def test_csv_raw_data_visitor(self):
self.data_source = common_pb.DataSource()
self.data_source.data_source_meta.name = 'fclh_test'
self.data_source.data_source_meta.partition_num = 1
self.raw_data_dir = path.join(
path.dirname(path.abspath(__file__)), "../csv_raw_data"
)
self.kvstore = DBClient("etcd", True)
self.kvstore.delete_prefix(common.data_source_kvstore_base_dir(self.data_source.data_source_meta.name))
self.assertEqual(self.data_source.data_source_meta.partition_num, 1)
partition_dir = path.join(self.raw_data_dir, common.partition_repr(0))
self.assertTrue(gfile.Exists(partition_dir))
manifest_manager = raw_data_manifest_manager.RawDataManifestManager(
self.kvstore, self.data_source)
manifest_manager.add_raw_data(
0, [dj_pb.RawDataMeta(file_path=path.join(partition_dir, "test_raw_data.csv"),
timestamp=timestamp_pb2.Timestamp(seconds=3))],
True)
raw_data_options = dj_pb.RawDataOptions(
raw_data_iter='CSV_DICT',
read_ahead_size=1<<20
)
rdm = raw_data_visitor.RawDataManager(self.kvstore, self.data_source,0)
self.assertTrue(rdm.check_index_meta_by_process_index(0))
rdv = raw_data_visitor.RawDataVisitor(self.kvstore, self.data_source, 0,
raw_data_options)
expected_index = 0
for (index, item) in rdv:
if index > 0 and index % 1024 == 0:
print("{} {}".format(index, item.raw_id))
self.assertEqual(index, expected_index)
expected_index += 1
self.assertEqual(expected_index, 4999)
def setUp(self) -> None:
logging.getLogger().setLevel(logging.DEBUG)
self._data_portal_name = 'test_data_portal_job_manager'
self._kvstore = DBClient('etcd', True)
self._portal_input_base_dir = './portal_input_dir'
self._portal_output_base_dir = './portal_output_dir'
self._raw_data_publish_dir = 'raw_data_publish_dir'
if gfile.Exists(self._portal_input_base_dir):
gfile.DeleteRecursively(self._portal_input_base_dir)
gfile.MakeDirs(self._portal_input_base_dir)
self._data_fnames = ['1001/{}.data'.format(i) for i in range(100)]
self._data_fnames_without_success = \
['1002/{}.data'.format(i) for i in range(100)]
self._csv_fnames = ['1003/{}.csv'.format(i) for i in range(100)]
self._unused_fnames = ['{}.xx'.format(100)]
self._all_fnames = self._data_fnames + \
self._data_fnames_without_success + \
self._csv_fnames + self._unused_fnames
all_fnames_with_success = ['1001/_SUCCESS'] + ['1003/_SUCCESS'] +\
self._all_fnames
for fname in all_fnames_with_success:
fpath = os.path.join(self._portal_input_base_dir, fname)
gfile.MakeDirs(os.path.dirname(fpath))
with gfile.Open(fpath, "w") as f:
f.write('xxx')
def _setUpMySQL(self):
self.kvstore_type = 'etcd'
self.leader_base_dir = 'bytefl_l'
self.follower_base_dir = 'bytefl_f'
os.environ['ETCD_BASE_DIR'] = self.leader_base_dir
self._kvstore_l = DBClient(self.kvstore_type, True)
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
self._kvstore_f = DBClient(self.kvstore_type, True)
def __init__(self, options, kvstore_type, use_mock_etcd=False):
self._lock = threading.Condition()
self._options = options
kvstore = DBClient(kvstore_type, use_mock_etcd)
pub_dir = self._options.raw_data_publish_dir
self._publisher = RawDataPublisher(kvstore, pub_dir)
self._process_pool_executor = \
concur_futures.ProcessPoolExecutor(
options.offload_processor_number
)
self._callback_submitter = None
# pre fock sub processor before launch grpc client
self._process_pool_executor.submit(min, 1, 2).result()
self._id_batch_fetcher = IdBatchFetcher(kvstore, self._options)
if self._options.role == common_pb.FLRole.Leader:
private_key = rsa.PrivateKey.load_pkcs1(options.rsa_key_pem)
self._psi_rsa_signer = LeaderPsiRsaSigner(
self._id_batch_fetcher,
options.batch_processor_options.max_flying_item,
self._options.max_flying_sign_batch,
self._options.slow_sign_threshold,
self._process_pool_executor,
private_key,
)
self._repr = 'leader-' + 'rsa_psi_preprocessor'
else:
public_key = rsa.PublicKey.load_pkcs1(options.rsa_key_pem)
self._callback_submitter = concur_futures.ThreadPoolExecutor(1)
self._psi_rsa_signer = FollowerPsiRsaSigner(
self._id_batch_fetcher,
options.batch_processor_options.max_flying_item,
self._options.max_flying_sign_batch,
self._options.max_flying_sign_rpc,
self._options.sign_rpc_timeout_ms,
self._options.slow_sign_threshold, self._options.stub_fanout,
self._process_pool_executor, self._callback_submitter,
public_key, self._options.leader_rsa_psi_signer_addr)
self._repr = 'follower-' + 'rsa_psi_preprocessor'
self._sort_run_dumper = SortRunDumper(options)
self._sort_run_merger = SortRunMerger(
dj_pb.SortRunMergerOptions(
merger_name='sort_run_merger_'+\
partition_repr(options.partition_id),
reader_options=dj_pb.RawDataOptions(
raw_data_iter=options.writer_options.output_writer,
compressed_type=options.writer_options.compressed_type,
read_ahead_size=\
options.sort_run_merger_read_ahead_buffer,
read_batch_size=\
options.sort_run_merger_read_batch_size
),
writer_options=options.writer_options,
output_file_dir=options.output_file_dir,
partition_id=options.partition_id
),
self._merger_comparator
)
self._produce_item_cnt = 0
self._comsume_item_cnt = 0
self._started = False
def __init__(self, listen_port, peer_addr, master_addr, rank_id,
kvstore_type, options):
master_channel = make_insecure_channel(
master_addr,
ChannelType.INTERNAL,
options=[('grpc.max_send_message_length', 2**31 - 1),
('grpc.max_receive_message_length', 2**31 - 1)])
self._master_client = dj_grpc.DataJoinMasterServiceStub(master_channel)
self._rank_id = rank_id
kvstore = DBClient(kvstore_type, options.use_mock_etcd)
data_source = self._sync_data_source()
self._data_source_name = data_source.data_source_meta.name
self._listen_port = listen_port
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
peer_channel = make_insecure_channel(
peer_addr,
ChannelType.REMOTE,
options=[('grpc.max_send_message_length', 2**31 - 1),
('grpc.max_receive_message_length', 2**31 - 1)])
peer_client = dj_grpc.DataJoinWorkerServiceStub(peer_channel)
self._data_join_worker = DataJoinWorker(peer_client,
self._master_client, rank_id,
kvstore, data_source, options)
dj_grpc.add_DataJoinWorkerServiceServicer_to_server(
self._data_join_worker, self._server)
self._role_repr = "leader" if data_source.role == \
common_pb.FLRole.Leader else "follower"
self._server.add_insecure_port('[::]:%d' % listen_port)
self._server_started = False
