def __init__(self, *args, **kwargs): super(RunTracker, self).__init__(*args, **kwargs) run_timestamp = time.time() cmd_line = ' '.join(['pants'] + sys.argv[1:]) # run_id is safe for use in paths. millis = int((run_timestamp * 1000) % 1000) run_id = 'pants_run_{}_{}_{}'.format( time.strftime('%Y_%m_%d_%H_%M_%S', time.localtime(run_timestamp)), millis, uuid.uuid4().hex) info_dir = os.path.join(self.get_options().pants_workdir, self.options_scope) self.run_info_dir = os.path.join(info_dir, run_id) self.run_info = RunInfo(os.path.join(self.run_info_dir, 'info')) self.run_info.add_basic_info(run_id, run_timestamp) self.run_info.add_info('cmd_line', cmd_line) # Create a 'latest' symlink, after we add_infos, so we're guaranteed that the file exists. link_to_latest = os.path.join(os.path.dirname(self.run_info_dir), 'latest') relative_symlink(self.run_info_dir, link_to_latest) # Time spent in a workunit, including its children. self.cumulative_timings = AggregatedTimings(os.path.join(self.run_info_dir, 'cumulative_timings')) # Time spent in a workunit, not including its children. self.self_timings = AggregatedTimings(os.path.join(self.run_info_dir, 'self_timings')) # Hit/miss stats for the artifact cache. self.artifact_cache_stats = \ ArtifactCacheStats(os.path.join(self.run_info_dir, 'artifact_cache_stats')) # Number of threads for foreground work. self._num_foreground_workers = self.get_options().num_foreground_workers # Number of threads for background work. self._num_background_workers = self.get_options().num_background_workers # We report to this Report. self.report = None # self._threadlocal.current_workunit contains the current workunit for the calling thread. # Note that multiple threads may share a name (e.g., all the threads in a pool). self._threadlocal = threading.local() # For main thread work. Created on start(). self._main_root_workunit = None # For background work. Created lazily if needed. self._background_worker_pool = None self._background_root_workunit = None # Trigger subproc pool init while our memory image is still clean (see SubprocPool docstring). SubprocPool.set_num_processes(self._num_foreground_workers) SubprocPool.foreground() self._aborted = False
def mock_artifact_cache_stats(self, expected_stats, expected_hit_or_miss_files=None): with temporary_dir() as tmp_dir: artifact_cache_stats = ArtifactCacheStats(tmp_dir) yield artifact_cache_stats self.assertEqual(expected_stats, artifact_cache_stats.get_all()) self.assertEqual(sorted(list(expected_hit_or_miss_files.keys())), sorted(os.listdir(tmp_dir))) for hit_or_miss_file in expected_hit_or_miss_files.keys(): with open(os.path.join(tmp_dir, hit_or_miss_file), 'r') as hit_or_miss_saved: self.assertEqual(expected_hit_or_miss_files[hit_or_miss_file], hit_or_miss_saved.read())
def initialize(self, all_options): """Create run_info and relevant directories, and return the run id. Must be called before `start`. """ if self.run_info: raise AssertionError( 'RunTracker.initialize must not be called multiple times.') # Initialize the run. # Select a globally unique ID for the run, that sorts by time. millis = int((self._run_timestamp * 1000) % 1000) # run_uuid is used as a part of run_id and also as a trace_id for Zipkin tracing run_uuid = uuid.uuid4().hex run_id = 'pants_run_{}_{}_{}'.format( time.strftime('%Y_%m_%d_%H_%M_%S', time.localtime(self._run_timestamp)), millis, run_uuid) info_dir = os.path.join(self.get_options().pants_workdir, self.options_scope) self.run_info_dir = os.path.join(info_dir, run_id) self.run_info = RunInfo(os.path.join(self.run_info_dir, 'info')) self.run_info.add_basic_info(run_id, self._run_timestamp) self.run_info.add_info('cmd_line', self._cmd_line) # Create a 'latest' symlink, after we add_infos, so we're guaranteed that the file exists. link_to_latest = os.path.join(os.path.dirname(self.run_info_dir), 'latest') relative_symlink(self.run_info_dir, link_to_latest) # Time spent in a workunit, including its children. self.cumulative_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'cumulative_timings')) # Time spent in a workunit, not including its children. self.self_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'self_timings')) # Hit/miss stats for the artifact cache. self.artifact_cache_stats = ArtifactCacheStats( os.path.join(self.run_info_dir, 'artifact_cache_stats')) # Daemon stats. self.pantsd_stats = PantsDaemonStats() self._all_options = all_options return (run_id, run_uuid)
