|
""" Overview ========
The multiprocess plugin enables you to distribute your test run among a set of worker processes that run tests in parallel. This can speed up CPU-bound test runs (as long as the number of work processeses is around the number of processors or cores available), but is mainly useful for IO-bound tests that spend most of their time waiting for data to arrive from someplace else.
.. note ::
See :doc:`../doc_tests/test_multiprocess/multiprocess` for additional documentation and examples. Use of this plugin on python 2.5 or earlier requires the multiprocessing_ module, also available from PyPI.
.. _multiprocessing : http://code.google.com/p/python-multiprocessing/
How tests are distributed =========================
The ideal case would be to dispatch each test to a worker process separately. This ideal is not attainable in all cases, however, because many test suites depend on context (class, module or package) fixtures.
The plugin can't know (unless you tell it -- see below!) if a context fixture can be called many times concurrently (is re-entrant), or if it can be shared among tests running in different processes. Therefore, if a context has fixtures, the default behavior is to dispatch the entire suite to a worker as a unit.
Controlling distribution ^^^^^^^^^^^^^^^^^^^^^^^^
There are two context-level variables that you can use to control this default behavior.
If a context's fixtures are re-entrant, set ``_multiprocess_can_split_ = True`` in the context, and the plugin will dispatch tests in suites bound to that context as if the context had no fixtures. This means that the fixtures will execute concurrently and multiple times, typically once per test.
If a context's fixtures can be shared by tests running in different processes -- such as a package-level fixture that starts an external http server or initializes a shared database -- then set ``_multiprocess_shared_ = True`` in the context. These fixtures will then execute in the primary nose process, and tests in those contexts will be individually dispatched to run in parallel.
How results are collected and reported ======================================
As each test or suite executes in a worker process, results (failures, errors, and specially handled exceptions like SkipTest) are collected in that process. When the worker process finishes, it returns results to the main nose process. There, any progress output is printed (dots!), and the results from the test run are combined into a consolidated result set. When results have been received for all dispatched tests, or all workers have died, the result summary is output as normal.
Beware! =======
Not all test suites will benefit from, or even operate correctly using, this plugin. For example, CPU-bound tests will run more slowly if you don't have multiple processors. There are also some differences in plugin interactions and behaviors due to the way in which tests are dispatched and loaded. In general, test loading under this plugin operates as if it were always in directed mode instead of discovered mode. For instance, doctests in test modules will always be found when using this plugin with the doctest plugin.
But the biggest issue you will face is probably concurrency. Unless you have kept your tests as religiously pure unit tests, with no side-effects, no ordering issues, and no external dependencies, chances are you will experience odd, intermittent and unexplainable failures and errors when using this plugin. This doesn't necessarily mean the plugin is broken; it may mean that your test suite is not safe for concurrency.
New Features in 1.1.0 =====================
* functions generated by test generators are now added to the worker queue making them multi-threaded. * fixed timeout functionality, now functions will be terminated with a TimedOutException exception when they exceed their execution time. The worker processes are not terminated. * added ``--process-restartworker`` option to restart workers once they are done, this helps control memory usage. Sometimes memory leaks can accumulate making long runs very difficult. * added global _instantiate_plugins to configure which plugins are started on the worker processes.