def __init__(self, peer_client, data_source_name, kvstore_type, options):
super(DataJoinMaster, self).__init__()
self._data_source_name = data_source_name
kvstore = DBClient(kvstore_type, options.use_mock_etcd)
self._options = options
self._fsm = MasterFSM(peer_client, data_source_name, kvstore,
self._options.batch_mode)
self._data_source_meta = \
self._fsm.get_data_source().data_source_meta
def __init__(self, listen_port, portal_name, kvstore_type, portal_options):
self._portal_name = portal_name
self._listen_port = listen_port
self._server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
kvstore = DBClient(kvstore_type, portal_options.use_mock_etcd)
self._data_portal_master = DataPortalMaster(portal_name, kvstore,
portal_options)
dp_grpc.add_DataPortalMasterServiceServicer_to_server(
self._data_portal_master, self._server)
self._server.add_insecure_port('[::]:%d' % listen_port)
self._server_started = False
def __init__(self, options, part_field, kvstore_type, use_mock_etcd=False):
self._options = options
self._part_field = part_field
kvstore = DBClient(kvstore_type, use_mock_etcd)
self._raw_data_batch_fetcher = RawDataBatchFetcher(kvstore, options)
self._next_part_index = None
self._dumped_process_index = None
self._flying_writers = []
self._dumped_file_metas = {}
self._worker_map = {}
self._started = False
self._part_finished = False
self._cond = threading.Condition()
def __init__(self, options):
self._fiter = None
self._index_meta = None
self._item = None
self._index = None
self._iter_failed = False
self._options = options
#_options will be None for example id visitor
if self._options and self._options.raw_data_cache_type == "disk":
#use leveldb to manager the disk storage by default
self._cache_type = DBClient("leveldb", False)
else:
self._cache_type = None
def __init__(self, options):
self._fiter = None
self._index_meta = None
self._item = None
self._index = None
self._iter_failed = False
self._options = options
try:
self._validator = Validator(options.validation_ratio)
except AttributeError:
self._validator = Validator()
#_options will be None for example id visitor
if self._options and self._options.raw_data_cache_type == "disk":
#use leveldb to manager the disk storage by default
self._cache_type = DBClient("leveldb", False)
else:
self._cache_type = None
class DataBlockVisitor(object):
def __init__(self, data_source_name, kvstore_type, use_mock_etcd=False):
self._kvstore = DBClient(kvstore_type, use_mock_etcd)
self._data_source = retrieve_data_source(self._kvstore,
data_source_name)
def LoadDataBlockRepByTimeFrame(self, start_time=None, end_time=None):
partition_num = self._data_source.data_source_meta.partition_num
data_block_fnames = {}
for partition_id in range(0, partition_num):
data_block_fnames[partition_id] = \
self._list_data_block(partition_id)
data_block_reps = {}
for partition_id, fnames in data_block_fnames.items():
manifest = self._sync_raw_data_manifest(partition_id)
for idx, fname in enumerate(fnames):
check_existed = (idx == len(fnames) - 1)
rep = self._make_data_block_rep(partition_id, fname,
check_existed)
filtered = True
reason = ''
if rep is None:
reason = 'failed to create data block rep'
elif end_time is not None and rep.end_time > end_time:
reason = 'excess time frame'
elif start_time is not None and rep.end_time <= start_time:
reason = 'less time frame'
elif self._filter_by_visible(rep.data_block_index, manifest):
reason = 'data block visible'
else:
data_block_reps[rep.block_id] = rep
filtered = False
if filtered:
logging.debug('skip %s since %s', fname, reason)
return data_block_reps
def LoadDataBlockReqByIndex(self, partition_id, data_block_index):
partition_num = self._data_source.data_source_meta.partition_num
if partition_id < 0 or partition_id >= partition_num:
raise IndexError("partition {} out range".format(partition_id))
dirpath = self._partition_data_block_dir(partition_id)
meta_fname = encode_data_block_meta_fname(self._data_source_name(),
partition_id,
data_block_index)
meta_fpath = os.path.join(dirpath, meta_fname)
meta = load_data_block_meta(meta_fpath)
manifest = self._sync_raw_data_manifest(partition_id)
if meta is not None and \
not self._filter_by_visible(meta.data_block_index, manifest):
fname = encode_data_block_fname(self._data_source_name(), meta)
return DataBlockRep(self._data_source_name(), fname, partition_id,
dirpath)
return None
def LoadDataBlockRepByBlockId(self, block_id):
block_info = decode_block_id(block_id)
dbr = self.LoadDataBlockReqByIndex(block_info['partition_id'],
block_info['data_block_index'])
if dbr:
assert dbr.block_id == block_id, \
"Invalid datablock, expected %s, but got %s), please "\
"check datasource!"%(block_id, dbr.block_id)
return dbr
def _list_data_block(self, partition_id):
dirpath = self._partition_data_block_dir(partition_id)
if gfile.Exists(dirpath) and gfile.IsDirectory(dirpath):
return [
f for f in gfile.ListDirectory(dirpath)
if f.endswith(DataBlockSuffix)
]
return []
def _partition_data_block_dir(self, partition_id):
return os.path.join(data_source_data_block_dir(self._data_source),
partition_repr(partition_id))
def _make_data_block_rep(self, partition_id, data_block_fname,
check_existed):
try:
rep = DataBlockRep(self._data_source.data_source_meta.name,
data_block_fname, partition_id,
self._partition_data_block_dir(partition_id),
check_existed)
except Exception as e: # pylint: disable=broad-except
logging.warning("Failed to create data block rep for %s in"\
"partition %d reason %s", data_block_fname,
partition_id, e)
return None
return rep
def _data_source_name(self):
return self._data_source.data_source_meta.name
def _sync_raw_data_manifest(self, partition_id):
kvstore_key = partition_manifest_kvstore_key(self._data_source_name(),
partition_id)
data = self._kvstore.get_data(kvstore_key)
assert data is not None, "raw data manifest of partition "\
"{} must be existed".format(partition_id)
return text_format.Parse(data, dj_pb.RawDataManifest())
def _filter_by_visible(self, index, manifest):
join_state = manifest.join_example_rep.state
if self._data_source.role == common_pb.FLRole.Follower and \
join_state != dj_pb.JoinExampleState.Joined:
return index > manifest.peer_dumped_index
return False
type=int,
default=None,
help='Max number of files in a job')
parser.add_argument('--start_date',
