def __init__(self, example_joiner_options, raw_data_options,
data_block_builder_options, kvstore, data_source,
partition_id):
self._lock = threading.Lock()
self._example_joiner_options = example_joiner_options
self._raw_data_options = raw_data_options
self._data_source = data_source
self._partition_id = partition_id
self._leader_visitor = \
ExampleIdVisitor(kvstore, self._data_source, self._partition_id)
self._follower_visitor = \
RawDataVisitor(kvstore, self._data_source,
self._partition_id, raw_data_options)
self._data_block_manager = \
DataBlockManager(self._data_source, self._partition_id)
meta = self._data_block_manager.get_lastest_data_block_meta()
if meta is None:
self._joiner_stats = JoinerStats(0, -1, -1)
else:
stats_info = meta.joiner_stats_info
self._joiner_stats = JoinerStats(stats_info.stats_cum_join_num,
stats_info.leader_stats_index,
stats_info.follower_stats_index)
self._data_block_builder_options = data_block_builder_options
self._data_block_builder = None
self._state_stale = False
self._follower_restart_index = 0
self._sync_example_id_finished = False
self._raw_data_finished = False
self._join_finished = False
ds_name = self._data_source.data_source_meta.name
self._metrics_tags = {
'data_source_name': ds_name,
'partition': partition_id,
'joiner_name': self.name()
}
self._optional_stats = OptionalStats(raw_data_options,
self._metrics_tags)
self._latest_dump_timestamp = time.time()
self._sync_state()
def __init__(self, kvstore, data_source, partition_id, raw_data_options,
data_block_builder_options):
self._lock = threading.Lock()
self._data_source = data_source
self._partition_id = partition_id
self._data_block_manager = \
DataBlockManager(data_source, partition_id)
self._raw_data_visitor = \
RawDataVisitor(kvstore, data_source,
partition_id, raw_data_options)
self._data_block_builder_options = data_block_builder_options
self._next_data_block_index = \
self._data_block_manager.get_dumped_data_block_count()
self._fly_data_block_meta = []
self._state_stale = False
self._synced_data_block_meta_finished = False
ds_name = self._data_source.data_source_meta.name
self._metrics_tags = {
'data_source_name': ds_name,
'partition': self._partition_id
}
self._optional_stats = OptionalStats(raw_data_options,
self._metrics_tags)
class DataBlockDumperManager(object):
def __init__(self, kvstore, data_source, partition_id, raw_data_options,
data_block_builder_options):
self._lock = threading.Lock()
self._data_source = data_source
self._partition_id = partition_id
self._data_block_manager = \
DataBlockManager(data_source, partition_id)
self._raw_data_visitor = \
RawDataVisitor(kvstore, data_source,
partition_id, raw_data_options)
self._data_block_builder_options = data_block_builder_options
self._next_data_block_index = \
self._data_block_manager.get_dumped_data_block_count()
self._fly_data_block_meta = []
self._state_stale = False
self._synced_data_block_meta_finished = False
ds_name = self._data_source.data_source_meta.name
self._metrics_tags = {
'data_source_name': ds_name,
'partition': self._partition_id
}
self._optional_stats = OptionalStats(raw_data_options,
self._metrics_tags)
def get_next_data_block_index(self):
with self._lock:
return self._next_data_block_index
def get_dumped_data_block_index(self):
return self._data_block_manager.get_dumped_data_block_count() - 1
def add_synced_data_block_meta(self, meta):
with self._lock:
if self._synced_data_block_meta_finished:
raise RuntimeError(
"data block dumper manager has been mark as "\
"no more data block meta"
)
if self._next_data_block_index != meta.data_block_index:
return False, self._next_data_block_index
self._fly_data_block_meta.append(meta)
self._next_data_block_index += 1
return True, self._next_data_block_index
def finish_sync_data_block_meta(self):
with self._lock:
self._synced_data_block_meta_finished = True
def need_dump(self):
with self._lock:
return len(self._fly_data_block_meta) > 0
def is_synced_data_block_meta_finished(self):
with self._lock:
return self._synced_data_block_meta_finished
@contextmanager
def make_data_block_dumper(self):
self._sync_with_data_block_manager()
self._acquire_state_stale()
yield self._dump_data_blocks
self._release_state_stale()
def _dump_data_blocks(self):
while self.need_dump():
meta = self._get_next_data_block_meta()
if meta is not None:
start_tm = time.time()
self._raw_data_visitor.active_visitor()
self._dump_data_block_by_meta(meta)
dump_duration = time.time() - start_tm
metrics.emit_timer(name='data_block_dump_duration',
value=int(dump_duration),
tags=self._metrics_tags)
def data_block_meta_sync_finished(self):
with self._lock:
return self._synced_data_block_meta_finished
def _acquire_state_stale(self):
with self._lock:
