blob: c732b8125870def8ae63d0b88a335d07651b9deb [file] [log] [blame]
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +00001#!/usr/bin/env python
maruelea586f32016-04-05 11:11:33 -07002# Copyright 2012 The LUCI Authors. All rights reserved.
maruelf1f5e2a2016-05-25 17:10:39 -07003# Use of this source code is governed under the Apache License, Version 2.0
4# that can be found in the LICENSE file.
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +00005
nodir55be77b2016-05-03 09:39:57 -07006"""Runs a command with optional isolated input/output.
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +00007
nodir55be77b2016-05-03 09:39:57 -07008Despite name "run_isolated", can run a generic non-isolated command specified as
9args.
10
11If input isolated hash is provided, fetches it, creates a tree of hard links,
12appends args to the command in the fetched isolated and runs it.
13To improve performance, keeps a local cache.
14The local cache can safely be deleted.
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -050015
nodirbe642ff2016-06-09 15:51:51 -070016Any ${EXECUTABLE_SUFFIX} on the command line will be replaced with ".exe" string
17on Windows and "" on other platforms.
18
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -050019Any ${ISOLATED_OUTDIR} on the command line will be replaced by the location of a
20temporary directory upon execution of the command specified in the .isolated
21file. All content written to this directory will be uploaded upon termination
22and the .isolated file describing this directory will be printed to stdout.
bpastene447c1992016-06-20 15:21:47 -070023
24Any ${SWARMING_BOT_FILE} on the command line will be replaced by the value of
25the --bot-file parameter. This file is used by a swarming bot to communicate
26state of the host to tasks. It is written to by the swarming bot's
27on_before_task() hook in the swarming server's custom bot_config.py.
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +000028"""
29
maruele2f2cb82016-07-13 14:41:03 -070030__version__ = '0.8.2'
maruel@chromium.orgdedbf492013-09-12 20:42:11 +000031
maruel064c0a32016-04-05 11:47:15 -070032import base64
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +000033import logging
34import optparse
35import os
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +000036import sys
37import tempfile
maruel064c0a32016-04-05 11:47:15 -070038import time
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +000039
vadimsh@chromium.orga4326472013-08-24 02:05:41 +000040from third_party.depot_tools import fix_encoding
41
Vadim Shtayura6b555c12014-07-23 16:22:18 -070042from utils import file_path
maruel12e30012015-10-09 11:55:35 -070043from utils import fs
maruel064c0a32016-04-05 11:47:15 -070044from utils import large
Marc-Antoine Ruelf74cffe2015-07-15 15:21:34 -040045from utils import logging_utils
Marc-Antoine Ruelcfb60852014-07-02 15:22:00 -040046from utils import on_error
Marc-Antoine Ruelc44f5722015-01-08 16:10:01 -050047from utils import subprocess42
vadimsh@chromium.orga4326472013-08-24 02:05:41 +000048from utils import tools
vadimsh@chromium.org3e97deb2013-08-24 00:56:44 +000049from utils import zip_package
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +000050
Vadim Shtayurae34e13a2014-02-02 11:23:26 -080051import auth
nodirbe642ff2016-06-09 15:51:51 -070052import cipd
maruel@chromium.orgdedbf492013-09-12 20:42:11 +000053import isolateserver
maruel@chromium.orgdedbf492013-09-12 20:42:11 +000054
vadimsh@chromium.orga4326472013-08-24 02:05:41 +000055
vadimsh@chromium.org85071062013-08-21 23:37:45 +000056# Absolute path to this file (can be None if running from zip on Mac).
tansella4949442016-06-23 22:34:32 -070057THIS_FILE_PATH = os.path.abspath(
58 __file__.decode(sys.getfilesystemencoding())) if __file__ else None
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +000059
60# Directory that contains this file (might be inside zip package).
tansella4949442016-06-23 22:34:32 -070061BASE_DIR = os.path.dirname(THIS_FILE_PATH) if __file__.decode(
62 sys.getfilesystemencoding()) else None
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +000063
64# Directory that contains currently running script file.
maruel@chromium.org814d23f2013-10-01 19:08:00 +000065if zip_package.get_main_script_path():
66 MAIN_DIR = os.path.dirname(
67 os.path.abspath(zip_package.get_main_script_path()))
68else:
69 # This happens when 'import run_isolated' is executed at the python
70 # interactive prompt, in that case __file__ is undefined.
71 MAIN_DIR = None
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +000072
maruele2f2cb82016-07-13 14:41:03 -070073
74# Magic variables that can be found in the isolate task command line.
75ISOLATED_OUTDIR_PARAMETER = '${ISOLATED_OUTDIR}'
76EXECUTABLE_SUFFIX_PARAMETER = '${EXECUTABLE_SUFFIX}'
77SWARMING_BOT_FILE_PARAMETER = '${SWARMING_BOT_FILE}'
78
79
csharp@chromium.orgff2a4662012-11-21 20:49:32 +000080# The name of the log file to use.
81RUN_ISOLATED_LOG_FILE = 'run_isolated.log'
82
maruele2f2cb82016-07-13 14:41:03 -070083
csharp@chromium.orge217f302012-11-22 16:51:53 +000084# The name of the log to use for the run_test_cases.py command
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +000085RUN_TEST_CASES_LOG = 'run_test_cases.log'
csharp@chromium.orge217f302012-11-22 16:51:53 +000086
vadimsh@chromium.org87d63262013-04-04 19:34:21 +000087
maruele2f2cb82016-07-13 14:41:03 -070088# Use short names for temporary directories. This is driven by Windows, which
89# imposes a relatively short maximum path length of 260 characters, often
90# referred to as MAX_PATH. It is relatively easy to create files with longer
91# path length. A use case is with recursive depedency treesV like npm packages.
92#
93# It is recommended to start the script with a `root_dir` as short as
94# possible.