"""
# 2.7+ except ImportError: from unittest import _WritelnDecorator except ImportError: import io
# this is a list of plugin classes that will be checked for and created inside # each worker process
# have to inherit KeyboardInterrupt to it will interrupt process properly self.value = value def __str__(self): return repr(self.value)
global Process, Queue, Pool, Event, Value, Array try: from multiprocessing import Manager, Process #prevent the server process created in the manager which holds Python #objects and allows other processes to manipulate them using proxies #to interrupt on SIGINT (keyboardinterrupt) so that the communication #channel between subprocesses and main process is still usable after #ctrl+C is received in the main process. old=signal.signal(signal.SIGINT, signal.SIG_IGN) m = Manager() #reset it back so main process will receive a KeyboardInterrupt #exception on ctrl+c signal.signal(signal.SIGINT, old) Queue, Pool, Event, Value, Array = ( m.Queue, m.Pool, m.Event, m.Value, m.Array ) except ImportError: warn("multiprocessing module is not available, multiprocess plugin " "cannot be used", RuntimeWarning)
try: self._id = case.id() except AttributeError: pass self._short_description = case.shortDescription() self._str = str(case)
return self._id
return self._short_description
def __str__(self): return self._str
""" Run tests in multiple processes. Requires processing module. """
""" Register command-line options. """ default=env.get('NOSE_PROCESSES', 0), dest="multiprocess_workers", metavar="NUM", help="Spread test run among this many processes. " "Set a number equal to the number of processors " "or cores in your machine for best results. " "Pass a negative number to have the number of " "processes automatically set to the number of " "cores. Passing 0 means to disable parallel " "testing. Default is 0 unless NOSE_PROCESSES is " "set. " "[NOSE_PROCESSES]") default=env.get('NOSE_PROCESS_TIMEOUT', 10), dest="multiprocess_timeout", metavar="SECONDS", help="Set timeout for return of results from each " "test runner process. Default is 10. " "[NOSE_PROCESS_TIMEOUT]") default=env.get('NOSE_PROCESS_RESTARTWORKER', False), dest="multiprocess_restartworker", help="If set, will restart each worker process once" " their tests are done, this helps control memory " "leaks from killing the system. " "[NOSE_PROCESS_RESTARTWORKER]")
""" Configure plugin. """ self.enabled = False return # don't start inside of a worker process return except (TypeError, ValueError): workers = 0 _import_mp() if Process is None: self.enabled = False return # Negative number of workers will cause multiprocessing to hang. # Set the number of workers to the CPU count to avoid this. if workers < 0: try: import multiprocessing workers = multiprocessing.cpu_count() except NotImplementedError: self.enabled = False return self.enabled = True self.config.multiprocess_workers = workers t = float(options.multiprocess_timeout) self.config.multiprocess_timeout = t r = int(options.multiprocess_restartworker) self.config.multiprocess_restartworker = r self.status['active'] = True
"""Remember loader class so MultiProcessTestRunner can instantiate the right loader. """ self.loaderClass = loader.__class__
"""Replace test runner with MultiProcessTestRunner. """ # replace with our runner class return MultiProcessTestRunner(stream=runner.stream, verbosity=self.config.verbosity, config=self.config, loaderClass=self.loaderClass)
raise TimedOutException()
# respond to SIGILL self.loaderClass = kw.pop('loaderClass', loader.defaultTestLoader) super(MultiProcessTestRunner, self).__init__(**kw)
# dispatch and collect results # put indexes only on queue because tests aren't picklable for case in self.nextBatch(test): log.debug("Next batch %s (%s)", case, type(case)) if (isinstance(case, nose.case.Test) and isinstance(case.test, failure.Failure)): log.debug("Case is a Failure") case(result) # run here to capture the failure continue # handle shared fixtures if isinstance(case, ContextSuite) and case.context is failure.Failure: log.debug("Case is a Failure") case(result) # run here to capture the failure continue elif isinstance(case, ContextSuite) and self.sharedFixtures(case): log.debug("%s has shared fixtures", case) try: case.setUp() except (KeyboardInterrupt, SystemExit): raise except: log.debug("%s setup failed", sys.exc_info()) result.addError(case, sys.exc_info()) else: to_teardown.append(case) if case.factory: ancestors=case.factory.context.get(case, []) for an in ancestors[:2]: #log.debug('reset ancestor %s', an) if getattr(an, '_multiprocess_shared_', False): an._multiprocess_can_split_=True #an._multiprocess_shared_=False self.collect(case, testQueue, tasks, to_teardown, result)
else: test_addr = self.addtask(testQueue,tasks,case) log.debug("Queued test %s (%s) to %s", len(tasks), test_addr, testQueue)
currentaddr = Value('c',bytes_('')) currentstart = Value('d',time.time()) keyboardCaught = Event() p = Process(target=runner, args=(iworker, testQueue, resultQueue, currentaddr, currentstart, keyboardCaught, shouldStop, self.loaderClass, result.__class__, pickle.dumps(self.config))) p.currentaddr = currentaddr p.currentstart = currentstart p.keyboardCaught = keyboardCaught old = signal.signal(signal.SIGILL, signalhandler) p.start() signal.signal(signal.SIGILL, old) return p
""" Execute the test (which may be a test suite). If the test is a suite, distribute it out among as many processes as have been configured, at as fine a level as is possible given the context fixtures defined in the suite or any sub-suites.
