OLD | NEW |
(Empty) | |
| 1 # Copyright 2013 The Chromium Authors. All rights reserved. |
| 2 # Use of this source code is governed by a BSD-style license that can be |
| 3 # found in the LICENSE file. |
| 4 |
| 5 """Dispatches tests, either sharding or replicating them. |
| 6 |
| 7 Performs the following steps: |
| 8 * Create a test collection factory, using the given tests |
| 9 - If sharding: test collection factory returns the same shared test collection |
| 10 to all test runners |
| 11 - If replciating: test collection factory returns a unique test collection to |
| 12 each test runner, with the same set of tests in each. |
| 13 * Create a test runner for each device. |
| 14 * Run each test runner in its own thread, grabbing tests from the test |
| 15 collection until there are no tests left. |
| 16 """ |
| 17 |
| 18 # TODO(jbudorick) Deprecate and remove this class after any relevant parts have |
| 19 # been ported to the new environment / test instance model. |
| 20 |
| 21 import logging |
| 22 import threading |
| 23 |
| 24 from pylib import constants |
| 25 from pylib.base import base_test_result |
| 26 from pylib.base import test_collection |
| 27 from pylib.device import device_errors |
| 28 from pylib.utils import reraiser_thread |
| 29 from pylib.utils import watchdog_timer |
| 30 |
| 31 |
| 32 DEFAULT_TIMEOUT = 7 * 60 # seven minutes |
| 33 |
| 34 |
| 35 class _ThreadSafeCounter(object): |
| 36 """A threadsafe counter.""" |
| 37 |
| 38 def __init__(self): |
| 39 self._lock = threading.Lock() |
| 40 self._value = 0 |
| 41 |
| 42 def GetAndIncrement(self): |
| 43 """Get the current value and increment it atomically. |
| 44 |
| 45 Returns: |
| 46 The value before incrementing. |
| 47 """ |
| 48 with self._lock: |
| 49 pre_increment = self._value |
| 50 self._value += 1 |
| 51 return pre_increment |
| 52 |
| 53 |
| 54 class _Test(object): |
| 55 """Holds a test with additional metadata.""" |
| 56 |
| 57 def __init__(self, test, tries=0): |
| 58 """Initializes the _Test object. |
| 59 |
| 60 Args: |
| 61 test: The test. |
| 62 tries: Number of tries so far. |
| 63 """ |
| 64 self.test = test |
| 65 self.tries = tries |
| 66 |
| 67 |
| 68 def _RunTestsFromQueue(runner, collection, out_results, watcher, |
| 69 num_retries, tag_results_with_device=False): |
| 70 """Runs tests from the collection until empty using the given runner. |
| 71 |
| 72 Adds TestRunResults objects to the out_results list and may add tests to the |
| 73 out_retry list. |
| 74 |
| 75 Args: |
| 76 runner: A TestRunner object used to run the tests. |
| 77 collection: A TestCollection from which to get _Test objects to run. |
| 78 out_results: A list to add TestRunResults to. |
| 79 watcher: A watchdog_timer.WatchdogTimer object, used as a shared timeout. |
| 80 num_retries: Number of retries for a test. |
| 81 tag_results_with_device: If True, appends the name of the device on which |
| 82 the test was run to the test name. Used when replicating to identify |
| 83 which device ran each copy of the test, and to ensure each copy of the |
| 84 test is recorded separately. |
| 85 """ |
| 86 |
| 87 def TagTestRunResults(test_run_results): |
| 88 """Tags all results with the last 4 digits of the device id. |
| 89 |
| 90 Used when replicating tests to distinguish the same tests run on different |
| 91 devices. We use a set to store test results, so the hash (generated from |
| 92 name and tag) must be unique to be considered different results. |
| 93 """ |
| 94 new_test_run_results = base_test_result.TestRunResults() |
| 95 for test_result in test_run_results.GetAll(): |
| 96 test_result.SetName('%s_%s' % (runner.device_serial[-4:], |
| 97 test_result.GetName())) |
| 98 new_test_run_results.AddResult(test_result) |
| 99 return new_test_run_results |
| 100 |
| 101 for test in collection: |
| 102 watcher.Reset() |
| 103 try: |
| 104 if not runner.device.IsOnline(): |
| 105 # Device is unresponsive, stop handling tests on this device. |
| 106 msg = 'Device %s is unresponsive.' % runner.device_serial |
| 107 logging.warning(msg) |
| 108 raise device_errors.DeviceUnreachableError(msg) |
| 109 result, retry = runner.RunTest(test.test) |
| 110 if tag_results_with_device: |
| 111 result = TagTestRunResults(result) |
| 112 test.tries += 1 |
| 113 if retry and test.tries <= num_retries: |
| 114 # Retry non-passing results, only record passing results. |
| 115 pass_results = base_test_result.TestRunResults() |
| 116 pass_results.AddResults(result.GetPass()) |