95# - ir stands for isolated_run
96# - io stands for isolated_out
97# - it stands for isolated_tmp
98ISOLATED_RUN_DIR = u'ir'
99ISOLATED_OUT_DIR = u'io'
100ISOLATED_TMP_DIR = u'it'
101
102
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +0000103def get_as_zip_package(executable=True):
104 """Returns ZipPackage with this module and all its dependencies.
105
106 If |executable| is True will store run_isolated.py as __main__.py so that
107 zip package is directly executable be python.
108 """
109 # Building a zip package when running from another zip package is
110 # unsupported and probably unneeded.
111 assert not zip_package.is_zipped_module(sys.modules[__name__])
vadimsh@chromium.org85071062013-08-21 23:37:45 +0000112 assert THIS_FILE_PATH
113 assert BASE_DIR
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +0000114 package = zip_package.ZipPackage(root=BASE_DIR)
115 package.add_python_file(THIS_FILE_PATH, '__main__.py' if executable else None)
Marc-Antoine Ruel8bee66d2014-08-28 19:02:07 -0400116 package.add_python_file(os.path.join(BASE_DIR, 'isolated_format.py'))
maruel@chromium.orgdedbf492013-09-12 20:42:11 +0000117 package.add_python_file(os.path.join(BASE_DIR, 'isolateserver.py'))
Vadim Shtayurae34e13a2014-02-02 11:23:26 -0800118 package.add_python_file(os.path.join(BASE_DIR, 'auth.py'))
nodirbe642ff2016-06-09 15:51:51 -0700119 package.add_python_file(os.path.join(BASE_DIR, 'cipd.py'))
vadimsh@chromium.org8b9d56b2013-08-21 22:24:35 +0000120 package.add_directory(os.path.join(BASE_DIR, 'third_party'))
121 package.add_directory(os.path.join(BASE_DIR, 'utils'))
122 return package
123
124
maruel03e11842016-07-14 10:50:16 -0700125def make_temp_dir(prefix, root_dir):
126 """Returns a new unique temporary directory."""
127 return unicode(tempfile.mkdtemp(prefix=prefix, dir=root_dir))
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000128
129
Marc-Antoine Ruel7124e392014-01-09 11:49:21 -0500130def change_tree_read_only(rootdir, read_only):
131 """Changes the tree read-only bits according to the read_only specification.
132
133 The flag can be 0, 1 or 2, which will affect the possibility to modify files
134 and create or delete files.
135 """
136 if read_only == 2:
137 # Files and directories (except on Windows) are marked read only. This
138 # inhibits modifying, creating or deleting files in the test directory,
139 # except on Windows where creating and deleting files is still possible.
Marc-Antoine Ruele4ad07e2014-10-15 20:22:29 -0400140 file_path.make_tree_read_only(rootdir)
Marc-Antoine Ruel7124e392014-01-09 11:49:21 -0500141 elif read_only == 1:
142 # Files are marked read only but not the directories. This inhibits
143 # modifying files but creating or deleting files is still possible.
Marc-Antoine Ruele4ad07e2014-10-15 20:22:29 -0400144 file_path.make_tree_files_read_only(rootdir)
Marc-Antoine Ruel7124e392014-01-09 11:49:21 -0500145 elif read_only in (0, None):
Marc-Antoine Ruelf1d827c2014-11-24 15:22:25 -0500146 # Anything can be modified.
Marc-Antoine Ruel7124e392014-01-09 11:49:21 -0500147 # TODO(maruel): This is currently dangerous as long as DiskCache.touch()
148 # is not yet changed to verify the hash of the content of the files it is
149 # looking at, so that if a test modifies an input file, the file must be
150 # deleted.
Marc-Antoine Ruele4ad07e2014-10-15 20:22:29 -0400151 file_path.make_tree_writeable(rootdir)
Marc-Antoine Ruel7124e392014-01-09 11:49:21 -0500152 else:
153 raise ValueError(
154 'change_tree_read_only(%s, %s): Unknown flag %s' %
155 (rootdir, read_only, read_only))
156
157
nodir90bc8dc2016-06-15 13:35:21 -0700158def process_command(command, out_dir, bot_file):
nodirbe642ff2016-06-09 15:51:51 -0700159 """Replaces variables in a command line.
160
161 Raises:
162 ValueError if a parameter is requested in |command| but its value is not
163 provided.
164 """
maruela9cfd6f2015-09-15 11:03:15 -0700165 def fix(arg):
nodirbe642ff2016-06-09 15:51:51 -0700166 arg = arg.replace(EXECUTABLE_SUFFIX_PARAMETER, cipd.EXECUTABLE_SUFFIX)
167 replace_slash = False
nodir55be77b2016-05-03 09:39:57 -0700168 if ISOLATED_OUTDIR_PARAMETER in arg:
nodirbe642ff2016-06-09 15:51:51 -0700169 if not out_dir:
maruel7f63a272016-07-12 12:40:36 -0700170 raise ValueError(
171 'output directory is requested in command, but not provided; '
172 'please specify one')
nodir55be77b2016-05-03 09:39:57 -0700173 arg = arg.replace(ISOLATED_OUTDIR_PARAMETER, out_dir)
nodirbe642ff2016-06-09 15:51:51 -0700174 replace_slash = True
nodir90bc8dc2016-06-15 13:35:21 -0700175 if SWARMING_BOT_FILE_PARAMETER in arg:
176 if bot_file:
177 arg = arg.replace(SWARMING_BOT_FILE_PARAMETER, bot_file)
178 replace_slash = True
179 else:
180 logging.warning('SWARMING_BOT_FILE_PARAMETER found in command, but no '
181 'bot_file specified. Leaving parameter unchanged.')
nodirbe642ff2016-06-09 15:51:51 -0700182 if replace_slash:
183 # Replace slashes only if parameters are present
nodir55be77b2016-05-03 09:39:57 -0700184 # because of arguments like '${ISOLATED_OUTDIR}/foo/bar'
185 arg = arg.replace('/', os.sep)
maruela9cfd6f2015-09-15 11:03:15 -0700186 return arg
187
188 return [fix(arg) for arg in command]
189
190
maruel6be7f9e2015-10-01 12:25:30 -0700191def run_command(command, cwd, tmp_dir, hard_timeout, grace_period):
192 """Runs the command.