type=str,
default=None,
help='Start date of input data, format %Y%m%d')
parser.add_argument('--end_date',
type=str,
default=None,
help='End date of input data, format %Y%m%d')
args = parser.parse_args()
set_logger()
use_mock_etcd = (args.kvstore_type == 'mock')
kvstore = DBClient(args.kvstore_type, use_mock_etcd)
kvstore_key = common.portal_kvstore_base_dir(args.data_portal_name)
portal_manifest = kvstore.get_data(kvstore_key)
data_portal_type = dp_pb.DataPortalType.PSI if \
args.data_portal_type == 'PSI' else dp_pb.DataPortalType.Streaming
if portal_manifest is None:
portal_manifest = dp_pb.DataPortalManifest(
name=args.data_portal_name,
data_portal_type=data_portal_type,
output_partition_num=args.output_partition_num,
input_file_wildcard=args.input_file_wildcard,
input_base_dir=args.input_base_dir,
output_base_dir=args.output_base_dir,
raw_data_publish_dir=args.raw_data_publish_dir,
processing_job_id=-1)
kvstore.set_data(kvstore_key, text_format.\
class TestDataPortalJobManager(unittest.TestCase):
def setUp(self) -> None:
logging.getLogger().setLevel(logging.DEBUG)
self._data_portal_name = 'test_data_portal_job_manager'
self._kvstore = DBClient('etcd', True)
self._portal_input_base_dir = './portal_input_dir'
self._portal_output_base_dir = './portal_output_dir'
self._raw_data_publish_dir = 'raw_data_publish_dir'
if gfile.Exists(self._portal_input_base_dir):
gfile.DeleteRecursively(self._portal_input_base_dir)
gfile.MakeDirs(self._portal_input_base_dir)
self._data_fnames = ['1001/{}.data'.format(i) for i in range(100)]
self._data_fnames_without_success = \
['1002/{}.data'.format(i) for i in range(100)]
self._csv_fnames = ['1003/{}.csv'.format(i) for i in range(100)]
self._unused_fnames = ['{}.xx'.format(100)]
self._all_fnames = self._data_fnames + \
self._data_fnames_without_success + \
self._csv_fnames + self._unused_fnames
all_fnames_with_success = ['1001/_SUCCESS'] + ['1003/_SUCCESS'] +\
self._all_fnames
for fname in all_fnames_with_success:
fpath = os.path.join(self._portal_input_base_dir, fname)
gfile.MakeDirs(os.path.dirname(fpath))
with gfile.Open(fpath, "w") as f:
f.write('xxx')
def tearDown(self) -> None:
gfile.DeleteRecursively(self._portal_input_base_dir)
def _list_input_dir(self, portal_options, file_wildcard,
target_fnames, max_files_per_job=8000):
portal_manifest = dp_pb.DataPortalManifest(
name=self._data_portal_name,
data_portal_type=dp_pb.DataPortalType.Streaming,
output_partition_num=4,
input_file_wildcard=file_wildcard,
input_base_dir=self._portal_input_base_dir,
output_base_dir=self._portal_output_base_dir,
raw_data_publish_dir=self._raw_data_publish_dir,
processing_job_id=-1,
next_job_id=0
)
self._kvstore.set_data(
common.portal_kvstore_base_dir(self._data_portal_name),
text_format.MessageToString(portal_manifest))
with Timer("DataPortalJobManager initialization"):
data_portal_job_manager = DataPortalJobManager(
self._kvstore, self._data_portal_name,
portal_options.long_running,
portal_options.check_success_tag,
portal_options.single_subfolder,
portal_options.files_per_job_limit,
max_files_per_job
)
portal_job = data_portal_job_manager._sync_processing_job()
target_fnames.sort()
fpaths = [os.path.join(self._portal_input_base_dir, f)
for f in target_fnames]
self.assertEqual(len(fpaths), len(portal_job.fpaths))
for index, fpath in enumerate(fpaths):
self.assertEqual(fpath, portal_job.fpaths[index])
def test_list_input_dir(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=True,
single_subfolder=False,
files_per_job_limit=None
)
self._list_input_dir(portal_options, "*.data", self._data_fnames)
def test_list_input_dir_single_folder(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=True,
files_per_job_limit=None,
)
self._list_input_dir(
portal_options, "*.data", self._data_fnames)
def test_list_input_dir_files_limit(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
files_per_job_limit=1,
)
self._list_input_dir(
portal_options, "*.data", self._data_fnames)
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
files_per_job_limit=150,
)
self._list_input_dir(
portal_options, "*.data", self._data_fnames)
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
files_per_job_limit=200,
)
self._list_input_dir(
portal_options, "*.data",
self._data_fnames + self._data_fnames_without_success)
def test_list_input_dir_over_limit(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
)
self._list_input_dir(
portal_options, "*.data", self._data_fnames, max_files_per_job=100)
self._list_input_dir(
portal_options, "*.data",
self._data_fnames + self._data_fnames_without_success,
max_files_per_job=200)
def test_list_input_dir_without_success_check(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
files_per_job_limit=None
)
self._list_input_dir(
portal_options, "*.data",
self._data_fnames + self._data_fnames_without_success)
def test_list_input_dir_without_wildcard(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=True,
single_subfolder=False,
files_per_job_limit=None
)
self._list_input_dir(
portal_options, None,
self._data_fnames + self._csv_fnames)
def test_list_input_dir_without_wildcard_and_success_check(self):
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=False,
single_subfolder=False,
files_per_job_limit=None
)
self._list_input_dir(portal_options, None, self._all_fnames)
parser.add_argument('--raw_data_sub_dir',
type=str,
required=True,
help='the mysql base dir to subscribe new raw data')
args = parser.parse_args()
data_source = common_pb.DataSource()
data_source.data_source_meta.name = args.data_source_name
data_source.data_source_meta.partition_num = args.partition_num
data_source.data_source_meta.start_time = args.start_time
data_source.data_source_meta.end_time = args.end_time
data_source.data_source_meta.negative_sampling_rate = \
args.negative_sampling_rate
if args.role.upper() == 'LEADER':
data_source.role = common_pb.FLRole.Leader
else:
assert args.role.upper() == 'FOLLOWER'
data_source.role = common_pb.FLRole.Follower
data_source.output_base_dir = args.output_base_dir
data_source.raw_data_sub_dir = args.raw_data_sub_dir
data_source.state = common_pb.DataSourceState.Init
kvstore = DBClient(args.kvstore_type)
master_kvstore_key = common.data_source_kvstore_base_dir(
data_source.data_source_meta.name)