""" log.debug("%s.run(%s) (%s)", self, test, os.getpid()) wrapper = self.config.plugins.prepareTest(test) if wrapper is not None: test = wrapper
# plugins can decorate or capture the output stream wrapped = self.config.plugins.setOutputStream(self.stream) if wrapped is not None: self.stream = wrapped
testQueue = Queue() resultQueue = Queue() tasks = [] completed = [] workers = [] to_teardown = [] shouldStop = Event()
result = self._makeResult() start = time.time()
self.collect(test, testQueue, tasks, to_teardown, result)
log.debug("Starting %s workers", self.config.multiprocess_workers) for i in range(self.config.multiprocess_workers): p = self.startProcess(i, testQueue, resultQueue, shouldStop, result) workers.append(p) log.debug("Started worker process %s", i+1)
total_tasks = len(tasks) # need to keep track of the next time to check for timeouts in case # more than one process times out at the same time. nexttimeout=self.config.multiprocess_timeout thrownError = None
try: while tasks: log.debug("Waiting for results (%s/%s tasks), next timeout=%.3fs", len(completed), total_tasks,nexttimeout) try: iworker, addr, newtask_addrs, batch_result = resultQueue.get( timeout=nexttimeout) log.debug('Results received for worker %d, %s, new tasks: %d', iworker,addr,len(newtask_addrs)) try: try: tasks.remove(addr) except ValueError: log.warn('worker %s failed to remove from tasks: %s', iworker,addr) total_tasks += len(newtask_addrs) tasks.extend(newtask_addrs) except KeyError: log.debug("Got result for unknown task? %s", addr) log.debug("current: %s",str(list(tasks)[0])) else: completed.append([addr,batch_result]) self.consolidate(result, batch_result) if (self.config.stopOnError and not result.wasSuccessful()): # set the stop condition shouldStop.set() break if self.config.multiprocess_restartworker: log.debug('joining worker %s',iworker) # wait for working, but not that important if worker # cannot be joined in fact, for workers that add to # testQueue, they will not terminate until all their # items are read workers[iworker].join(timeout=1) if not shouldStop.is_set() and not testQueue.empty(): log.debug('starting new process on worker %s',iworker) workers[iworker] = self.startProcess(iworker, testQueue, resultQueue, shouldStop, result) except Empty: log.debug("Timed out with %s tasks pending " "(empty testQueue=%r): %s", len(tasks),testQueue.empty(),str(tasks)) any_alive = False for iworker, w in enumerate(workers): if w.is_alive(): worker_addr = bytes_(w.currentaddr.value,'ascii') timeprocessing = time.time() - w.currentstart.value if ( len(worker_addr) == 0 and timeprocessing > self.config.multiprocess_timeout-0.1): log.debug('worker %d has finished its work item, ' 'but is not exiting? do we wait for it?', iworker) else: any_alive = True if (len(worker_addr) > 0 and timeprocessing > self.config.multiprocess_timeout-0.1): log.debug('timed out worker %s: %s', iworker,worker_addr) w.currentaddr.value = bytes_('') # If the process is in C++ code, sending a SIGILL # might not send a python KeybordInterrupt exception # therefore, send multiple signals until an # exception is caught. If this takes too long, then # terminate the process w.keyboardCaught.clear() startkilltime = time.time() while not w.keyboardCaught.is_set() and w.is_alive(): if time.time()-startkilltime > self.waitkilltime: # have to terminate... log.error("terminating worker %s",iworker) w.terminate() # there is a small probability that the # terminated process might send a result, # which has to be specially handled or # else processes might get orphaned. workers[iworker] = w = self.startProcess(iworker, testQueue, resultQueue, shouldStop, result) break os.kill(w.pid, signal.SIGILL) time.sleep(0.1) if not any_alive and testQueue.empty(): log.debug("All workers dead") break nexttimeout=self.config.multiprocess_timeout for w in workers: if w.is_alive() and len(w.currentaddr.value) > 0: timeprocessing = time.time()-w.currentstart.value if timeprocessing <= self.config.multiprocess_timeout: nexttimeout = min(nexttimeout, self.config.multiprocess_timeout-timeprocessing) log.debug("Completed %s tasks (%s remain)", len(completed), len(tasks))
except (KeyboardInterrupt, SystemExit) as e: log.info('parent received ctrl-c when waiting for test results') thrownError = e #resultQueue.get(False)
result.addError(test, sys.exc_info())
try: for case in to_teardown: log.debug("Tearing down shared fixtures for %s", case) try: case.tearDown() except (KeyboardInterrupt, SystemExit): raise except: result.addError(case, sys.exc_info())
stop = time.time()