193
194 Returns:
195 tuple(process exit code, bool if had a hard timeout)
196 """
maruela9cfd6f2015-09-15 11:03:15 -0700197 logging.info('run_command(%s, %s)' % (command, cwd))
marueleb5fbee2015-09-17 13:01:36 -0700198
199 env = os.environ.copy()
200 if sys.platform == 'darwin':
tansella4949442016-06-23 22:34:32 -0700201 env['TMPDIR'] = tmp_dir.encode(sys.getfilesystemencoding())
marueleb5fbee2015-09-17 13:01:36 -0700202 elif sys.platform == 'win32':
tansella4949442016-06-23 22:34:32 -0700203 env['TEMP'] = tmp_dir.encode(sys.getfilesystemencoding())
marueleb5fbee2015-09-17 13:01:36 -0700204 else:
tansella4949442016-06-23 22:34:32 -0700205 env['TMP'] = tmp_dir.encode(sys.getfilesystemencoding())
maruel6be7f9e2015-10-01 12:25:30 -0700206 exit_code = None
207 had_hard_timeout = False
maruela9cfd6f2015-09-15 11:03:15 -0700208 with tools.Profiler('RunTest'):
maruel6be7f9e2015-10-01 12:25:30 -0700209 proc = None
210 had_signal = []
maruela9cfd6f2015-09-15 11:03:15 -0700211 try:
maruel6be7f9e2015-10-01 12:25:30 -0700212 # TODO(maruel): This code is imperfect. It doesn't handle well signals
213 # during the download phase and there's short windows were things can go
214 # wrong.
215 def handler(signum, _frame):
216 if proc and not had_signal:
217 logging.info('Received signal %d', signum)
218 had_signal.append(True)
maruel556d9052015-10-05 11:12:44 -0700219 raise subprocess42.TimeoutExpired(command, None)
maruel6be7f9e2015-10-01 12:25:30 -0700220
221 proc = subprocess42.Popen(command, cwd=cwd, env=env, detached=True)
222 with subprocess42.set_signal_handler(subprocess42.STOP_SIGNALS, handler):
223 try:
224 exit_code = proc.wait(hard_timeout or None)
225 except subprocess42.TimeoutExpired:
226 if not had_signal:
227 logging.warning('Hard timeout')
228 had_hard_timeout = True
229 logging.warning('Sending SIGTERM')
230 proc.terminate()
231
232 # Ignore signals in grace period. Forcibly give the grace period to the
233 # child process.
234 if exit_code is None:
235 ignore = lambda *_: None
236 with subprocess42.set_signal_handler(subprocess42.STOP_SIGNALS, ignore):
237 try:
238 exit_code = proc.wait(grace_period or None)
239 except subprocess42.TimeoutExpired:
240 # Now kill for real. The user can distinguish between the
241 # following states:
242 # - signal but process exited within grace period,
243 # hard_timed_out will be set but the process exit code will be
244 # script provided.
245 # - processed exited late, exit code will be -9 on posix.
246 logging.warning('Grace exhausted; sending SIGKILL')
247 proc.kill()
248 logging.info('Waiting for proces exit')
249 exit_code = proc.wait()
maruela9cfd6f2015-09-15 11:03:15 -0700250 except OSError:
251 # This is not considered to be an internal error. The executable simply
252 # does not exit.
maruela72f46e2016-02-24 11:05:45 -0800253 sys.stderr.write(
254 '<The executable does not exist or a dependent library is missing>\n'
255 '<Check for missing .so/.dll in the .isolate or GN file>\n'
256 '<Command: %s>\n' % command)
257 if os.environ.get('SWARMING_TASK_ID'):
258 # Give an additional hint when running as a swarming task.
259 sys.stderr.write(
260 '<See the task\'s page for commands to help diagnose this issue '
261 'by reproducing the task locally>\n')
maruela9cfd6f2015-09-15 11:03:15 -0700262 exit_code = 1
263 logging.info(
264 'Command finished with exit code %d (%s)',
265 exit_code, hex(0xffffffff & exit_code))
maruel6be7f9e2015-10-01 12:25:30 -0700266 return exit_code, had_hard_timeout
maruela9cfd6f2015-09-15 11:03:15 -0700267
268
nodir6f801882016-04-29 14:41:50 -0700269def fetch_and_measure(isolated_hash, storage, cache, outdir):
270 """Fetches an isolated and returns (bundle, stats)."""
271 start = time.time()
272 bundle = isolateserver.fetch_isolated(
273 isolated_hash=isolated_hash,
274 storage=storage,
275 cache=cache,
276 outdir=outdir)
277 return bundle, {
278 'duration': time.time() - start,
279 'initial_number_items': cache.initial_number_items,
280 'initial_size': cache.initial_size,
281 'items_cold': base64.b64encode(large.pack(sorted(cache.added))),
282 'items_hot': base64.b64encode(
283 large.pack(sorted(set(cache.linked) - set(cache.added)))),
284 }
285
286
maruela9cfd6f2015-09-15 11:03:15 -0700287def delete_and_upload(storage, out_dir, leak_temp_dir):
288 """Deletes the temporary run directory and uploads results back.
289
290 Returns:
nodir6f801882016-04-29 14:41:50 -0700291 tuple(outputs_ref, success, stats)
maruel064c0a32016-04-05 11:47:15 -0700292 - outputs_ref: a dict referring to the results archived back to the isolated
293 server, if applicable.