raw_data = kvstore.get_data(master_kvstore_key)
if raw_data is None:
logging.info("data source %s is not existed", args.data_source_name)
common.commit_data_source(kvstore, data_source)
logging.info("apply new data source %s", args.data_source_name)
else:
logging.info("data source %s has been existed", args.data_source_name)
def test_api(self):
logging.getLogger().setLevel(logging.DEBUG)
os.environ['ETCD_BASE_DIR'] = 'bytefl_l'
data_source_name = 'test_data_source'
kvstore_l = DBClient('etcd', True)
os.environ['ETCD_BASE_DIR'] = 'bytefl_f'
kvstore_f = DBClient('etcd', True)
kvstore_l.delete_prefix(
common.data_source_kvstore_base_dir(data_source_name))
kvstore_f.delete_prefix(
common.data_source_kvstore_base_dir(data_source_name))
data_source_l = common_pb.DataSource()
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"
data_source_f = common_pb.DataSource()
data_source_f.role = common_pb.FLRole.Follower
data_source_f.state = common_pb.DataSourceState.Init
data_source_f.output_base_dir = "./ds_output_f"
data_source_meta = common_pb.DataSourceMeta()
data_source_meta.name = data_source_name
data_source_meta.partition_num = 1
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(kvstore_l, data_source_l)
data_source_f.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(kvstore_f, data_source_f)
master_addr_l = 'localhost:4061'
master_addr_f = 'localhost:4062'
options = dj_pb.DataJoinMasterOptions(use_mock_etcd=True)
os.environ['ETCD_BASE_DIR'] = 'bytefl_l'
master_l = data_join_master.DataJoinMasterService(
int(master_addr_l.split(':')[1]), master_addr_f, data_source_name,
'etcd', options)
master_l.start()
os.environ['ETCD_BASE_DIR'] = 'bytefl_f'
master_f = data_join_master.DataJoinMasterService(
int(master_addr_f.split(':')[1]), master_addr_l, data_source_name,
'etcd', options)
master_f.start()
channel_l = make_insecure_channel(master_addr_l, ChannelType.INTERNAL)
client_l = dj_grpc.DataJoinMasterServiceStub(channel_l)
channel_f = make_insecure_channel(master_addr_f, ChannelType.INTERNAL)
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 = client_l.GetDataSourceStatus(req_l)
dss_f = 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)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=-1,
join_example=empty_pb2.Empty())
rdrsp = client_f.RequestJoinPartition(rdreq)
self.assertTrue(rdrsp.status.code == 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.join_example_rep.rank_id, 0)
self.assertEqual(rdrsp.manifest.join_example_rep.state,
dj_pb.JoinExampleState.Joining)
self.assertEqual(rdrsp.manifest.partition_id, 0)
#check idempotent
rdrsp = client_f.RequestJoinPartition(rdreq)
self.assertTrue(rdrsp.status.code == 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.join_example_rep.rank_id, 0)
self.assertEqual(rdrsp.manifest.join_example_rep.state,
dj_pb.JoinExampleState.Joining)
self.assertEqual(rdrsp.manifest.partition_id, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
join_example=empty_pb2.Empty())
rdrsp = client_l.RequestJoinPartition(rdreq)
self.assertTrue(rdrsp.status.code == 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.join_example_rep.rank_id, 0)
self.assertEqual(rdrsp.manifest.join_example_rep.state,
dj_pb.JoinExampleState.Joining)
self.assertEqual(rdrsp.manifest.partition_id, 0)
#check idempotent
rdrsp = client_l.RequestJoinPartition(rdreq)
self.assertTrue(rdrsp.status.code == 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.join_example_rep.rank_id, 0)
self.assertEqual(rdrsp.manifest.join_example_rep.state,
dj_pb.JoinExampleState.Joining)
self.assertEqual(rdrsp.manifest.partition_id, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=1,
partition_id=-1,
sync_example_id=empty_pb2.Empty())
rdrsp = client_l.RequestJoinPartition(rdreq)
self.assertEqual(rdrsp.status.code, 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.sync_example_id_rep.rank_id, 1)
self.assertEqual(rdrsp.manifest.sync_example_id_rep.state,
dj_pb.SyncExampleIdState.Syncing)
self.assertEqual(rdrsp.manifest.partition_id, 0)
#check idempotent
rdrsp = client_l.RequestJoinPartition(rdreq)
self.assertEqual(rdrsp.status.code, 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.sync_example_id_rep.rank_id, 1)
self.assertEqual(rdrsp.manifest.sync_example_id_rep.state,
dj_pb.SyncExampleIdState.Syncing)
self.assertEqual(rdrsp.manifest.partition_id, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=1,
partition_id=0,
sync_example_id=empty_pb2.Empty())
rdrsp = client_f.RequestJoinPartition(rdreq)
self.assertEqual(rdrsp.status.code, 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.sync_example_id_rep.rank_id, 1)
self.assertEqual(rdrsp.manifest.sync_example_id_rep.state,
dj_pb.SyncExampleIdState.Syncing)
self.assertEqual(rdrsp.manifest.partition_id, 0)
#check idempotent
rdrsp = client_f.RequestJoinPartition(rdreq)
self.assertEqual(rdrsp.status.code, 0)
self.assertTrue(rdrsp.HasField('manifest'))
self.assertEqual(rdrsp.manifest.sync_example_id_rep.rank_id, 1)
self.assertEqual(rdrsp.manifest.sync_example_id_rep.state,
dj_pb.SyncExampleIdState.Syncing)
self.assertEqual(rdrsp.manifest.partition_id, 0)
rdreq1 = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=1,
partition_id=0,
sync_example_id=empty_pb2.Empty())
try:
rsp = client_l.FinishJoinPartition(rdreq1)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
rdreq2 = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
join_example=empty_pb2.Empty())
try:
rsp = client_l.FinishJoinPartition(rdreq2)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
manifest_l = client_l.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_l is not None)
self.assertFalse(manifest_l.finished)
self.assertEqual(manifest_l.next_process_index, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_l.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 0)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=False,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='a',
timestamp=timestamp_pb2.Timestamp(seconds=3))
]))
rsp = client_l.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_l = client_l.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_l is not None)
self.assertFalse(manifest_l.finished)
