milvus/shards/mishards/service_handler.py
XuPeng-SH 0a134dab82 Mishards synced to 0.6.0 release (#725)
* [skip ci](shards): update to support milvus 0.6.0

* [skip ci](shards): add show partitions

* [skip ci](shards): add drop partition

* [skip ci](shards): add create partition

* [skip ci](shards): update search for partition

* [skip ci](shards): fix search partion bugs

* [skip ci](shards): update all in one demo

* [skip ci](shards/doc): update image in readme
2019-12-09 10:55:32 +08:00

548 lines
21 KiB
Python

import logging
import time
import datetime
from collections import defaultdict
import multiprocessing
from concurrent.futures import ThreadPoolExecutor
from milvus.grpc_gen import milvus_pb2, milvus_pb2_grpc, status_pb2
from milvus.grpc_gen.milvus_pb2 import TopKQueryResult
from milvus.client.abstract import Range
from milvus.client import types as Types
from mishards import (db, settings, exceptions)
from mishards.grpc_utils import mark_grpc_method
from mishards.grpc_utils.grpc_args_parser import GrpcArgsParser as Parser
from mishards import utilities
logger = logging.getLogger(__name__)
class ServiceHandler(milvus_pb2_grpc.MilvusServiceServicer):
MAX_NPROBE = 2048
MAX_TOPK = 2048
def __init__(self, tracer, router, max_workers=multiprocessing.cpu_count(), **kwargs):
self.table_meta = {}
self.error_handlers = {}
self.tracer = tracer
self.router = router
self.max_workers = max_workers
def _reduce(self, source_ids, ids, source_diss, diss, k, reverse):
if source_diss[k - 1] <= diss[0]:
return source_ids, source_diss
if diss[k - 1] <= source_diss[0]:
return ids, diss
source_diss.extend(diss)
diss_t = enumerate(source_diss)
diss_m_rst = sorted(diss_t, key=lambda x: x[1])[:k]
diss_m_out = [id_ for _, id_ in diss_m_rst]
source_ids.extend(ids)
id_m_out = [source_ids[i] for i, _ in diss_m_rst]
return id_m_out, diss_m_out
def _do_merge(self, files_n_topk_results, topk, reverse=False, **kwargs):
status = status_pb2.Status(error_code=status_pb2.SUCCESS,
reason="Success")
if not files_n_topk_results:
return status, [], []
merge_id_results = []
merge_dis_results = []
calc_time = time.time()
for files_collection in files_n_topk_results:
if isinstance(files_collection, tuple):
status, _ = files_collection
return status, [], []
row_num = files_collection.row_num
# row_num is equal to 0, result is empty
if not row_num:
continue
ids = files_collection.ids
diss = files_collection.distances # distance collections
# TODO: batch_len is equal to topk, may need to compare with topk
batch_len = len(ids) // row_num
for row_index in range(row_num):
id_batch = ids[row_index * batch_len: (row_index + 1) * batch_len]
dis_batch = diss[row_index * batch_len: (row_index + 1) * batch_len]
if len(merge_id_results) < row_index:
raise ValueError("merge error")
elif len(merge_id_results) == row_index:
# TODO: may bug here
merge_id_results.append(id_batch)
merge_dis_results.append(dis_batch)
else:
merge_id_results[row_index], merge_dis_results[row_index] = \
self._reduce(merge_id_results[row_index], id_batch,
merge_dis_results[row_index], dis_batch,
batch_len,
reverse)
calc_time = time.time() - calc_time
logger.info('Merge takes {}'.format(calc_time))
id_mrege_list = []
dis_mrege_list = []
for id_results, dis_results in zip(merge_id_results, merge_dis_results):
id_mrege_list.extend(id_results)
dis_mrege_list.extend(dis_results)
return status, id_mrege_list, dis_mrege_list
def _do_query(self,
context,
table_id,
table_meta,
vectors,
topk,
nprobe,
range_array=None,
partition_tags=None,
**kwargs):
metadata = kwargs.get('metadata', None)
range_array = [
utilities.range_to_date(r, metadata=metadata) for r in range_array
] if range_array else None
routing = {}
p_span = None if self.tracer.empty else context.get_active_span(
).context
with self.tracer.start_span('get_routing', child_of=p_span):
routing = self.router.routing(table_id,
range_array=range_array,
partition_tags=partition_tags,
metadata=metadata)
logger.info('Routing: {}'.format(routing))
metadata = kwargs.get('metadata', None)
rs = []
all_topk_results = []
def search(addr, query_params, vectors, topk, nprobe, **kwargs):
logger.info(
'Send Search Request: addr={};params={};nq={};topk={};nprobe={}'
.format(addr, query_params, len(vectors), topk, nprobe))