294 - success: False if something occurred that means that the task must
295 forcibly be considered a failure, e.g. zombie processes were left
296 behind.
nodir6f801882016-04-29 14:41:50 -0700297 - stats: uploading stats.
maruela9cfd6f2015-09-15 11:03:15 -0700298 """
299
300 # Upload out_dir and generate a .isolated file out of this directory. It is
301 # only done if files were written in the directory.
302 outputs_ref = None
maruel064c0a32016-04-05 11:47:15 -0700303 cold = []
304 hot = []
nodir6f801882016-04-29 14:41:50 -0700305 start = time.time()
306
maruel12e30012015-10-09 11:55:35 -0700307 if fs.isdir(out_dir) and fs.listdir(out_dir):
maruela9cfd6f2015-09-15 11:03:15 -0700308 with tools.Profiler('ArchiveOutput'):
309 try:
maruel064c0a32016-04-05 11:47:15 -0700310 results, f_cold, f_hot = isolateserver.archive_files_to_storage(
maruela9cfd6f2015-09-15 11:03:15 -0700311 storage, [out_dir], None)
312 outputs_ref = {
313 'isolated': results[0][0],
314 'isolatedserver': storage.location,
315 'namespace': storage.namespace,
316 }
maruel064c0a32016-04-05 11:47:15 -0700317 cold = sorted(i.size for i in f_cold)
318 hot = sorted(i.size for i in f_hot)
maruela9cfd6f2015-09-15 11:03:15 -0700319 except isolateserver.Aborted:
320 # This happens when a signal SIGTERM was received while uploading data.
321 # There is 2 causes:
322 # - The task was too slow and was about to be killed anyway due to
323 # exceeding the hard timeout.
324 # - The amount of data uploaded back is very large and took too much
325 # time to archive.
326 sys.stderr.write('Received SIGTERM while uploading')
327 # Re-raise, so it will be treated as an internal failure.
328 raise
nodir6f801882016-04-29 14:41:50 -0700329
330 success = False
maruela9cfd6f2015-09-15 11:03:15 -0700331 try:
maruel12e30012015-10-09 11:55:35 -0700332 if (not leak_temp_dir and fs.isdir(out_dir) and
maruel6eeea7d2015-09-16 12:17:42 -0700333 not file_path.rmtree(out_dir)):
maruela9cfd6f2015-09-15 11:03:15 -0700334 logging.error('Had difficulties removing out_dir %s', out_dir)
nodir6f801882016-04-29 14:41:50 -0700335 else:
336 success = True
maruela9cfd6f2015-09-15 11:03:15 -0700337 except OSError as e:
338 # When this happens, it means there's a process error.
maruel12e30012015-10-09 11:55:35 -0700339 logging.exception('Had difficulties removing out_dir %s: %s', out_dir, e)
nodir6f801882016-04-29 14:41:50 -0700340 stats = {
341 'duration': time.time() - start,
342 'items_cold': base64.b64encode(large.pack(cold)),
343 'items_hot': base64.b64encode(large.pack(hot)),
344 }
345 return outputs_ref, success, stats
maruela9cfd6f2015-09-15 11:03:15 -0700346
347
marueleb5fbee2015-09-17 13:01:36 -0700348def map_and_run(
nodir55be77b2016-05-03 09:39:57 -0700349 command, isolated_hash, storage, cache, leak_temp_dir, root_dir,
nodir90bc8dc2016-06-15 13:35:21 -0700350 hard_timeout, grace_period, bot_file, extra_args, install_packages_fn):
nodir55be77b2016-05-03 09:39:57 -0700351 """Runs a command with optional isolated input/output.
352
353 See run_tha_test for argument documentation.
354
355 Returns metadata about the result.
356 """
357 assert bool(command) ^ bool(isolated_hash)
maruela9cfd6f2015-09-15 11:03:15 -0700358 result = {
maruel064c0a32016-04-05 11:47:15 -0700359 'duration': None,
maruela9cfd6f2015-09-15 11:03:15 -0700360 'exit_code': None,
maruel6be7f9e2015-10-01 12:25:30 -0700361 'had_hard_timeout': False,
maruela9cfd6f2015-09-15 11:03:15 -0700362 'internal_failure': None,
maruel064c0a32016-04-05 11:47:15 -0700363 'stats': {
nodir55715712016-06-03 12:28:19 -0700364 # 'isolated': {
nodirbe642ff2016-06-09 15:51:51 -0700365 # 'cipd': {
366 # 'duration': 0.,
367 # 'get_client_duration': 0.,
368 # },
nodir55715712016-06-03 12:28:19 -0700369 # 'download': {
370 # 'duration': 0.,
371 # 'initial_number_items': 0,
372 # 'initial_size': 0,
373 # 'items_cold': '<large.pack()>',
374 # 'items_hot': '<large.pack()>',
375 # },
376 # 'upload': {
377 # 'duration': 0.,
378 # 'items_cold': '<large.pack()>',
379 # 'items_hot': '<large.pack()>',
380 # },
maruel064c0a32016-04-05 11:47:15 -0700381 # },
382 },
maruela9cfd6f2015-09-15 11:03:15 -0700383 'outputs_ref': None,
nodirbe642ff2016-06-09 15:51:51 -0700384 'version': 5,
maruela9cfd6f2015-09-15 11:03:15 -0700385 }
nodirbe642ff2016-06-09 15:51:51 -0700386
marueleb5fbee2015-09-17 13:01:36 -0700387 if root_dir:
nodire5028a92016-04-29 14:38:21 -0700388 file_path.ensure_tree(root_dir, 0700)
marueleb5fbee2015-09-17 13:01:36 -0700389 else:
maruel03e11842016-07-14 10:50:16 -0700390 root_dir = (
391 os.path.dirname(cache.cache_dir) if cache.cache_dir else os.getcwd())
maruele2f2cb82016-07-13 14:41:03 -0700392 # See comment for these constants.
