '.join(sys.argv))
mbligh4608b002010-01-05 18:22:35 +0000653
Dan Shicf4d2032015-03-12 15:04:21 -0700654 if parser.options.write_pidfile and results:
mbligh4608b002010-01-05 18:22:35 +0000655 pid_file_manager = pidfile.PidFileManager(parser.options.pidfile_label,
656 results)
jadmanskid5ab8c52008-12-03 16:27:07 +0000657 pid_file_manager.open_file()
mblighff7d61f2008-12-22 14:53:35 +0000658 else:
659 pid_file_manager = None
mbligha46678d2008-05-01 20:00:01 +0000660
jadmanskif22fea82008-11-26 20:57:07 +0000661 autotest.BaseAutotest.set_install_in_tmpdir(
662 parser.options.install_in_tmpdir)
663
Dan Shia1ecd5c2013-06-06 11:21:31 -0700664 timer = None
665 try:
666 # Take the first argument as control file name, get the test name from
667 # the control file. If the test name exists in the list of tests with
668 # run time measurement enabled, start a timer to begin measurement.
669 if (len(parser.args) > 0 and parser.args[0] != '' and
670 parser.options.machines):
Dan Shibbc16132013-07-09 16:23:59 -0700671 try:
672 test_name = control_data.parse_control(parser.args[0],
673 raise_warnings=True).name
674 except control_data.ControlVariableException:
675 logging.debug('Failed to retrieve test name from control file.')
676 test_name = None
Dan Shia1ecd5c2013-06-06 11:21:31 -0700677 if test_name in measure_run_time_tests:
678 machines = parser.options.machines.replace(',', ' '
679 ).strip().split()
Dan Shi8eac5af2014-09-17 00:15:15 -0700680 try:
681 afe = frontend.AFE()
682 board = server_utils.get_board_from_afe(machines[0], afe)
Gabe Black1e1c41b2015-02-04 23:55:15 -0800683 timer = autotest_stats.Timer('autoserv_run_time.%s.%s' %
684 (board, test_name))
Dan Shi8eac5af2014-09-17 00:15:15 -0700685 timer.start()
686 except (urllib2.HTTPError, urllib2.URLError):
687 # Ignore error if RPC failed to get board
688 pass
Dan Shia1ecd5c2013-06-06 11:21:31 -0700689 except control_data.ControlVariableException as e:
690 logging.error(str(e))
jadmanski0afbb632008-06-06 21:10:57 +0000691 exit_code = 0
Prashanth B6285f6a2014-05-08 18:01:27 -0700692 # TODO(beeps): Extend this to cover different failure modes.
693 # Testing exceptions are matched against labels sent to autoserv. Eg,
694 # to allow only the hostless job to run, specify
695 # testing_exceptions: test_suite in the shadow_config. To allow both
696 # the hostless job and dummy_Pass to run, specify
697 # testing_exceptions: test_suite,dummy_Pass. You can figure out
698 # what label autoserv is invoked with by looking through the logs of a test
699 # for the autoserv command's -l option.
Dan Shia06f3e22015-09-03 16:15:15 -0700700 testing_exceptions = _CONFIG.get_config_value(
Prashanth B6285f6a2014-05-08 18:01:27 -0700701 'AUTOSERV', 'testing_exceptions', type=list, default=[])
Dan Shia06f3e22015-09-03 16:15:15 -0700702 test_mode = _CONFIG.get_config_value(
Prashanth B6285f6a2014-05-08 18:01:27 -0700703 'AUTOSERV', 'testing_mode', type=bool, default=False)
Prashanth Balasubramanianf8b83712014-11-06 15:58:21 -0800704 test_mode = (results_mocker and test_mode and not
705 any([ex in parser.options.label
706 for ex in testing_exceptions]))
707 is_task = (parser.options.verify or parser.options.repair or
708 parser.options.provision or parser.options.reset or
709 parser.options.cleanup or parser.options.collect_crashinfo)
jadmanski0afbb632008-06-06 21:10:57 +0000710 try:
711 try:
Prashanth B6285f6a2014-05-08 18:01:27 -0700712 if test_mode:
Prashanth Balasubramanianf8b83712014-11-06 15:58:21 -0800713 # The parser doesn't run on tasks anyway, so we can just return
714 # happy signals without faking results.
715 if not is_task:
716 machine = parser.options.results.split('/')[-1]
717
718 # TODO(beeps): The proper way to do this would be to
719 # refactor job creation so we can invoke job.record
720 # directly. To do that one needs to pipe the test_name
721 # through run_autoserv and bail just before invoking
722 # the server job. See the comment in
723 # puppylab/results_mocker for more context.
724 results_mocker.ResultsMocker(
Prashanth Balasubramanian22dd2262014-11-28 18:19:18 -0800725 test_name if test_name else 'unknown-test',
726 parser.options.results, machine
Prashanth Balasubramanianf8b83712014-11-06 15:58:21 -0800727 ).mock_results()
728 return
Prashanth B6285f6a2014-05-08 18:01:27 -0700729 else:
Dan Shic68fefb2015-04-07 10:10:52 -0700730 run_autoserv(pid_file_manager, results, parser, ssp_url,
731 use_ssp)
Aviv Keshet5c40ec62013-08-20 12:11:12 -0700732 except SystemExit as e:
jadmanski0afbb632008-06-06 21:10:57 +0000733 exit_code = e.code
Aviv Keshet5c40ec62013-08-20 12:11:12 -0700734 if exit_code:
735 logging.exception(e)
736 except Exception as e:
jadmanski0afbb632008-06-06 21:10:57 +0000737 # If we don't know what happened, we'll classify it as
738 # an 'abort' and return 1.
Aviv Keshet5c40ec62013-08-20 12:11:12 -0700739 logging.exception(e)
jadmanski0afbb632008-06-06 21:10:57 +0000740 exit_code = 1
741 finally:
mblighff7d61f2008-12-22 14:53:35 +0000742 if pid_file_manager:
743 pid_file_manager.close_file(exit_code)
Dan Shia1ecd5c2013-06-06 11:21:31 -0700744 if timer:
745 timer.stop()
Fang Deng042c1472014-10-23 13:56:41 -0700746 # Record the autoserv duration time. Must be called
747 # just before the system exits to ensure accuracy.
748 duration_secs = (datetime.datetime.now() - start_time).total_seconds()
749 record_autoserv(parser.options, duration_secs)
jadmanski0afbb632008-06-06 21:10:57 +0000750 sys.exit(exit_code)
mblighfaf0cd42007-11-19 16:00:24 +0000751
mblighbb421852008-03-11 22:36:16 +0000752
mbligha46678d2008-05-01 20:00:01 +0000753if __name__ == '__main__':
jadmanski0afbb632008-06-06 21:10:57 +0000754 main()