self._state_stale = True
def _release_state_stale(self):
with self._lock:
self._state_stale = False
def _get_next_data_block_meta(self):
with self._lock:
if len(self._fly_data_block_meta) == 0:
return None
return self._fly_data_block_meta[0]
@contextmanager
def _make_data_block_builder(self, meta):
assert self._partition_id == meta.partition_id, \
"partition id of building data block meta mismatch "\
"{} != {}".format(self._partition_id, meta.partition_id)
builder = None
expt = None
try:
builder = DataBlockBuilder(
common.data_source_data_block_dir(self._data_source),
self._data_source.data_source_meta.name, self._partition_id,
meta.data_block_index, self._data_block_builder_options)
builder.init_by_meta(meta)
builder.set_data_block_manager(self._data_block_manager)
yield builder
except Exception as e: # pylint: disable=broad-except
logging.warning("Failed make data block builder, " \
"reason %s", e)
expt = e
if builder is not None:
del builder
if expt is not None:
raise expt
def _dump_data_block_by_meta(self, meta):
assert meta is not None, "input data block must not be None"
with self._make_data_block_builder(meta) as data_block_builder:
try:
if meta.leader_start_index == 0:
self._raw_data_visitor.reset()
else:
assert meta.leader_start_index > 0, \
"leader start index must be positive"
self._raw_data_visitor.seek(meta.leader_start_index - 1)
except StopIteration:
logging.fatal("raw data finished before when seek to %d",
meta.leader_start_index - 1)
traceback.print_stack()
os._exit(-1) # pylint: disable=protected-access
match_index = 0
example_num = len(meta.example_ids)
for (index, item) in self._raw_data_visitor:
example_id = item.example_id
joined = False
# Elements in meta.example_ids maybe duplicated
while match_index < example_num and \
example_id == meta.example_ids[match_index]:
data_block_builder.write_item(item)
self._optional_stats.update_stats(item, kind='joined')
match_index += 1
joined = True
if not joined:
self._optional_stats.update_stats(item, kind='unjoined')
if match_index >= example_num:
break
if index >= meta.leader_end_index:
break
if match_index < example_num:
logging.fatal(
"Data lose corrupt! only match %d/%d example "
"for data block %s", match_index, example_num,
meta.block_id)
traceback.print_stack()
os._exit(-1) # pylint: disable=protected-access
dumped_meta = data_block_builder.finish_data_block(True)
self._optional_stats.emit_optional_stats()
assert dumped_meta == meta, "the generated dumped meta should "\
"be the same with input mata"
with self._lock:
assert self._fly_data_block_meta[0] == meta
self._fly_data_block_meta.pop(0)
def _is_state_stale(self):
with self._lock:
return self._state_stale
def _sync_with_data_block_manager(self):
if self._is_state_stale():
self._evict_dumped_data_block_meta()
def _evict_dumped_data_block_meta(self):
next_data_block_index = \
self._data_block_manager.get_dumped_data_block_count()
with self._lock:
skip_count = 0
for meta in self._fly_data_block_meta:
if meta.data_block_index >= next_data_block_index:
break
skip_count += 1
self._fly_data_block_meta = \
self._fly_data_block_meta[skip_count:]
class ExampleJoiner(object):
def __init__(self, example_joiner_options, raw_data_options,
data_block_builder_options, kvstore, data_source,
partition_id):
self._lock = threading.Lock()
self._example_joiner_options = example_joiner_options
self._raw_data_options = raw_data_options
self._data_source = data_source
self._partition_id = partition_id
self._leader_visitor = \
ExampleIdVisitor(kvstore, self._data_source, self._partition_id)
self._follower_visitor = \
RawDataVisitor(kvstore, self._data_source,
self._partition_id, raw_data_options)
self._data_block_manager = \
DataBlockManager(self._data_source, self._partition_id)
meta = self._data_block_manager.get_lastest_data_block_meta()
if meta is None:
self._joiner_stats = JoinerStats(0, -1, -1)
else:
stats_info = meta.joiner_stats_info
self._joiner_stats = JoinerStats(stats_info.stats_cum_join_num,
stats_info.leader_stats_index,
stats_info.follower_stats_index)
self._data_block_builder_options = data_block_builder_options
self._data_block_builder = None
self._state_stale = False
self._follower_restart_index = 0
self._sync_example_id_finished = False
self._raw_data_finished = False
self._join_finished = False
ds_name = self._data_source.data_source_meta.name
self._metrics_tags = {
'data_source_name': ds_name,
'partition': partition_id,
'joiner_name': self.name()
}
self._optional_stats = OptionalStats(raw_data_options,
self._metrics_tags)
self._latest_dump_timestamp = time.time()
self._sync_state()
@contextmanager
def make_example_joiner(self):
state_stale = self._is_state_stale()
self._acuqire_state_stale()
yield self._inner_joiner(state_stale)