self.assertEqual(manifest_l.next_process_index, 1)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_l.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 3)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=False,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='b',
timestamp=timestamp_pb2.Timestamp(seconds=5))
]))
rsp = client_l.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_l = client_l.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_l is not None)
self.assertFalse(manifest_l.finished)
self.assertEqual(manifest_l.next_process_index, 2)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_l.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 5)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=True,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='b',
timestamp=timestamp_pb2.Timestamp(seconds=5)),
dj_pb.RawDataMeta(
file_path='a',
timestamp=timestamp_pb2.Timestamp(seconds=3))
]))
rsp = client_l.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_l = client_l.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_l is not None)
self.assertFalse(manifest_l.finished)
self.assertEqual(manifest_l.next_process_index, 2)
rsp = client_l.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 5)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
)
manifest_f = client_f.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_f is not None)
self.assertFalse(manifest_f.finished)
self.assertEqual(manifest_f.next_process_index, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_f.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 0)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=False,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='a',
timestamp=timestamp_pb2.Timestamp(seconds=1))
]))
rsp = client_f.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_f = client_f.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_f is not None)
self.assertFalse(manifest_f.finished)
self.assertEqual(manifest_f.next_process_index, 1)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_f.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 1)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=False,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='b',
timestamp=timestamp_pb2.Timestamp(seconds=2))
]))
rsp = client_f.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_f = client_f.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_f is not None)
self.assertFalse(manifest_f.finished)
self.assertEqual(manifest_f.next_process_index, 2)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_f.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 2)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=True,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='a',
timestamp=timestamp_pb2.Timestamp(seconds=1)),
dj_pb.RawDataMeta(
file_path='b',
timestamp=timestamp_pb2.Timestamp(seconds=2))
]))
rsp = client_f.AddRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_f = client_f.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_f is not None)
self.assertFalse(manifest_f.finished)
self.assertEqual(manifest_f.next_process_index, 2)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
)
rsp = client_f.GetRawDataLatestTimeStamp(rdreq)
self.assertEqual(rsp.status.code, 0)
self.assertTrue(rsp.HasField('timestamp'))
self.assertEqual(rsp.timestamp.seconds, 2)
self.assertEqual(rsp.timestamp.nanos, 0)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
finish_raw_data=empty_pb2.Empty())
rsp = client_l.FinishRawData(rdreq)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_l.FinishRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_l = client_l.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_l is not None)
self.assertTrue(manifest_l.finished)
self.assertEqual(manifest_l.next_process_index, 2)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_l.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=False,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='x',
timestamp=timestamp_pb2.Timestamp(seconds=4))
]))
try:
rsp = client_l.AddRawData(rdreq)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
try:
rsp = client_f.FinishJoinPartition(rdreq2)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
rsp = client_l.FinishJoinPartition(rdreq1)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_l.FinishJoinPartition(rdreq1)
self.assertEqual(rsp.code, 0)
rsp = client_f.FinishJoinPartition(rdreq1)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_f.FinishJoinPartition(rdreq1)
self.assertEqual(rsp.code, 0)
try:
rsp = client_f.FinishJoinPartition(rdreq2)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
finish_raw_data=empty_pb2.Empty())
rsp = client_f.FinishRawData(rdreq)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_f.FinishRawData(rdreq)
self.assertEqual(rsp.code, 0)
manifest_f = client_f.QueryRawDataManifest(rdreq)
self.assertTrue(manifest_f is not None)
self.assertTrue(manifest_f.finished)
self.assertEqual(manifest_f.next_process_index, 2)
rdreq = dj_pb.RawDataRequest(
data_source_meta=data_source_f.data_source_meta,
rank_id=0,
partition_id=0,
added_raw_data_metas=dj_pb.AddedRawDataMetas(
dedup=True,
raw_data_metas=[
dj_pb.RawDataMeta(
file_path='x',
timestamp=timestamp_pb2.Timestamp(seconds=3))
]))
try:
rsp = client_f.AddRawData(rdreq)
except Exception as e:
self.assertTrue(True)
else:
self.assertTrue(False)
rsp = client_f.FinishJoinPartition(rdreq2)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_f.FinishJoinPartition(rdreq2)
self.assertEqual(rsp.code, 0)
rsp = client_l.FinishJoinPartition(rdreq2)
self.assertEqual(rsp.code, 0)
#check idempotent
rsp = client_l.FinishJoinPartition(rdreq2)
self.assertEqual(rsp.code, 0)
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 = client_l.GetDataSourceStatus(req_l)
dss_f = 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)
master_l.stop()
master_f.stop()
def __init__(self, data_source_name, kvstore_type, use_mock_etcd=False):
self._kvstore = DBClient(kvstore_type, use_mock_etcd)
self._data_source = retrieve_data_source(self._kvstore,
data_source_name)
class RsaPsi(unittest.TestCase):
def _setUpMySQL(self):
self.kvstore_type = 'etcd'
self.leader_base_dir = 'bytefl_l'
self.follower_base_dir = 'bytefl_f'
os.environ['ETCD_BASE_DIR'] = self.leader_base_dir
self._kvstore_l = DBClient(self.kvstore_type, True)
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
