def get_grpc_server(user_model,
seldon_metrics,
annotations={},
trace_interceptor=None):
seldon_model = SeldonModelGRPC(user_model, seldon_metrics)
options = []
if ANNOTATION_GRPC_MAX_MSG_SIZE in annotations:
max_msg = int(annotations[ANNOTATION_GRPC_MAX_MSG_SIZE])
logger.info("Setting grpc max message and receive length to %d",
max_msg)
options.append(("grpc.max_message_length", max_msg))
options.append(("grpc.max_send_message_length", max_msg))
options.append(("grpc.max_receive_message_length", max_msg))
server = grpc.server(futures.ThreadPoolExecutor(max_workers=10),
options=options)
if trace_interceptor:
from grpc_opentracing.grpcext import intercept_server
server = intercept_server(server, trace_interceptor)
prediction_pb2_grpc.add_GenericServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_ModelServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_TransformerServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_OutputTransformerServicer_to_server(
seldon_model, server)
prediction_pb2_grpc.add_CombinerServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_RouterServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_SeldonServicer_to_server(seldon_model, server)
return server
def get_grpc_server(user_model, debug=False, annotations={}):
seldon_model = SeldonTransformerGRPC(user_model)
options = []
if ANNOTATION_GRPC_MAX_MSG_SIZE in annotations:
max_msg = int(annotations[ANNOTATION_GRPC_MAX_MSG_SIZE])
logger.info("Setting grpc max message to %d", max_msg)
options.append(('grpc.max_message_length', max_msg))
server = grpc.server(futures.ThreadPoolExecutor(
max_workers=10), options=options)
prediction_pb2_grpc.add_ModelServicer_to_server(seldon_model, server)
return server
def get_grpc_server(user_model,
seldon_metrics,
annotations={},
trace_interceptor=None,
num_threads=1):
seldon_model = SeldonModelGRPC(user_model, seldon_metrics)
options = [
("grpc.so_reuseport", 1),
]
if ANNOTATION_GRPC_MAX_MSG_SIZE in annotations:
max_msg = int(annotations[ANNOTATION_GRPC_MAX_MSG_SIZE])
logger.info("Setting grpc max message and receive length to %d",
max_msg)
options.append(("grpc.max_message_length", max_msg))
options.append(("grpc.max_send_message_length", max_msg))
options.append(("grpc.max_receive_message_length", max_msg))
server = grpc.server(futures.ThreadPoolExecutor(max_workers=num_threads),
options=options)
if trace_interceptor:
from grpc_opentracing.grpcext import intercept_server
server = intercept_server(server, trace_interceptor)
prediction_pb2_grpc.add_GenericServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_ModelServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_TransformerServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_OutputTransformerServicer_to_server(
seldon_model, server)
prediction_pb2_grpc.add_CombinerServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_RouterServicer_to_server(seldon_model, server)
prediction_pb2_grpc.add_SeldonServicer_to_server(seldon_model, server)
SERVICE_NAMES = (
prediction_pb2.DESCRIPTOR.services_by_name["Generic"].full_name,
prediction_pb2.DESCRIPTOR.services_by_name["Model"].full_name,
prediction_pb2.DESCRIPTOR.services_by_name["Router"].full_name,
prediction_pb2.DESCRIPTOR.services_by_name["Transformer"].full_name,
prediction_pb2.DESCRIPTOR.services_by_name["OutputTransformer"].
full_name,
prediction_pb2.DESCRIPTOR.services_by_name["Combiner"].full_name,
prediction_pb2.DESCRIPTOR.services_by_name["Seldon"].full_name,
reflection.SERVICE_NAME,
)
reflection.enable_server_reflection(SERVICE_NAMES, server)
return server