393 run_dir = make_temp_dir(ISOLATED_RUN_DIR, root_dir)
maruel03e11842016-07-14 10:50:16 -0700394 # storage should be normally set but don't crash if it is not. This can happen
395 # as Swarming task can run without an isolate server.
maruele2f2cb82016-07-13 14:41:03 -0700396 out_dir = make_temp_dir(ISOLATED_OUT_DIR, root_dir) if storage else None
397 tmp_dir = make_temp_dir(ISOLATED_TMP_DIR, root_dir)
nodir55be77b2016-05-03 09:39:57 -0700398 cwd = run_dir
maruela9cfd6f2015-09-15 11:03:15 -0700399
nodir55be77b2016-05-03 09:39:57 -0700400 try:
nodir90bc8dc2016-06-15 13:35:21 -0700401 cipd_stats = install_packages_fn(run_dir)
402 if cipd_stats:
403 result['stats']['cipd'] = cipd_stats
404
nodir55be77b2016-05-03 09:39:57 -0700405 if isolated_hash:
nodir55715712016-06-03 12:28:19 -0700406 isolated_stats = result['stats'].setdefault('isolated', {})
407 bundle, isolated_stats['download'] = fetch_and_measure(
nodir55be77b2016-05-03 09:39:57 -0700408 isolated_hash=isolated_hash,
409 storage=storage,
410 cache=cache,
411 outdir=run_dir)
412 if not bundle.command:
413 # Handle this as a task failure, not an internal failure.
414 sys.stderr.write(
415 '<The .isolated doesn\'t declare any command to run!>\n'
416 '<Check your .isolate for missing \'command\' variable>\n')
417 if os.environ.get('SWARMING_TASK_ID'):
418 # Give an additional hint when running as a swarming task.
419 sys.stderr.write('<This occurs at the \'isolate\' step>\n')
420 result['exit_code'] = 1
421 return result
422
423 change_tree_read_only(run_dir, bundle.read_only)
424 cwd = os.path.normpath(os.path.join(cwd, bundle.relative_cwd))
425 command = bundle.command + extra_args
nodirbe642ff2016-06-09 15:51:51 -0700426
nodir34d673c2016-05-24 09:30:48 -0700427 command = tools.fix_python_path(command)
nodir90bc8dc2016-06-15 13:35:21 -0700428 command = process_command(command, out_dir, bot_file)
maruela9cfd6f2015-09-15 11:03:15 -0700429 file_path.ensure_command_has_abs_path(command, cwd)
nodirbe642ff2016-06-09 15:51:51 -0700430
maruel064c0a32016-04-05 11:47:15 -0700431 sys.stdout.flush()
432 start = time.time()
433 try:
434 result['exit_code'], result['had_hard_timeout'] = run_command(
nodirbe642ff2016-06-09 15:51:51 -0700435 command, cwd, tmp_dir, hard_timeout, grace_period)
maruel064c0a32016-04-05 11:47:15 -0700436 finally:
437 result['duration'] = max(time.time() - start, 0)
maruela9cfd6f2015-09-15 11:03:15 -0700438 except Exception as e:
nodir90bc8dc2016-06-15 13:35:21 -0700439 # An internal error occurred. Report accordingly so the swarming task will
440 # be retried automatically.
maruel12e30012015-10-09 11:55:35 -0700441 logging.exception('internal failure: %s', e)
maruela9cfd6f2015-09-15 11:03:15 -0700442 result['internal_failure'] = str(e)
443 on_error.report(None)
444 finally:
445 try:
446 if leak_temp_dir:
447 logging.warning(
448 'Deliberately leaking %s for later examination', run_dir)
marueleb5fbee2015-09-17 13:01:36 -0700449 else:
maruel84537cb2015-10-16 14:21:28 -0700450 # On Windows rmtree(run_dir) call above has a synchronization effect: it
451 # finishes only when all task child processes terminate (since a running
452 # process locks *.exe file). Examine out_dir only after that call
453 # completes (since child processes may write to out_dir too and we need
454 # to wait for them to finish).
455 if fs.isdir(run_dir):
456 try:
457 success = file_path.rmtree(run_dir)
458 except OSError as e:
459 logging.error('Failure with %s', e)
460 success = False
461 if not success:
462 print >> sys.stderr, (
463 'Failed to delete the run directory, forcibly failing\n'
464 'the task because of it. No zombie process can outlive a\n'
465 'successful task run and still be marked as successful.\n'
466 'Fix your stuff.')
467 if result['exit_code'] == 0:
468 result['exit_code'] = 1
469 if fs.isdir(tmp_dir):
470 try:
471 success = file_path.rmtree(tmp_dir)
472 except OSError as e:
473 logging.error('Failure with %s', e)
474 success = False
475 if not success:
476 print >> sys.stderr, (
477 'Failed to delete the temporary directory, forcibly failing\n'
478 'the task because of it. No zombie process can outlive a\n'
479 'successful task run and still be marked as successful.\n'
480 'Fix your stuff.')