self._kvstore_f = DBClient(self.kvstore_type, True)
def _setUpDataSource(self):
self._data_source_name = 'test_data_source'
self._kvstore_l.delete_prefix(
common.data_source_kvstore_base_dir(self._data_source_name))
self._kvstore_f.delete_prefix(
common.data_source_kvstore_base_dir(self._data_source_name))
self._data_source_l = common_pb.DataSource()
self._data_source_l.role = common_pb.FLRole.Leader
self._data_source_l.state = common_pb.DataSourceState.Init
self._data_source_l.output_base_dir = "./ds_output_l"
self._raw_data_dir_l = "./raw_data_l"
self._data_source_l.raw_data_sub_dir = "./raw_data_sub_dir_l"
self._data_source_f = common_pb.DataSource()
self._data_source_f.role = common_pb.FLRole.Follower
self._data_source_f.state = common_pb.DataSourceState.Init
self._data_source_f.output_base_dir = "./ds_output_f"
self._raw_data_dir_f = "./raw_data_f"
self._data_source_f.raw_data_sub_dir = "./raw_data_sub_dir_f"
data_source_meta = common_pb.DataSourceMeta()
data_source_meta.name = self._data_source_name
data_source_meta.partition_num = 4
data_source_meta.start_time = 0
data_source_meta.end_time = 100000000
self._data_source_l.data_source_meta.MergeFrom(data_source_meta)
self._data_source_f.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(self._kvstore_l, self._data_source_l)
common.commit_data_source(self._kvstore_f, self._data_source_f)
def _generate_input_csv(self, cands, base_dir):
if not gfile.Exists(base_dir):
gfile.MakeDirs(base_dir)
fpaths = []
random.shuffle(cands)
csv_writers = []
partition_num = self._data_source_l.data_source_meta.partition_num
for partition_id in range(partition_num):
fpath = os.path.join(base_dir,
str(partition_id) + common.RawDataFileSuffix)
fpaths.append(fpath)
csv_writers.append(csv_dict_writer.CsvDictWriter(fpath))
for item in cands:
partition_id = CityHash32(item) % partition_num
raw = OrderedDict()
raw['raw_id'] = item
raw['feat_0'] = 'leader-' + str((partition_id << 30) + 0) + item
raw['feat_1'] = 'leader-' + str((partition_id << 30) + 1) + item
raw['feat_2'] = 'leader-' + str((partition_id << 30) + 2) + item
csv_writers[partition_id].write(raw)
for csv_writer in csv_writers:
csv_writer.close()
return fpaths
def _generate_input_tf_record(self, cands, base_dir):
if not gfile.Exists(base_dir):
gfile.MakeDirs(base_dir)
fpaths = []
random.shuffle(cands)
tfr_writers = []
partition_num = self._data_source_l.data_source_meta.partition_num
for partition_id in range(partition_num):
fpath = os.path.join(base_dir,
str(partition_id) + common.RawDataFileSuffix)
fpaths.append(fpath)
tfr_writers.append(tf.io.TFRecordWriter(fpath))
for item in cands:
partition_id = CityHash32(item) % partition_num
feat = {}
feat['raw_id'] = tf.train.Feature(bytes_list=tf.train.BytesList(
value=[item.encode()]))
f0 = 'follower' + str((partition_id << 30) + 0) + item
f1 = 'follower' + str((partition_id << 30) + 1) + item
f2 = 'follower' + str((partition_id << 30) + 2) + item
feat['feat_0'] = tf.train.Feature(bytes_list=tf.train.BytesList(
value=[f0.encode()]))
feat['feat_1'] = tf.train.Feature(bytes_list=tf.train.BytesList(
value=[f1.encode()]))
feat['feat_2'] = tf.train.Feature(bytes_list=tf.train.BytesList(
value=[f2.encode()]))
example = tf.train.Example(features=tf.train.Features(
feature=feat))
tfr_writers[partition_id].write(example.SerializeToString())
for tfr_writer in tfr_writers:
tfr_writer.close()
return fpaths
def _setUpRsaPsiConf(self):
self._input_dir_l = './rsa_psi_raw_input_l'
self._input_dir_f = './rsa_psi_raw_input_f'
self._pre_processor_ouput_dir_l = './pre_processor_output_dir_l'
self._pre_processor_ouput_dir_f = './pre_processor_output_dir_f'
key_dir = path.join(path.dirname(path.abspath(__file__)), '../rsa_key')
self._rsa_public_key_path = path.join(key_dir, 'rsa_psi.pub')
self._rsa_private_key_path = path.join(key_dir, 'rsa_psi')
self._raw_data_pub_dir_l = self._data_source_l.raw_data_sub_dir
self._raw_data_pub_dir_f = self._data_source_f.raw_data_sub_dir
def _gen_psi_input_raw_data(self):
self._intersection_ids = set(
['{:09}'.format(i) for i in range(0, 1 << 16) if i % 3 == 0])
self._rsa_raw_id_l = set([
'{:09}'.format(i) for i in range(0, 1 << 16) if i % 2 == 0
]) | self._intersection_ids
self._rsa_raw_id_f = set([
'{:09}'.format(i) for i in range(0, 1 << 16) if i % 2 == 1
]) | self._intersection_ids
self._input_dir_l = './rsa_psi_raw_input_l'
self._input_dir_f = './rsa_psi_raw_input_f'
self._psi_raw_data_fpaths_l = self._generate_input_csv(
list(self._rsa_raw_id_l), self._input_dir_l)
self._psi_raw_data_fpaths_f = self._generate_input_tf_record(
list(self._rsa_raw_id_f), self._input_dir_f)
def _remove_existed_dir(self):
if gfile.Exists(self._input_dir_l):
gfile.DeleteRecursively(self._input_dir_l)
if gfile.Exists(self._input_dir_f):
gfile.DeleteRecursively(self._input_dir_f)
if gfile.Exists(self._pre_processor_ouput_dir_l):
gfile.DeleteRecursively(self._pre_processor_ouput_dir_l)
if gfile.Exists(self._pre_processor_ouput_dir_f):
gfile.DeleteRecursively(self._pre_processor_ouput_dir_f)
if gfile.Exists(self._data_source_l.output_base_dir):
gfile.DeleteRecursively(self._data_source_l.output_base_dir)
if gfile.Exists(self._raw_data_dir_l):
gfile.DeleteRecursively(self._raw_data_dir_l)
if gfile.Exists(self._data_source_f.output_base_dir):
gfile.DeleteRecursively(self._data_source_f.output_base_dir)
if gfile.Exists(self._raw_data_dir_f):
gfile.DeleteRecursively(self._raw_data_dir_f)
def _launch_masters(self):
self._master_addr_l = 'localhost:4061'
self._master_addr_f = 'localhost:4062'
master_options = dj_pb.DataJoinMasterOptions(use_mock_etcd=True)
os.environ['ETCD_BASE_DIR'] = self.leader_base_dir
self._master_l = data_join_master.DataJoinMasterService(
int(self._master_addr_l.split(':')[1]), self._master_addr_f,
self._data_source_name, self.kvstore_type, master_options)
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
self._master_f = data_join_master.DataJoinMasterService(
int(self._master_addr_f.split(':')[1]), self._master_addr_l,
self._data_source_name, self.kvstore_type, master_options)
self._master_f.start()
self._master_l.start()
channel_l = make_insecure_channel(self._master_addr_l,
ChannelType.INTERNAL)
self._master_client_l = dj_grpc.DataJoinMasterServiceStub(channel_l)
channel_f = make_insecure_channel(self._master_addr_f,