# first write since can freeze on shutting down processes result.printErrors() result.printSummary(start, stop) self.config.plugins.finalize(result)
if thrownError is None: log.debug("Tell all workers to stop") for w in workers: if w.is_alive(): testQueue.put('STOP', block=False)
# wait for the workers to end for iworker,worker in enumerate(workers): if worker.is_alive(): log.debug('joining worker %s',iworker) worker.join() if worker.is_alive(): log.debug('failed to join worker %s',iworker) except (KeyboardInterrupt, SystemExit): log.info('parent received ctrl-c when shutting down: stop all processes') for worker in workers: if worker.is_alive(): worker.terminate()
if thrownError: raise thrownError else: raise
return result
arg = None if isinstance(case,nose.case.Test) and hasattr(case.test,'arg'): # this removes the top level descriptor and allows real function # name to be returned case.test.descriptor = None arg = case.test.arg test_addr = MultiProcessTestRunner.address(case) testQueue.put((test_addr,arg), block=False) if arg is not None: test_addr += str(arg) if tasks is not None: tasks.append(test_addr) return test_addr
if hasattr(case, 'address'): file, mod, call = case.address() elif hasattr(case, 'context'): file, mod, call = test_address(case.context) else: raise Exception("Unable to convert %s to address" % case) parts = [] if file is None: if mod is None: raise Exception("Unaddressable case %s" % case) else: parts.append(mod) else: # strip __init__.py(c) from end of file part # if present, having it there confuses loader dirname, basename = os.path.split(file) if basename.startswith('__init__'): file = dirname parts.append(file) if call is not None: parts.append(call) return ':'.join(map(str, parts))
# allows tests or suites to mark themselves as not safe # for multiprocess execution if hasattr(test, 'context'): if not getattr(test.context, '_multiprocess_', True): return
if ((isinstance(test, ContextSuite) and test.hasFixtures(self.checkCanSplit)) or not getattr(test, 'can_split', True) or not isinstance(test, unittest.TestSuite)): # regular test case, or a suite with context fixtures
# special case: when run like nosetests path/to/module.py # the top-level suite has only one item, and it shares # the same context as that item. In that case, we want the # item, not the top-level suite if isinstance(test, ContextSuite): contained = list(test) if (len(contained) == 1 and getattr(contained[0], 'context', None) == test.context): test = contained[0] yield test else: # Suite is without fixtures at this level; but it may have # fixtures at any deeper level, so we need to examine it all # the way down to the case level for case in test: for batch in self.nextBatch(case): yield batch
""" Callback that we use to check whether the fixtures found in a context or ancestor are ones we care about.
Contexts can tell us that their fixtures are reentrant by setting _multiprocess_can_split_. So if we see that, we return False to disregard those fixtures. """ if not fixt: return False if getattr(context, '_multiprocess_can_split_', False): return False return True
context = getattr(case, 'context', None) if not context: return False return getattr(context, '_multiprocess_shared_', False)
log.debug("batch result is %s" , batch_result) try: output, testsRun, failures, errors, errorClasses = batch_result except ValueError: log.debug("result in unexpected format %s", batch_result) failure.Failure(*sys.exc_info())(result) return self.stream.write(output) result.testsRun += testsRun result.failures.extend(failures) result.errors.extend(errors) for key, (storage, label, isfail) in list(errorClasses.items()): if key not in result.errorClasses: # Ordinarily storage is result attribute # but it's only processed through the errorClasses # dict, so it's ok to fake it here result.errorClasses[key] = ([], label, isfail) mystorage, _junk, _junk = result.errorClasses[key] mystorage.extend(storage) log.debug("Ran %s tests (total: %s)", testsRun, result.testsRun)
keyboardCaught, shouldStop, loaderClass, resultClass, config): try: try: return __runner(ix, testQueue, resultQueue, currentaddr, currentstart, keyboardCaught, shouldStop, loaderClass, resultClass, config) except KeyboardInterrupt: log.debug('Worker %s keyboard interrupt, stopping',ix) except Empty: log.debug("Worker %s timed out waiting for tasks", ix)
keyboardCaught, shouldStop, loaderClass, resultClass, config):