def initialize(self, all_options): """Create run_info and relevant directories, and return the run id. Must be called before `start`. """ if self.run_info: raise AssertionError( 'RunTracker.initialize must not be called multiple times.') # Initialize the run. info_dir = os.path.join(self.get_options().pants_workdir, self.options_scope) self.run_info_dir = os.path.join(info_dir, self.run_id) self.run_info = RunInfo(os.path.join(self.run_info_dir, 'info')) self.run_info.add_basic_info(self.run_id, self._run_timestamp) self.run_info.add_info('cmd_line', self._cmd_line) if self.get_options().parent_build_id: self.run_info.add_info('parent_build_id', self.get_options().parent_build_id) # Create a 'latest' symlink, after we add_infos, so we're guaranteed that the file exists. link_to_latest = os.path.join(os.path.dirname(self.run_info_dir), 'latest') relative_symlink(self.run_info_dir, link_to_latest) # Time spent in a workunit, including its children. self.cumulative_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'cumulative_timings')) # Time spent in a workunit, not including its children. self.self_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'self_timings')) # Hit/miss stats for the artifact cache. self.artifact_cache_stats = ArtifactCacheStats( os.path.join(self.run_info_dir, 'artifact_cache_stats')) # Daemon stats. self.pantsd_stats = PantsDaemonStats() self._all_options = all_options return (self.run_id, self.run_uuid)
def __init__(self, info_dir, stats_upload_url=None, stats_upload_timeout=2, num_foreground_workers=8, num_background_workers=8): self.run_timestamp = time.time( ) # A double, so we get subsecond precision for ids. cmd_line = ' '.join(['./pants'] + sys.argv[1:]) # run_id is safe for use in paths. millis = (self.run_timestamp * 1000) % 1000 run_id = 'pants_run_%s_%d' % \ (time.strftime('%Y_%m_%d_%H_%M_%S', time.localtime(self.run_timestamp)), millis) self.run_info_dir = os.path.join(info_dir, run_id) self.run_info = RunInfo(os.path.join(self.run_info_dir, 'info')) self.run_info.add_basic_info(run_id, self.run_timestamp) self.run_info.add_info('cmd_line', cmd_line) self.stats_url = stats_upload_url self.stats_timeout = stats_upload_timeout # Create a 'latest' symlink, after we add_infos, so we're guaranteed that the file exists. link_to_latest = os.path.join(os.path.dirname(self.run_info_dir), 'latest') try: if os.path.lexists(link_to_latest): os.unlink(link_to_latest) os.symlink(self.run_info_dir, link_to_latest) except OSError as e: # Another run may beat us to deletion or creation. if not (e.errno == errno.EEXIST or e.errno == errno.ENOENT): raise # Time spent in a workunit, including its children. self.cumulative_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'cumulative_timings')) # Time spent in a workunit, not including its children. self.self_timings = AggregatedTimings( os.path.join(self.run_info_dir, 'self_timings')) # Hit/miss stats for the artifact cache. self.artifact_cache_stats = \ ArtifactCacheStats(os.path.join(self.run_info_dir, 'artifact_cache_stats')) # Number of threads for foreground work. self._num_foreground_workers = num_foreground_workers # Number of threads for background work. self._num_background_workers = num_background_workers # We report to this Report. self.report = None # self._threadlocal.current_workunit contains the current workunit for the calling thread. # Note that multiple threads may share a name (e.g., all the threads in a pool). self._threadlocal = threading.local() # For main thread work. Created on start(). self._main_root_workunit = None # For concurrent foreground work. Created lazily if needed. # Associated with the main thread's root workunit. self._foreground_worker_pool = None # For background work. Created lazily if needed. self._background_worker_pool = None self._background_root_workunit = None # Trigger subproc pool init while our memory image is still clean (see SubprocPool docstring) SubprocPool.foreground() self._aborted = False