def test_all_pipeline(self):
self._wait_timestamp(self._missing_datetime_l - timedelta(hours=1),
self._missing_datetime_f - timedelta(hours=1))
self._generate_portal_input_data(self._missing_datetime_l,
self._event_time_filter_l,
self._missing_start_index_l, 1 << 13,
self._portal_manifest_l)
self._generate_portal_input_data(self._missing_datetime_f,
self._event_time_filter_f,
self._missing_start_index_f, 1 << 13,
self._portal_manifest_f)
self._wait_timestamp(self._dt_l - timedelta(hours=1),
self._dt_f - timedelta(hours=1))
self._generate_portal_input_data(self._dt_l, self._event_time_filter_l,
self._item_num_l, 1 << 13,
self._portal_manifest_l)
self._dt_l += timedelta(hours=1)
self.assertEqual(self._dt_f, self._dt_l)
self._wait_timestamp(self._dt_l - timedelta(hours=1),
self._dt_f - timedelta(hours=1))
data_source_l = self._master_client_l.GetDataSource(empty_pb2.Empty())
data_source_f = self._master_client_f.GetDataSource(empty_pb2.Empty())
rd_puber_l = raw_data_publisher.RawDataPublisher(
self._etcd_l, self._raw_data_pub_dir_l)
rd_puber_f = raw_data_publisher.RawDataPublisher(
self._etcd_f, self._raw_data_pub_dir_f)
for partition_id in range(
data_source_l.data_source_meta.partition_num):
rd_puber_f.finish_raw_data(partition_id)
rd_puber_l.finish_raw_data(partition_id)
while True:
req_l = dj_pb.DataSourceRequest(
data_source_meta=self._data_source_l.data_source_meta)
req_f = dj_pb.DataSourceRequest(
data_source_meta=self._data_source_f.data_source_meta)
dss_l = self._master_client_l.GetDataSourceStatus(req_l)
dss_f = self._master_client_f.GetDataSourceStatus(req_f)
self.assertEqual(dss_l.role, common_pb.FLRole.Leader)
self.assertEqual(dss_f.role, common_pb.FLRole.Follower)
if dss_l.state == common_pb.DataSourceState.Finished and \
dss_f.state == common_pb.DataSourceState.Finished:
break
else:
time.sleep(2)
logging.info("masters turn into Finished state")
def _preprocess_rsa_psi_follower(self):
processors = []
rsa_key_pem = None
with gfile.GFile(self._rsa_public_key_path, 'rb') as f:
rsa_key_pem = f.read()
self._follower_rsa_psi_sub_dir = 'follower_rsa_psi_sub_dir'
rd_publisher = raw_data_publisher.RawDataPublisher(
self._kvstore_f, self._follower_rsa_psi_sub_dir)
for partition_id in range(
self._data_source_f.data_source_meta.partition_num):
rd_publisher.publish_raw_data(
partition_id, [self._psi_raw_data_fpaths_f[partition_id]])
rd_publisher.finish_raw_data(partition_id)
options = dj_pb.RsaPsiPreProcessorOptions(
preprocessor_name='follower-rsa-psi-processor',
role=common_pb.FLRole.Follower,
rsa_key_pem=rsa_key_pem,
input_file_subscribe_dir=self._follower_rsa_psi_sub_dir,
output_file_dir=self._pre_processor_ouput_dir_f,
raw_data_publish_dir=self._raw_data_pub_dir_f,
partition_id=partition_id,
leader_rsa_psi_signer_addr=self._rsa_psi_signer_addr,
offload_processor_number=1,
max_flying_sign_batch=128,
max_flying_sign_rpc=64,
sign_rpc_timeout_ms=100000,
stub_fanout=2,
slow_sign_threshold=8,
sort_run_merger_read_ahead_buffer=1 << 20,
sort_run_merger_read_batch_size=128,
batch_processor_options=dj_pb.BatchProcessorOptions(
batch_size=1024, max_flying_item=1 << 14),
input_raw_data=dj_pb.RawDataOptions(raw_data_iter='TF_RECORD',
read_ahead_size=1 << 20),
writer_options=dj_pb.WriterOptions(output_writer='CSV_DICT'))
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
processor = rsa_psi_preprocessor.RsaPsiPreProcessor(
options, self.kvstore_type, True)
processor.start_process()
processors.append(processor)
for processor in processors:
processor.wait_for_finished()
def setUp(self):
etcd_name = 'test_etcd'
etcd_addrs = 'localhost:2379'
etcd_base_dir_l = 'byefl_l'
etcd_base_dir_f = 'byefl_f'
data_source_name = 'test_data_source'
etcd_l = EtcdClient(etcd_name, etcd_addrs, etcd_base_dir_l, True)
etcd_f = EtcdClient(etcd_name, etcd_addrs, etcd_base_dir_f, True)