ChannelType.INTERNAL)
self._master_client_f = dj_grpc.DataJoinMasterServiceStub(channel_f)
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.Processing and \
dss_f.state == common_pb.DataSourceState.Processing:
break
else:
time.sleep(2)
logging.info("masters turn into Processing state")
def _launch_workers(self):
worker_options_l = 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='SORT_RUN_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=1024, max_flying_item=4096),
data_block_builder_options=dj_pb.WriterOptions(
output_writer='CSV_DICT'))
worker_options_f = dj_pb.DataJoinWorkerOptions(
use_mock_etcd=True,
raw_data_options=dj_pb.RawDataOptions(raw_data_iter='CSV_DICT',
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='SORT_RUN_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=1024, max_flying_item=4096),
data_block_builder_options=dj_pb.WriterOptions(
output_writer='TF_RECORD'))
self._worker_addrs_l = [
'localhost:4161', 'localhost:4162', 'localhost:4163',
'localhost:4164'
]
self._worker_addrs_f = [
'localhost:5161', 'localhost:5162', 'localhost:5163',
'localhost:5164'
]
self._workers_l = []
self._workers_f = []
for rank_id in range(4):
worker_addr_l = self._worker_addrs_l[rank_id]
worker_addr_f = self._worker_addrs_f[rank_id]
os.environ['ETCD_BASE_DIR'] = self.leader_base_dir
self._workers_l.append(
data_join_worker.DataJoinWorkerService(
int(worker_addr_l.split(':')[1]), worker_addr_f,
self._master_addr_l, rank_id, self.kvstore_type,
worker_options_l))
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
self._workers_f.append(
data_join_worker.DataJoinWorkerService(
int(worker_addr_f.split(':')[1]), worker_addr_l,
self._master_addr_f, rank_id, self.kvstore_type,
worker_options_f))
for w in self._workers_l:
w.start()
for w in self._workers_f:
w.start()
def _launch_rsa_psi_signer(self):
self._rsa_psi_signer_addr = 'localhost:6171'
rsa_private_key_pem = None
with gfile.GFile(self._rsa_private_key_path, 'rb') as f:
rsa_private_key_pem = f.read()
rsa_private_key = rsa.PrivateKey.load_pkcs1(rsa_private_key_pem)
self._rsa_psi_signer = rsa_psi_signer.RsaPsiSigner(
rsa_private_key, 1, 500)
self._rsa_psi_signer.start(
int(self._rsa_psi_signer_addr.split(':')[1]), 512)
def _stop_workers(self):
for w in self._workers_f:
w.stop()
for w in self._workers_l:
w.stop()
def _stop_masters(self):
self._master_f.stop()
self._master_l.stop()
def _stop_rsa_psi_signer(self):
self._rsa_psi_signer.stop()
def setUp(self):
self._setUpMySQL()
self._setUpDataSource()
self._setUpRsaPsiConf()
self._remove_existed_dir()
self._gen_psi_input_raw_data()
self._launch_masters()
self._launch_workers()
self._launch_rsa_psi_signer()
def _preprocess_rsa_psi_leader(self):
processors = []
rsa_key_pem = None
with gfile.GFile(self._rsa_private_key_path, 'rb') as f:
rsa_key_pem = f.read()
for partition_id in range(
self._data_source_l.data_source_meta.partition_num):
options = dj_pb.RsaPsiPreProcessorOptions(
preprocessor_name='leader-rsa-psi-processor',
role=common_pb.FLRole.Leader,
rsa_key_pem=rsa_key_pem,
input_file_paths=[self._psi_raw_data_fpaths_l[partition_id]],
output_file_dir=self._pre_processor_ouput_dir_l,
raw_data_publish_dir=self._raw_data_pub_dir_l,
partition_id=partition_id,
offload_processor_number=1,
max_flying_sign_batch=128,
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='CSV_DICT',
read_ahead_size=1 << 20),
writer_options=dj_pb.WriterOptions(output_writer='TF_RECORD'))
os.environ['ETCD_BASE_DIR'] = self.leader_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 _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 test_all_pipeline(self):
start_tm = time.time()
self._preprocess_rsa_psi_follower()
logging.warning("Follower Preprocess cost %d seconds",
time.time() - start_tm)
start_tm = time.time()
self._preprocess_rsa_psi_leader()
logging.warning("Leader Preprocess cost %f seconds",
time.time() - start_tm)
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 tearDown(self):
self._stop_workers()
self._stop_masters()
self._stop_rsa_psi_signer()
self._remove_existed_dir()
import etcd3
from fedlearner.common.db_client import DBClient
from fedlearner.common.db_client import get_kvstore_config
MySQL_client = DBClient('mysql')
database, addr, username, password, base_dir = \
get_kvstore_config('etcd')
(host, port) = addr.split(':')
options = [('grpc.max_send_message_length', 2**31 - 1),
('grpc.max_receive_message_length', 2**31 - 1)]
clnt = etcd3.client(host=host, port=port, grpc_options=options)
for (data, key) in clnt.get_prefix('/', sort_order='ascend'):
if not isinstance(key.key, str):
key = key.key.decoder()
if not isinstance(data, str):
data = data.decoder()
MySQL_client.set_data(key, data)
def test_api(self):
logging.getLogger().setLevel(logging.DEBUG)
kvstore_type = 'etcd'
db_base_dir = 'dp_test'
os.environ['ETCD_BASE_DIR'] = db_base_dir
data_portal_name = 'test_data_source'
kvstore = DBClient(kvstore_type, True)
kvstore.delete_prefix(db_base_dir)
portal_input_base_dir = './portal_upload_dir'
portal_output_base_dir = './portal_output_dir'
raw_data_publish_dir = 'raw_data_publish_dir'
portal_manifest = dp_pb.DataPortalManifest(
name=data_portal_name,
data_portal_type=dp_pb.DataPortalType.Streaming,
output_partition_num=4,
input_file_wildcard="*.done",
input_base_dir=portal_input_base_dir,
output_base_dir=portal_output_base_dir,
raw_data_publish_dir=raw_data_publish_dir,
processing_job_id=-1,
next_job_id=0)
kvstore.set_data(common.portal_kvstore_base_dir(data_portal_name),
text_format.MessageToString(portal_manifest))
if gfile.Exists(portal_input_base_dir):
gfile.DeleteRecursively(portal_input_base_dir)
gfile.MakeDirs(portal_input_base_dir)
all_fnames = ['1001/{}.done'.format(i) for i in range(100)]
all_fnames.append('{}.xx'.format(100))
all_fnames.append('1001/_SUCCESS')
for fname in all_fnames:
fpath = os.path.join(portal_input_base_dir, fname)
gfile.MakeDirs(os.path.dirname(fpath))
with gfile.Open(fpath, "w") as f:
f.write('xxx')
portal_master_addr = 'localhost:4061'
portal_options = dp_pb.DataPotraMasterlOptions(
use_mock_etcd=True,
long_running=False,
check_success_tag=True,
)
data_portal_master = DataPortalMasterService(
int(portal_master_addr.split(':')[1]), data_portal_name,
kvstore_type, portal_options)
data_portal_master.start()