481 if result['exit_code'] == 0:
482 result['exit_code'] = 1
maruela9cfd6f2015-09-15 11:03:15 -0700483
marueleb5fbee2015-09-17 13:01:36 -0700484 # This deletes out_dir if leak_temp_dir is not set.
nodir9130f072016-05-27 13:59:08 -0700485 if out_dir:
nodir55715712016-06-03 12:28:19 -0700486 isolated_stats = result['stats'].setdefault('isolated', {})
487 result['outputs_ref'], success, isolated_stats['upload'] = (
nodir9130f072016-05-27 13:59:08 -0700488 delete_and_upload(storage, out_dir, leak_temp_dir))
maruela9cfd6f2015-09-15 11:03:15 -0700489 if not success and result['exit_code'] == 0:
490 result['exit_code'] = 1
491 except Exception as e:
492 # Swallow any exception in the main finally clause.
nodir9130f072016-05-27 13:59:08 -0700493 if out_dir:
494 logging.exception('Leaking out_dir %s: %s', out_dir, e)
maruela9cfd6f2015-09-15 11:03:15 -0700495 result['internal_failure'] = str(e)
496 return result
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -0500497
498
Marc-Antoine Ruel0ec868b2015-08-12 14:12:46 -0400499def run_tha_test(
nodir55be77b2016-05-03 09:39:57 -0700500 command, isolated_hash, storage, cache, leak_temp_dir, result_json,
bpastene3ae09522016-06-10 17:12:59 -0700501 root_dir, hard_timeout, grace_period, bot_file, extra_args,
nodir90bc8dc2016-06-15 13:35:21 -0700502 install_packages_fn):
nodir55be77b2016-05-03 09:39:57 -0700503 """Runs an executable and records execution metadata.
504
505 Either command or isolated_hash must be specified.
506
507 If isolated_hash is specified, downloads the dependencies in the cache,
508 hardlinks them into a temporary directory and runs the command specified in
509 the .isolated.
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -0500510
511 A temporary directory is created to hold the output files. The content inside
512 this directory will be uploaded back to |storage| packaged as a .isolated
513 file.
514
515 Arguments:
nodir55be77b2016-05-03 09:39:57 -0700516 command: the command to run, a list of strings. Mutually exclusive with
517 isolated_hash.
Marc-Antoine Ruel35b58432014-12-08 17:40:40 -0500518 isolated_hash: the SHA-1 of the .isolated file that must be retrieved to
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -0500519 recreate the tree of files to run the target executable.
nodir55be77b2016-05-03 09:39:57 -0700520 The command specified in the .isolated is executed.
521 Mutually exclusive with command argument.
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -0500522 storage: an isolateserver.Storage object to retrieve remote objects. This
523 object has a reference to an isolateserver.StorageApi, which does
524 the actual I/O.
525 cache: an isolateserver.LocalCache to keep from retrieving the same objects
526 constantly by caching the objects retrieved. Can be on-disk or
527 in-memory.
Kenneth Russell61d42352014-09-15 11:41:16 -0700528 leak_temp_dir: if true, the temporary directory will be deliberately leaked
529 for later examination.
maruela9cfd6f2015-09-15 11:03:15 -0700530 result_json: file path to dump result metadata into. If set, the process
nodirbe642ff2016-06-09 15:51:51 -0700531 exit code is always 0 unless an internal error occurred.
nodir90bc8dc2016-06-15 13:35:21 -0700532 root_dir: path to the directory to use to create the temporary directory. If
marueleb5fbee2015-09-17 13:01:36 -0700533 not specified, a random temporary directory is created.
maruel6be7f9e2015-10-01 12:25:30 -0700534 hard_timeout: kills the process if it lasts more than this amount of
535 seconds.
536 grace_period: number of seconds to wait between SIGTERM and SIGKILL.
Marc-Antoine Ruel2283ad12014-02-09 11:14:57 -0500537 extra_args: optional arguments to add to the command stated in the .isolate
nodir55be77b2016-05-03 09:39:57 -0700538 file. Ignored if isolate_hash is empty.
nodir90bc8dc2016-06-15 13:35:21 -0700539 install_packages_fn: function (dir) => cipd_stats. Installs packages.
maruela9cfd6f2015-09-15 11:03:15 -0700540
541 Returns:
542 Process exit code that should be used.
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000543 """
nodir55be77b2016-05-03 09:39:57 -0700544 assert bool(command) ^ bool(isolated_hash)
545 extra_args = extra_args or []
nodirbe642ff2016-06-09 15:51:51 -0700546
nodir55be77b2016-05-03 09:39:57 -0700547 if any(ISOLATED_OUTDIR_PARAMETER in a for a in (command or extra_args)):
548 assert storage is not None, 'storage is None although outdir is specified'
549
maruela76b9ee2015-12-15 06:18:08 -0800550 if result_json:
551 # Write a json output file right away in case we get killed.
552 result = {
553 'exit_code': None,
554 'had_hard_timeout': False,
555 'internal_failure': 'Was terminated before completion',
556 'outputs_ref': None,
nodirbe642ff2016-06-09 15:51:51 -0700557 'version': 5,
maruela76b9ee2015-12-15 06:18:08 -0800558 }
559 tools.write_json(result_json, result, dense=True)
560
maruela9cfd6f2015-09-15 11:03:15 -0700561 # run_isolated exit code. Depends on if result_json is used or not.
562 result = map_and_run(
nodir55be77b2016-05-03 09:39:57 -0700563 command, isolated_hash, storage, cache, leak_temp_dir, root_dir,
nodir90bc8dc2016-06-15 13:35:21 -0700564 hard_timeout, grace_period, bot_file, extra_args, install_packages_fn)
maruela9cfd6f2015-09-15 11:03:15 -0700565 logging.info('Result:\n%s', tools.format_json(result, dense=True))
bpastene3ae09522016-06-10 17:12:59 -0700566
maruela9cfd6f2015-09-15 11:03:15 -0700567 if result_json:
maruel05d5a882015-09-21 13:59:02 -0700568 # We've found tests to delete 'work' when quitting, causing an exception
569 # here. Try to recreate the directory if necessary.
nodire5028a92016-04-29 14:38:21 -0700570 file_path.ensure_tree(os.path.dirname(result_json))
maruela9cfd6f2015-09-15 11:03:15 -0700571 tools.write_json(result_json, result, dense=True)
572 # Only return 1 if there was an internal error.
573 return int(bool(result['internal_failure']))
maruel@chromium.org781ccf62013-09-17 19:39:47 +0000574
maruela9cfd6f2015-09-15 11:03:15 -0700575 # Marshall into old-style inline output.