etcd_l.delete_prefix(
common.data_source_etcd_base_dir(data_source_name))
etcd_f.delete_prefix(
common.data_source_etcd_base_dir(data_source_name))
data_source_l = common_pb.DataSource()
self.raw_data_pub_dir_l = './raw_data_pub_dir_l'
data_source_l.raw_data_sub_dir = self.raw_data_pub_dir_l
data_source_l.role = common_pb.FLRole.Leader
data_source_l.state = common_pb.DataSourceState.Init
data_source_l.data_block_dir = "./data_block_l"
data_source_l.raw_data_dir = "./raw_data_l"
data_source_l.example_dumped_dir = "./example_dumped_l"
data_source_f = common_pb.DataSource()
self.raw_data_pub_dir_f = './raw_data_pub_dir_f'
data_source_f.role = common_pb.FLRole.Follower
data_source_f.raw_data_sub_dir = self.raw_data_pub_dir_f
data_source_f.state = common_pb.DataSourceState.Init
data_source_f.data_block_dir = "./data_block_f"
data_source_f.raw_data_dir = "./raw_data_f"
data_source_f.example_dumped_dir = "./example_dumped_f"
data_source_meta = common_pb.DataSourceMeta()
data_source_meta.name = data_source_name
data_source_meta.partition_num = 2
data_source_meta.start_time = 0
data_source_meta.end_time = 100000000
data_source_l.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(etcd_l, data_source_l)
data_source_f.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(etcd_f, data_source_f)
master_options = dj_pb.DataJoinMasterOptions(use_mock_etcd=True)
master_addr_l = 'localhost:4061'
master_addr_f = 'localhost:4062'
master_l = data_join_master.DataJoinMasterService(
int(master_addr_l.split(':')[1]),
master_addr_f,
data_source_name,
etcd_name,
etcd_base_dir_l,
etcd_addrs,
master_options,
)
master_l.start()
master_f = data_join_master.DataJoinMasterService(
int(master_addr_f.split(':')[1]), master_addr_l, data_source_name,
etcd_name, etcd_base_dir_f, etcd_addrs, master_options)
master_f.start()
channel_l = make_insecure_channel(master_addr_l, ChannelType.INTERNAL)
master_client_l = dj_grpc.DataJoinMasterServiceStub(channel_l)
channel_f = make_insecure_channel(master_addr_f, ChannelType.INTERNAL)
master_client_f = dj_grpc.DataJoinMasterServiceStub(channel_f)
while True:
req_l = dj_pb.DataSourceRequest(
data_source_meta=data_source_l.data_source_meta)
req_f = dj_pb.DataSourceRequest(
data_source_meta=data_source_f.data_source_meta)
dss_l = master_client_l.GetDataSourceStatus(req_l)
dss_f = master_client_f.GetDataSourceStatus(req_f)
self.assertEqual(dss_l.role, common_pb.FLRole.Leader)
self.assertEqual(dss_f.role, common_pb.FLRole.Follower)
if dss_l.state == common_pb.DataSourceState.Processing and \
dss_f.state == common_pb.DataSourceState.Processing:
break
else:
time.sleep(2)
self.master_client_l = master_client_l
self.master_client_f = master_client_f
self.master_addr_l = master_addr_l
self.master_addr_f = master_addr_f
self.etcd_l = etcd_l
self.etcd_f = etcd_f
self.data_source_l = data_source_l
self.data_source_f = data_source_f
self.master_l = master_l
self.master_f = master_f
self.data_source_name = data_source_name,
self.etcd_name = etcd_name
self.etcd_addrs = etcd_addrs
self.etcd_base_dir_l = etcd_base_dir_l
self.etcd_base_dir_f = etcd_base_dir_f
self.raw_data_publisher_l = raw_data_publisher.RawDataPublisher(
self.etcd_l, self.raw_data_pub_dir_l)
self.raw_data_publisher_f = raw_data_publisher.RawDataPublisher(
self.etcd_f, self.raw_data_pub_dir_f)
if gfile.Exists(data_source_l.data_block_dir):
gfile.DeleteRecursively(data_source_l.data_block_dir)
if gfile.Exists(data_source_l.example_dumped_dir):
gfile.DeleteRecursively(data_source_l.example_dumped_dir)
if gfile.Exists(data_source_l.raw_data_dir):
gfile.DeleteRecursively(data_source_l.raw_data_dir)
if gfile.Exists(data_source_f.data_block_dir):
gfile.DeleteRecursively(data_source_f.data_block_dir)
if gfile.Exists(data_source_f.example_dumped_dir):
gfile.DeleteRecursively(data_source_f.example_dumped_dir)