config = pickle.loads(config) dummy_parser = config.parserClass() if _instantiate_plugins is not None: for pluginclass in _instantiate_plugins: plugin = pluginclass() plugin.addOptions(dummy_parser,{}) config.plugins.addPlugin(plugin) config.plugins.configure(config.options,config) config.plugins.begin() log.debug("Worker %s executing, pid=%d", ix,os.getpid()) loader = loaderClass(config=config) loader.suiteClass.suiteClass = NoSharedFixtureContextSuite
def get(): return testQueue.get(timeout=config.multiprocess_timeout)
def makeResult(): stream = _WritelnDecorator(StringIO()) result = resultClass(stream, descriptions=1, verbosity=config.verbosity, config=config) plug_result = config.plugins.prepareTestResult(result) if plug_result: return plug_result return result
def batch(result): failures = [(TestLet(c), err) for c, err in result.failures] errors = [(TestLet(c), err) for c, err in result.errors] errorClasses = {} for key, (storage, label, isfail) in list(result.errorClasses.items()): errorClasses[key] = ([(TestLet(c), err) for c, err in storage], label, isfail) return ( result.stream.getvalue(), result.testsRun, failures, errors, errorClasses) for test_addr, arg in iter(get, 'STOP'): if shouldStop.is_set(): log.exception('Worker %d STOPPED',ix) break result = makeResult() test = loader.loadTestsFromNames([test_addr]) test.testQueue = testQueue test.tasks = [] test.arg = arg log.debug("Worker %s Test is %s (%s)", ix, test_addr, test) try: if arg is not None: test_addr = test_addr + str(arg) currentaddr.value = bytes_(test_addr) currentstart.value = time.time() test(result) currentaddr.value = bytes_('') resultQueue.put((ix, test_addr, test.tasks, batch(result))) except KeyboardInterrupt as e: #TimedOutException: timeout = isinstance(e, TimedOutException) if timeout: keyboardCaught.set() if len(currentaddr.value): if timeout: msg = 'Worker %s timed out, failing current test %s' else: msg = 'Worker %s keyboard interrupt, failing current test %s' log.exception(msg,ix,test_addr) currentaddr.value = bytes_('') failure.Failure(*sys.exc_info())(result) resultQueue.put((ix, test_addr, test.tasks, batch(result))) else: if timeout: msg = 'Worker %s test %s timed out' else: msg = 'Worker %s test %s keyboard interrupt' log.debug(msg,ix,test_addr) resultQueue.put((ix, test_addr, test.tasks, batch(result))) if not timeout: raise except SystemExit: currentaddr.value = bytes_('') log.exception('Worker %s system exit',ix) raise except: currentaddr.value = bytes_('') log.exception("Worker %s error running test or returning " "results",ix) failure.Failure(*sys.exc_info())(result) resultQueue.put((ix, test_addr, test.tasks, batch(result))) if config.multiprocess_restartworker: break log.debug("Worker %s ending", ix)
""" Context suite that never fires shared fixtures.
When a context sets _multiprocess_shared_, fixtures in that context are executed by the main process. Using this suite class prevents them from executing in the runner process as well.
""" if getattr(context, '_multiprocess_shared_', False): return super(NoSharedFixtureContextSuite, self).setupContext(context)
if getattr(context, '_multiprocess_shared_', False): return super(NoSharedFixtureContextSuite, self).teardownContext(context) """Run tests in suite inside of suite fixtures. """ # proxy the result for myself log.debug("suite %s (%s) run called, tests: %s", id(self), self, self._tests) if self.resultProxy: result, orig = self.resultProxy(result, self), result else: result, orig = result, result try: #log.debug('setUp for %s', id(self)); self.setUp() except KeyboardInterrupt: raise except: self.error_context = 'setup' result.addError(self, self._exc_info()) return try: for test in self._tests: if (isinstance(test,nose.case.Test) and self.arg is not None): test.test.arg = self.arg else: test.arg = self.arg test.testQueue = self.testQueue test.tasks = self.tasks if result.shouldStop: log.debug("stopping") break # each nose.case.Test will create its own result proxy # so the cases need the original result, to avoid proxy # chains #log.debug('running test %s in suite %s', test, self); try: test(orig) except KeyboardInterrupt as e: timeout = isinstance(e, TimedOutException) if timeout: msg = 'Timeout when running test %s in suite %s' else: msg = 'KeyboardInterrupt when running test %s in suite %s' log.debug(msg, test, self) err = (TimedOutException,TimedOutException(str(test)), sys.exc_info()[2]) test.config.plugins.addError(test,err) orig.addError(test,err) if not timeout: raise finally: self.has_run = True try: #log.debug('tearDown for %s', id(self)); self.tearDown() except KeyboardInterrupt: raise except: self.error_context = 'teardown' result.addError(self, self._exc_info()) |