channel = make_insecure_channel(portal_master_addr,
ChannelType.INTERNAL)
portal_master_cli = dp_grpc.DataPortalMasterServiceStub(channel)
recv_manifest = portal_master_cli.GetDataPortalManifest(
empty_pb2.Empty())
self.assertEqual(recv_manifest.name, portal_manifest.name)
self.assertEqual(recv_manifest.data_portal_type,
portal_manifest.data_portal_type)
self.assertEqual(recv_manifest.output_partition_num,
portal_manifest.output_partition_num)
self.assertEqual(recv_manifest.input_file_wildcard,
portal_manifest.input_file_wildcard)
self.assertEqual(recv_manifest.input_base_dir,
portal_manifest.input_base_dir)
self.assertEqual(recv_manifest.output_base_dir,
portal_manifest.output_base_dir)
self.assertEqual(recv_manifest.raw_data_publish_dir,
portal_manifest.raw_data_publish_dir)
self.assertEqual(recv_manifest.next_job_id, 1)
self.assertEqual(recv_manifest.processing_job_id, 0)
self._check_portal_job(kvstore, all_fnames, portal_manifest, 0)
mapped_partition = set()
task_0 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=0))
task_0_1 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=0))
self.assertEqual(task_0, task_0_1)
self.assertTrue(task_0.HasField('map_task'))
mapped_partition.add(task_0.map_task.partition_id)
self._check_map_task(task_0.map_task, all_fnames,
task_0.map_task.partition_id, portal_manifest)
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(rank_id=0,
partition_id=task_0.map_task.partition_id,
part_state=dp_pb.PartState.kIdMap))
task_1 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=0))
self.assertTrue(task_1.HasField('map_task'))
mapped_partition.add(task_1.map_task.partition_id)
self._check_map_task(task_1.map_task, all_fnames,
task_1.map_task.partition_id, portal_manifest)
task_2 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=1))
self.assertTrue(task_2.HasField('map_task'))
mapped_partition.add(task_2.map_task.partition_id)
self._check_map_task(task_2.map_task, all_fnames,
task_2.map_task.partition_id, portal_manifest)
task_3 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=2))
self.assertTrue(task_3.HasField('map_task'))
mapped_partition.add(task_3.map_task.partition_id)
self._check_map_task(task_3.map_task, all_fnames,
task_3.map_task.partition_id, portal_manifest)
self.assertEqual(len(mapped_partition),
portal_manifest.output_partition_num)
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(rank_id=0,
partition_id=task_1.map_task.partition_id,
part_state=dp_pb.PartState.kIdMap))
pending_1 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=4))
self.assertTrue(pending_1.HasField('pending'))
pending_2 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=3))
self.assertTrue(pending_2.HasField('pending'))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(rank_id=1,
partition_id=task_2.map_task.partition_id,
part_state=dp_pb.PartState.kIdMap))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(rank_id=2,
partition_id=task_3.map_task.partition_id,
part_state=dp_pb.PartState.kIdMap))
reduce_partition = set()
task_4 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=0))
task_4_1 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=0))
self.assertEqual(task_4, task_4_1)
self.assertTrue(task_4.HasField('reduce_task'))
reduce_partition.add(task_4.reduce_task.partition_id)
self._check_reduce_task(task_4.reduce_task,
task_4.reduce_task.partition_id,
portal_manifest)
task_5 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=1))
self.assertTrue(task_5.HasField('reduce_task'))
reduce_partition.add(task_5.reduce_task.partition_id)
self._check_reduce_task(task_5.reduce_task,
task_5.reduce_task.partition_id,
portal_manifest)
task_6 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=2))
self.assertTrue(task_6.HasField('reduce_task'))
reduce_partition.add(task_6.reduce_task.partition_id)
self._check_reduce_task(task_6.reduce_task,
task_6.reduce_task.partition_id,
portal_manifest)
task_7 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=3))
self.assertTrue(task_7.HasField('reduce_task'))
reduce_partition.add(task_7.reduce_task.partition_id)
self.assertEqual(len(reduce_partition), 4)
self._check_reduce_task(task_7.reduce_task,
task_7.reduce_task.partition_id,
portal_manifest)
task_8 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=5))
self.assertTrue(task_8.HasField('pending'))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(
rank_id=0,
partition_id=task_4.reduce_task.partition_id,
part_state=dp_pb.PartState.kEventTimeReduce))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(
rank_id=1,
partition_id=task_5.reduce_task.partition_id,
part_state=dp_pb.PartState.kEventTimeReduce))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(
rank_id=2,
partition_id=task_6.reduce_task.partition_id,
part_state=dp_pb.PartState.kEventTimeReduce))
portal_master_cli.FinishTask(
dp_pb.FinishTaskRequest(
rank_id=3,
partition_id=task_7.reduce_task.partition_id,
part_state=dp_pb.PartState.kEventTimeReduce))
task_9 = portal_master_cli.RequestNewTask(
dp_pb.NewTaskRequest(rank_id=5))
self.assertTrue(task_9.HasField('finished'))
data_portal_master.stop()
gfile.DeleteRecursively(portal_input_base_dir)
def setUp(self):
self.kvstore_type = 'etcd'
self.leader_base_dir = 'bytefl_l'
self.follower_base_dir = 'bytefl_f'
data_source_name = 'test_data_source'
os.environ['ETCD_BASE_DIR'] = self.leader_base_dir
kvstore_l = DBClient(self.kvstore_type, True)
os.environ['ETCD_BASE_DIR'] = self.follower_base_dir
kvstore_f = DBClient(self.kvstore_type, True)
kvstore_l.delete_prefix(
common.data_source_kvstore_base_dir(data_source_name))
kvstore_f.delete_prefix(
common.data_source_kvstore_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(kvstore_l, data_source_l)
data_source_f.data_source_meta.MergeFrom(data_source_meta)
common.commit_data_source(kvstore_f, data_source_f)
self.kvstore_l = kvstore_l
self.kvstore_f = kvstore_f
self.data_source_l = data_source_l
self.data_source_f = data_source_f
self.data_source_name = data_source_name
self.raw_data_publisher_l = raw_data_publisher.RawDataPublisher(
self.kvstore_l, self.raw_data_pub_dir_l)
self.raw_data_publisher_f = raw_data_publisher.RawDataPublisher(
self.kvstore_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,
optional_fields=['label']),
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