| 117 out_results.append(pass_results) |
| 118 logging.warning('Will retry test %s, try #%s.', retry, test.tries) |
| 119 collection.add(_Test(test=retry, tries=test.tries)) |
| 120 else: |
| 121 # All tests passed or retry limit reached. Either way, record results. |
| 122 out_results.append(result) |
| 123 except: |
| 124 # An unhandleable exception, ensure tests get run by another device and |
| 125 # reraise this exception on the main thread. |
| 126 collection.add(test) |
| 127 raise |
| 128 finally: |
| 129 # Retries count as separate tasks so always mark the popped test as done. |
| 130 collection.test_completed() |
| 131 |
| 132 |
| 133 def _SetUp(runner_factory, device, out_runners, threadsafe_counter): |
| 134 """Creates a test runner for each device and calls SetUp() in parallel. |
| 135 |
| 136 Note: if a device is unresponsive the corresponding TestRunner will not be |
| 137 added to out_runners. |
| 138 |
| 139 Args: |
| 140 runner_factory: Callable that takes a device and index and returns a |
| 141 TestRunner object. |
| 142 device: The device serial number to set up. |
| 143 out_runners: List to add the successfully set up TestRunner object. |
| 144 threadsafe_counter: A _ThreadSafeCounter object used to get shard indices. |
| 145 """ |
| 146 try: |
| 147 index = threadsafe_counter.GetAndIncrement() |
| 148 logging.warning('Creating shard %s for device %s.', index, device) |
| 149 runner = runner_factory(device, index) |
| 150 runner.SetUp() |
| 151 out_runners.append(runner) |
| 152 except device_errors.DeviceUnreachableError as e: |
| 153 logging.warning('Failed to create shard for %s: [%s]', device, e) |
| 154 |
| 155 |
| 156 def _RunAllTests(runners, test_collection_factory, num_retries, timeout=None, |
| 157 tag_results_with_device=False): |
| 158 """Run all tests using the given TestRunners. |
| 159 |
| 160 Args: |
| 161 runners: A list of TestRunner objects. |
| 162 test_collection_factory: A callable to generate a TestCollection object for |
| 163 each test runner. |
| 164 num_retries: Number of retries for a test. |
| 165 timeout: Watchdog timeout in seconds. |
| 166 tag_results_with_device: If True, appends the name of the device on which |
| 167 the test was run to the test name. Used when replicating to identify |
| 168 which device ran each copy of the test, and to ensure each copy of the |
| 169 test is recorded separately. |
| 170 |
| 171 Returns: |
| 172 A tuple of (TestRunResults object, exit code) |
| 173 """ |
| 174 logging.warning('Running tests with %s test runners.' % (len(runners))) |
| 175 results = [] |
| 176 exit_code = 0 |
| 177 run_results = base_test_result.TestRunResults() |
| 178 watcher = watchdog_timer.WatchdogTimer(timeout) |
| 179 test_collections = [test_collection_factory() for _ in runners] |
| 180 |
| 181 threads = [ |
| 182 reraiser_thread.ReraiserThread( |
| 183 _RunTestsFromQueue, |
| 184 [r, tc, results, watcher, num_retries, tag_results_with_device], |
| 185 name=r.device_serial[-4:]) |
| 186 for r, tc in zip(runners, test_collections)] |
| 187 |
| 188 workers = reraiser_thread.ReraiserThreadGroup(threads) |
| 189 workers.StartAll() |
| 190 |
| 191 # Catch DeviceUnreachableErrors and set a warning exit code |
| 192 try: |
| 193 workers.JoinAll(watcher) |
| 194 except device_errors.DeviceUnreachableError as e: |
| 195 logging.error(e) |
| 196 |
| 197 if not all((len(tc) == 0 for tc in test_collections)): |
| 198 logging.error('Only ran %d tests (all devices are likely offline).' % |
| 199 len(results)) |
| 200 for tc in test_collections: |
| 201 run_results.AddResults(base_test_result.BaseTestResult( |
| 202 t, base_test_result.ResultType.UNKNOWN) for t in tc.test_names()) |
| 203 |
| 204 for r in results: |
| 205 run_results.AddTestRunResults(r) |
| 206 if not run_results.DidRunPass(): |
| 207 exit_code = constants.ERROR_EXIT_CODE |
| 208 return (run_results, exit_code) |
| 209 |
| 210 |
| 211 def _CreateRunners(runner_factory, devices, timeout=None): |
| 212 """Creates a test runner for each device and calls SetUp() in parallel. |
| 213 |
| 214 Note: if a device is unresponsive the corresponding TestRunner will not be |
| 215 included in the returned list. |
| 216 |
| 217 Args: |
| 218 runner_factory: Callable that takes a device and index and returns a |
| 219 TestRunner object. |
| 220 devices: List of device serial numbers as strings. |
| 221 timeout: Watchdog timeout in seconds, defaults to the default timeout. |
| 222 |
| 223 Returns: |
| 224 A list of TestRunner objects. |