576 if result['outputs_ref']:
577 data = {
578 'hash': result['outputs_ref']['isolated'],
579 'namespace': result['outputs_ref']['namespace'],
580 'storage': result['outputs_ref']['isolatedserver'],
581 }
Marc-Antoine Ruelc44f5722015-01-08 16:10:01 -0500582 sys.stdout.flush()
maruela9cfd6f2015-09-15 11:03:15 -0700583 print(
584 '[run_isolated_out_hack]%s[/run_isolated_out_hack]' %
585 tools.format_json(data, dense=True))
maruelb76604c2015-11-11 11:53:44 -0800586 sys.stdout.flush()
maruela9cfd6f2015-09-15 11:03:15 -0700587 return result['exit_code'] or int(bool(result['internal_failure']))
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000588
589
nodir90bc8dc2016-06-15 13:35:21 -0700590def install_packages(
nodirff531b42016-06-23 13:05:06 -0700591 run_dir, packages, service_url, client_package_name,
nodir90bc8dc2016-06-15 13:35:21 -0700592 client_version, cache_dir=None, timeout=None):
593 """Installs packages. Returns stats.
nodirbe642ff2016-06-09 15:51:51 -0700594
595 Args:
nodir90bc8dc2016-06-15 13:35:21 -0700596 run_dir (str): root of installation.
nodirff531b42016-06-23 13:05:06 -0700597 packages: packages to install, dict {path: [(package_name, version)].
nodirbe642ff2016-06-09 15:51:51 -0700598 service_url (str): CIPD server url, e.g.
599 "https://chrome-infra-packages.appspot.com."
nodir90bc8dc2016-06-15 13:35:21 -0700600 client_package_name (str): CIPD package name of CIPD client.
601 client_version (str): Version of CIPD client.
nodirbe642ff2016-06-09 15:51:51 -0700602 cache_dir (str): where to keep cache of cipd clients, packages and tags.
603 timeout: max duration in seconds that this function can take.
nodirbe642ff2016-06-09 15:51:51 -0700604 """
605 assert cache_dir
nodirff531b42016-06-23 13:05:06 -0700606 if not packages:
nodir90bc8dc2016-06-15 13:35:21 -0700607 return None
608
nodirbe642ff2016-06-09 15:51:51 -0700609 timeoutfn = tools.sliding_timeout(timeout)
nodirbe642ff2016-06-09 15:51:51 -0700610 start = time.time()
nodirbe642ff2016-06-09 15:51:51 -0700611 cache_dir = os.path.abspath(cache_dir)
612
nodir90bc8dc2016-06-15 13:35:21 -0700613 run_dir = os.path.abspath(run_dir)
nodir90bc8dc2016-06-15 13:35:21 -0700614
nodirbe642ff2016-06-09 15:51:51 -0700615 get_client_start = time.time()
616 client_manager = cipd.get_client(
617 service_url, client_package_name, client_version, cache_dir,
618 timeout=timeoutfn())
619 with client_manager as client:
620 get_client_duration = time.time() - get_client_start
nodirff531b42016-06-23 13:05:06 -0700621 for path, packages in sorted(packages.iteritems()):
nodir90bc8dc2016-06-15 13:35:21 -0700622 site_root = os.path.abspath(os.path.join(run_dir, path))
623 if not site_root.startswith(run_dir):
624 raise cipd.Error('Invalid CIPD package path "%s"' % path)
625
626 # Do not clean site_root before installation because it may contain other
627 # site roots.
628 file_path.ensure_tree(site_root, 0770)
nodirbe642ff2016-06-09 15:51:51 -0700629 client.ensure(
630 site_root, packages,
631 cache_dir=os.path.join(cache_dir, 'cipd_internal'),
632 timeout=timeoutfn())
nodirbe642ff2016-06-09 15:51:51 -0700633 file_path.make_tree_files_read_only(site_root)
nodir90bc8dc2016-06-15 13:35:21 -0700634
635 total_duration = time.time() - start
636 logging.info(
637 'Installing CIPD client and packages took %d seconds', total_duration)
638
639 return {
640 'duration': total_duration,
641 'get_client_duration': get_client_duration,
642 }
nodirbe642ff2016-06-09 15:51:51 -0700643
644
645def create_option_parser():
Marc-Antoine Ruelf74cffe2015-07-15 15:21:34 -0400646 parser = logging_utils.OptionParserWithLogging(
nodir55be77b2016-05-03 09:39:57 -0700647 usage='%prog <options> [command to run or extra args]',
maruel@chromium.orgdedbf492013-09-12 20:42:11 +0000648 version=__version__,
649 log_file=RUN_ISOLATED_LOG_FILE)
maruela9cfd6f2015-09-15 11:03:15 -0700650 parser.add_option(
maruel36a963d2016-04-08 17:15:49 -0700651 '--clean', action='store_true',
652 help='Cleans the cache, trimming it necessary and remove corrupted items '
653 'and returns without executing anything; use with -v to know what '
654 'was done')
655 parser.add_option(
maruela9cfd6f2015-09-15 11:03:15 -0700656 '--json',
657 help='dump output metadata to json file. When used, run_isolated returns '
658 'non-zero only on internal failure')
maruel6be7f9e2015-10-01 12:25:30 -0700659 parser.add_option(
maruel5c9e47b2015-12-18 13:02:30 -0800660 '--hard-timeout', type='float', help='Enforce hard timeout in execution')
maruel6be7f9e2015-10-01 12:25:30 -0700661 parser.add_option(
maruel5c9e47b2015-12-18 13:02:30 -0800662 '--grace-period', type='float',
maruel6be7f9e2015-10-01 12:25:30 -0700663 help='Grace period between SIGTERM and SIGKILL')
bpastene3ae09522016-06-10 17:12:59 -0700664 parser.add_option(
665 '--bot-file',
666 help='Path to a file describing the state of the host. The content is '
667 'defined by on_before_task() in bot_config.')