self._release_state_stale()
@classmethod
def name(cls):
return 'BASE_EXAMPLE_JOINER'
def get_data_block_meta_by_index(self, index):
with self._lock:
manager = self._data_block_manager
return self._join_finished, \
manager.get_data_block_meta_by_index(index)
def get_dumped_data_block_count(self):
return self._data_block_manager.get_dumped_data_block_count()
def is_join_finished(self):
with self._lock:
return self._join_finished
def set_sync_example_id_finished(self):
with self._lock:
self._sync_example_id_finished = True
def set_raw_data_finished(self):
with self._lock:
self._raw_data_finished = True
def is_sync_example_id_finished(self):
with self._lock:
return self._sync_example_id_finished
def is_raw_data_finished(self):
with self._lock:
return self._raw_data_finished
def need_join(self):
with self._lock:
if self._join_finished:
return False
if self._state_stale or self._sync_example_id_finished:
return True
if self._follower_visitor.is_visitor_stale() or \
self._leader_visitor.is_visitor_stale():
return True
if not self._follower_visitor.finished() and \
not self._leader_visitor.finished():
return True
return self._need_finish_data_block_since_interval()
def _prepare_join(self, state_stale):
if state_stale:
self._sync_state()
self._reset_data_block_builder()
sync_example_id_finished = self.is_sync_example_id_finished()
raw_data_finished = self.is_raw_data_finished()
self._active_visitors()
return sync_example_id_finished, raw_data_finished
def _inner_joiner(self, reset_state):
raise NotImplementedError(
"_inner_joiner not implement for base class: %s" %
ExampleJoiner.name())
def _is_state_stale(self):
with self._lock:
return self._state_stale
def _active_visitors(self):
self._leader_visitor.active_visitor()
self._follower_visitor.active_visitor()
def _sync_state(self):
meta = self._data_block_manager.get_lastest_data_block_meta()
if meta is not None:
try:
self._leader_visitor.seek(meta.leader_end_index)
except StopIteration:
logging.warning("leader visitor finished")
try:
self._follower_visitor.seek(meta.follower_restart_index)
except StopIteration:
logging.warning("follower visitor finished")
else:
self._leader_visitor.reset()
self._follower_visitor.reset()
def _get_data_block_builder(self, create_if_no_existed):
if self._data_block_builder is None and create_if_no_existed:
data_block_index = \
self._data_block_manager.get_dumped_data_block_count()
self._data_block_builder = DataBlockBuilder(
common.data_source_data_block_dir(self._data_source),
self._data_source.data_source_meta.name, self._partition_id,
data_block_index, self._data_block_builder_options,
self._example_joiner_options.data_block_dump_threshold)
self._data_block_builder.set_data_block_manager(
self._data_block_manager)
self._data_block_builder.set_follower_restart_index(
self._follower_restart_index)
return self._data_block_builder
def _finish_data_block(self):
if self._data_block_builder is not None:
self._data_block_builder.set_join_stats_info(
self._create_join_stats_info())
meta = self._data_block_builder.finish_data_block(
True, self._metrics_tags)
self._optional_stats.emit_optional_stats()
self._reset_data_block_builder()
self._update_latest_dump_timestamp()
return meta
return None
def _create_join_stats_info(self):
builder = self._get_data_block_builder(False)
nstats_cum_join_num = self._joiner_stats.calc_stats_joined_num()
nactual_cum_join_num = 0 if builder is None \
else builder.example_count()
meta = self._data_block_manager.get_lastest_data_block_meta()
if meta is not None:
nactual_cum_join_num += meta.joiner_stats_info.actual_cum_join_num
return dj_pb.JoinerStatsInfo(
stats_cum_join_num=nstats_cum_join_num,
actual_cum_join_num=nactual_cum_join_num,
leader_stats_index=self._joiner_stats.get_leader_stats_index(),
follower_stats_index=self._joiner_stats.get_follower_stats_index())
def _reset_data_block_builder(self):
builder = None
with self._lock:
builder = self._data_block_builder
self._data_block_builder = None
if builder is not None:
del builder
def _update_latest_dump_timestamp(self):
data_block_dump_duration = time.time() - self._latest_dump_timestamp
metrics.emit_timer(name='data_block_dump_duration',
value=int(data_block_dump_duration),
tags=self._metrics_tags)
self._latest_dump_timestamp = time.time()
def _acuqire_state_stale(self):
with self._lock:
self._state_stale = True
def _release_state_stale(self):
with self._lock:
self._state_stale = False
def _set_join_finished(self):
with self._lock:
self._join_finished = True
def _need_finish_data_block_since_interval(self):
dump_interval = self._example_joiner_options.data_block_dump_interval
duration_since_dump = time.time() - self._latest_dump_timestamp
return 0 < dump_interval <= duration_since_dump