if gfile.Exists(data_source_f.raw_data_dir):
gfile.DeleteRecursively(data_source_f.raw_data_dir)
self.worker_options = dj_pb.DataJoinWorkerOptions(
use_mock_etcd=True,
raw_data_options=dj_pb.RawDataOptions(raw_data_iter='TF_RECORD',
compressed_type=''),
example_id_dump_options=dj_pb.ExampleIdDumpOptions(
example_id_dump_interval=1, example_id_dump_threshold=1024),
example_joiner_options=dj_pb.ExampleJoinerOptions(
example_joiner='STREAM_JOINER',
min_matching_window=64,
max_matching_window=256,
data_block_dump_interval=30,
data_block_dump_threshold=1000),
batch_processor_options=dj_pb.BatchProcessorOptions(
batch_size=512, max_flying_item=2048),
data_block_builder_options=dj_pb.WriterOptions(
output_writer='TF_RECORD'))
self.total_index = 1 << 13
def setUp(self):
etcd_name = 'test_etcd'
etcd_addrs = 'localhost:2379'
etcd_base_dir_l = 'byefl_l'
etcd_base_dir_f= 'byefl_f'
data_source_name = 'test_data_source'
etcd_l = EtcdClient(etcd_name, etcd_addrs, etcd_base_dir_l, True)
etcd_f = EtcdClient(etcd_name, etcd_addrs, etcd_base_dir_f, True)
etcd_l.delete_prefix(common.data_source_etcd_base_dir(data_source_name))
etcd_f.delete_prefix(common.data_source_etcd_base_dir(data_source_name))
data_source_l = common_pb.DataSource()
self.raw_data_pub_dir_l = './raw_data_pub_dir_l'
data_source_l.raw_data_sub_dir = self.raw_data_pub_dir_l
data_source_l.role = common_pb.FLRole.Leader
data_source_l.state = common_pb.DataSourceState.Init
data_source_l.output_base_dir = "./ds_output_l"
self.raw_data_dir_l = "./raw_data_l"
data_source_f = common_pb.DataSource()
self.raw_data_pub_dir_f = './raw_data_pub_dir_f'
data_source_f.role = common_pb.FLRole.Follower
data_source_f.raw_data_sub_dir = self.raw_data_pub_dir_f
data_source_f.state = common_pb.DataSourceState.Init
data_source_f.output_base_dir = "./ds_output_f"
self.raw_data_dir_f = "./raw_data_f"
data_source_meta = common_pb.DataSourceMeta()
data_source_meta.name = data_source_name
data_source_meta.partition_num = 2
data_source_meta.start_time = 0
data_source_meta.end_time = 100000000
data_source_l.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(etcd_l, data_source_l)
data_source_f.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(etcd_f, data_source_f)
self.etcd_l = etcd_l
self.etcd_f = etcd_f
self.data_source_l = data_source_l
self.data_source_f = data_source_f
self.data_source_name = data_source_name
self.etcd_name = etcd_name
self.etcd_addrs = etcd_addrs
self.etcd_base_dir_l = etcd_base_dir_l
self.etcd_base_dir_f = etcd_base_dir_f
self.raw_data_publisher_l = raw_data_publisher.RawDataPublisher(
self.etcd_l, self.raw_data_pub_dir_l
)
self.raw_data_publisher_f = raw_data_publisher.RawDataPublisher(
self.etcd_f, self.raw_data_pub_dir_f
)
if gfile.Exists(data_source_l.output_base_dir):
gfile.DeleteRecursively(data_source_l.output_base_dir)
if gfile.Exists(self.raw_data_dir_l):
gfile.DeleteRecursively(self.raw_data_dir_l)
if gfile.Exists(data_source_f.output_base_dir):
gfile.DeleteRecursively(data_source_f.output_base_dir)
if gfile.Exists(self.raw_data_dir_f):
gfile.DeleteRecursively(self.raw_data_dir_f)
self.worker_options = dj_pb.DataJoinWorkerOptions(
use_mock_etcd=True,
raw_data_options=dj_pb.RawDataOptions(
raw_data_iter='TF_RECORD',
read_ahead_size=1<<20,
read_batch_size=128
),
example_id_dump_options=dj_pb.ExampleIdDumpOptions(
example_id_dump_interval=1,
example_id_dump_threshold=1024
),
example_joiner_options=dj_pb.ExampleJoinerOptions(
example_joiner='STREAM_JOINER',
min_matching_window=64,
max_matching_window=256,
data_block_dump_interval=30,
data_block_dump_threshold=1000
),
batch_processor_options=dj_pb.BatchProcessorOptions(
batch_size=512,
max_flying_item=2048
),
data_block_builder_options=dj_pb.WriterOptions(
output_writer='TF_RECORD'
)
)
self.total_index = 1 << 12