| 225 """ |
| 226 logging.warning('Creating %s test runners.' % len(devices)) |
| 227 runners = [] |
| 228 counter = _ThreadSafeCounter() |
| 229 threads = reraiser_thread.ReraiserThreadGroup( |
| 230 [reraiser_thread.ReraiserThread(_SetUp, |
| 231 [runner_factory, d, runners, counter], |
| 232 name=str(d)[-4:]) |
| 233 for d in devices]) |
| 234 threads.StartAll() |
| 235 threads.JoinAll(watchdog_timer.WatchdogTimer(timeout)) |
| 236 return runners |
| 237 |
| 238 |
| 239 def _TearDownRunners(runners, timeout=None): |
| 240 """Calls TearDown() for each test runner in parallel. |
| 241 |
| 242 Args: |
| 243 runners: A list of TestRunner objects. |
| 244 timeout: Watchdog timeout in seconds, defaults to the default timeout. |
| 245 """ |
| 246 threads = reraiser_thread.ReraiserThreadGroup( |
| 247 [reraiser_thread.ReraiserThread(r.TearDown, name=r.device_serial[-4:]) |
| 248 for r in runners]) |
| 249 threads.StartAll() |
| 250 threads.JoinAll(watchdog_timer.WatchdogTimer(timeout)) |
| 251 |
| 252 |
| 253 def ApplyMaxPerRun(tests, max_per_run): |
| 254 """Rearrange the tests so that no group contains more than max_per_run tests. |
| 255 |
| 256 Args: |
| 257 tests: |
| 258 max_per_run: |
| 259 |
| 260 Returns: |
| 261 A list of tests with no more than max_per_run per run. |
| 262 """ |
| 263 tests_expanded = [] |
| 264 for test_group in tests: |
| 265 if type(test_group) != str: |
| 266 # Do not split test objects which are not strings. |
| 267 tests_expanded.append(test_group) |
| 268 else: |
| 269 test_split = test_group.split(':') |
| 270 for i in range(0, len(test_split), max_per_run): |
| 271 tests_expanded.append(':'.join(test_split[i:i+max_per_run])) |
| 272 return tests_expanded |
| 273 |
| 274 |
| 275 def RunTests(tests, runner_factory, devices, shard=True, |
| 276 test_timeout=DEFAULT_TIMEOUT, setup_timeout=DEFAULT_TIMEOUT, |
| 277 num_retries=2, max_per_run=256): |
| 278 """Run all tests on attached devices, retrying tests that don't pass. |
| 279 |
| 280 Args: |
| 281 tests: List of tests to run. |
| 282 runner_factory: Callable that takes a device and index and returns a |
| 283 TestRunner object. |
| 284 devices: List of attached devices. |
| 285 shard: True if we should shard, False if we should replicate tests. |
| 286 - Sharding tests will distribute tests across all test runners through a |
| 287 shared test collection. |
| 288 - Replicating tests will copy all tests to each test runner through a |
| 289 unique test collection for each test runner. |
| 290 test_timeout: Watchdog timeout in seconds for running tests. |
| 291 setup_timeout: Watchdog timeout in seconds for creating and cleaning up |
| 292 test runners. |
| 293 num_retries: Number of retries for a test. |
| 294 max_per_run: Maximum number of tests to run in any group. |
| 295 |
| 296 Returns: |
| 297 A tuple of (base_test_result.TestRunResults object, exit code). |
| 298 """ |
| 299 if not tests: |
| 300 logging.critical('No tests to run.') |
| 301 return (base_test_result.TestRunResults(), constants.ERROR_EXIT_CODE) |
| 302 |
| 303 tests_expanded = ApplyMaxPerRun(tests, max_per_run) |
| 304 if shard: |
| 305 # Generate a shared TestCollection object for all test runners, so they |
| 306 # draw from a common pool of tests. |
| 307 shared_test_collection = test_collection.TestCollection( |
| 308 [_Test(t) for t in tests_expanded]) |
| 309 test_collection_factory = lambda: shared_test_collection |
| 310 tag_results_with_device = False |
| 311 log_string = 'sharded across devices' |
| 312 else: |
| 313 # Generate a unique TestCollection object for each test runner, but use |
| 314 # the same set of tests. |
| 315 test_collection_factory = lambda: test_collection.TestCollection( |
| 316 [_Test(t) for t in tests_expanded]) |
| 317 tag_results_with_device = True |
| 318 log_string = 'replicated on each device' |
| 319 |
| 320 logging.info('Will run %d tests (%s): %s', |
| 321 len(tests_expanded), log_string, str(tests_expanded)) |
| 322 runners = _CreateRunners(runner_factory, devices, setup_timeout) |
| 323 try: |
| 324 return _RunAllTests(runners, test_collection_factory, |
| 325 num_retries, test_timeout, tag_results_with_device) |
| 326 finally: |
| 327 try: |
| 328 _TearDownRunners(runners, setup_timeout) |
| 329 except device_errors.DeviceUnreachableError as e: |
| 330 logging.warning('Device unresponsive during TearDown: [%s]', e) |
| 331 except Exception as e: |
| 332 logging.error('Unexpected exception caught during TearDown: %s' % str(e)) |
OLD | NEW |