Marc-Antoine Ruel1687b5e2014-02-06 17:47:53 -0500668 data_group = optparse.OptionGroup(parser, 'Data source')
669 data_group.add_option(
Marc-Antoine Ruel185ded42015-01-28 20:49:18 -0500670 '-s', '--isolated',
nodir55be77b2016-05-03 09:39:57 -0700671 help='Hash of the .isolated to grab from the isolate server.')
Marc-Antoine Ruelf7d737d2014-12-10 15:36:29 -0500672 isolateserver.add_isolate_server_options(data_group)
Marc-Antoine Ruel1687b5e2014-02-06 17:47:53 -0500673 parser.add_option_group(data_group)
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000674
Marc-Antoine Ruela57d7db2014-10-15 20:31:19 -0400675 isolateserver.add_cache_options(parser)
nodirbe642ff2016-06-09 15:51:51 -0700676
677 cipd.add_cipd_options(parser)
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000678
Kenneth Russell61d42352014-09-15 11:41:16 -0700679 debug_group = optparse.OptionGroup(parser, 'Debugging')
680 debug_group.add_option(
681 '--leak-temp-dir',
682 action='store_true',
nodirbe642ff2016-06-09 15:51:51 -0700683 help='Deliberately leak isolate\'s temp dir for later examination. '
684 'Default: %default')
marueleb5fbee2015-09-17 13:01:36 -0700685 debug_group.add_option(
686 '--root-dir', help='Use a directory instead of a random one')
Kenneth Russell61d42352014-09-15 11:41:16 -0700687 parser.add_option_group(debug_group)
688
Vadim Shtayurae34e13a2014-02-02 11:23:26 -0800689 auth.add_auth_options(parser)
nodirbe642ff2016-06-09 15:51:51 -0700690
691 parser.set_defaults(cache='cache', cipd_cache='cipd_cache')
692 return parser
693
694
695def main(args):
696 parser = create_option_parser()
Marc-Antoine Ruel90c98162013-12-18 15:11:57 -0500697 options, args = parser.parse_args(args)
maruel36a963d2016-04-08 17:15:49 -0700698
699 cache = isolateserver.process_cache_options(options)
700 if options.clean:
701 if options.isolated:
702 parser.error('Can\'t use --isolated with --clean.')
703 if options.isolate_server:
704 parser.error('Can\'t use --isolate-server with --clean.')
705 if options.json:
706 parser.error('Can\'t use --json with --clean.')
707 cache.cleanup()
708 return 0
709
nodir55be77b2016-05-03 09:39:57 -0700710 if not options.isolated and not args:
711 parser.error('--isolated or command to run is required.')
712
Vadim Shtayura5d1efce2014-02-04 10:55:43 -0800713 auth.process_auth_options(parser, options)
nodir55be77b2016-05-03 09:39:57 -0700714
715 isolateserver.process_isolate_server_options(
716 parser, options, True, False)
717 if not options.isolate_server:
718 if options.isolated:
719 parser.error('--isolated requires --isolate-server')
720 if ISOLATED_OUTDIR_PARAMETER in args:
721 parser.error(
722 '%s in args requires --isolate-server' % ISOLATED_OUTDIR_PARAMETER)
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000723
nodir90bc8dc2016-06-15 13:35:21 -0700724 if options.root_dir:
725 options.root_dir = unicode(os.path.abspath(options.root_dir))
maruel12e30012015-10-09 11:55:35 -0700726 if options.json:
727 options.json = unicode(os.path.abspath(options.json))
nodir55be77b2016-05-03 09:39:57 -0700728
nodirbe642ff2016-06-09 15:51:51 -0700729 cipd.validate_cipd_options(parser, options)
730
nodir90bc8dc2016-06-15 13:35:21 -0700731 install_packages_fn = lambda run_dir: install_packages(
nodirff531b42016-06-23 13:05:06 -0700732 run_dir, cipd.parse_package_args(options.cipd_packages),
733 options.cipd_server, options.cipd_client_package,
734 options.cipd_client_version, cache_dir=options.cipd_cache)
nodirbe642ff2016-06-09 15:51:51 -0700735
736 try:
nodir90bc8dc2016-06-15 13:35:21 -0700737 command = [] if options.isolated else args
738 if options.isolate_server:
739 storage = isolateserver.get_storage(
740 options.isolate_server, options.namespace)
741 with storage:
742 # Hashing schemes used by |storage| and |cache| MUST match.
743 assert storage.hash_algo == cache.hash_algo
nodirbe642ff2016-06-09 15:51:51 -0700744 return run_tha_test(
nodir90bc8dc2016-06-15 13:35:21 -0700745 command, options.isolated, storage, cache, options.leak_temp_dir,
746 options.json, options.root_dir, options.hard_timeout,
747 options.grace_period, options.bot_file, args, install_packages_fn)
748 else:
749 return run_tha_test(
750 command, options.isolated, None, cache, options.leak_temp_dir,
751 options.json, options.root_dir, options.hard_timeout,
752 options.grace_period, options.bot_file, args, install_packages_fn)
nodirbe642ff2016-06-09 15:51:51 -0700753 except cipd.Error as ex:
754 print >> sys.stderr, ex.message
755 return 1
maruel@chromium.org9c72d4e2012-09-28 19:20:25 +0000756
757
758if __name__ == '__main__':
maruel8e4e40c2016-05-30 06:21:07 -0700759 subprocess42.inhibit_os_error_reporting()
csharp@chromium.orgbfb98742013-03-26 20:28:36 +0000760 # Ensure that we are always running with the correct encoding.
vadimsh@chromium.orga4326472013-08-24 02:05:41 +0000761 fix_encoding.fix_encoding()
Marc-Antoine Ruel90c98162013-12-18 15:11:57 -0500762 sys.exit(main(sys.argv[1:]))