conn = self.router.query_conn(addr, metadata=metadata)
start = time.time()
span = kwargs.get('span', None)
span = span if span else (None if self.tracer.empty else
context.get_active_span().context)
with self.tracer.start_span('search_{}'.format(addr),
child_of=span):
ret = conn.search_vectors_in_files(table_name=query_params['table_id'],
file_ids=query_params['file_ids'],
query_records=vectors,
top_k=topk,
nprobe=nprobe)
end = time.time()
all_topk_results.append(ret)
with self.tracer.start_span('do_search', child_of=p_span) as span:
with ThreadPoolExecutor(max_workers=self.max_workers) as pool:
for addr, params in routing.items():
res = pool.submit(search,
addr,
params,
vectors,
topk,
nprobe,
span=span)
rs.append(res)
for res in rs:
res.result()
reverse = table_meta.metric_type == Types.MetricType.IP
with self.tracer.start_span('do_merge', child_of=p_span):
return self._do_merge(all_topk_results,
topk,
reverse=reverse,
metadata=metadata)
def _create_table(self, table_schema):
return self.router.connection().create_table(table_schema)
@mark_grpc_method
def CreateTable(self, request, context):
_status, _table_schema = Parser.parse_proto_TableSchema(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
logger.info('CreateTable {}'.format(_table_schema['table_name']))
_status = self._create_table(_table_schema)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
def _has_table(self, table_name, metadata=None):
return self.router.connection(metadata=metadata).has_table(table_name)
@mark_grpc_method
def HasTable(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return milvus_pb2.BoolReply(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
bool_reply=False)
logger.info('HasTable {}'.format(_table_name))
_status, _bool = self._has_table(_table_name,
metadata={'resp_class': milvus_pb2.BoolReply})
return milvus_pb2.BoolReply(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
bool_reply=_bool)
@mark_grpc_method
def CreatePartition(self, request, context):
_table_name, _partition_name, _tag = Parser.parse_proto_PartitionParam(request)
_status = self.router.connection().create_partition(_table_name, _partition_name, _tag)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
@mark_grpc_method
def DropPartition(self, request, context):
_table_name, _partition_name, _tag = Parser.parse_proto_PartitionParam(request)
_status = self.router.connection().drop_partition(_table_name, _tag)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
@mark_grpc_method
def ShowPartitions(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return milvus_pb2.PartitionList(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
partition_array=[])
logger.info('ShowPartitions {}'.format(_table_name))
_status, partition_array = self.router.connection().show_partitions(_table_name)
return milvus_pb2.PartitionList(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
partition_array=[milvus_pb2.PartitionParam(table_name=param.table_name,
tag=param.tag,
partition_name=param.partition_name)
for param in partition_array])
def _delete_table(self, table_name):
return self.router.connection().delete_table(table_name)
@mark_grpc_method
def DropTable(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
logger.info('DropTable {}'.format(_table_name))
_status = self._delete_table(_table_name)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
def _create_index(self, table_name, index):
return self.router.connection().create_index(table_name, index)
@mark_grpc_method
def CreateIndex(self, request, context):
_status, unpacks = Parser.parse_proto_IndexParam(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
_table_name, _index = unpacks
logger.info('CreateIndex {}'.format(_table_name))
# TODO: interface create_table incompleted
_status = self._create_index(_table_name, _index)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
def _add_vectors(self, param, metadata=None):
return self.router.connection(metadata=metadata).add_vectors(
None, None, insert_param=param)
@mark_grpc_method
def Insert(self, request, context):
logger.info('Insert')
# TODO: Ths SDK interface add_vectors() could update, add a key 'row_id_array'
_status, _ids = self._add_vectors(
metadata={'resp_class': milvus_pb2.VectorIds}, param=request)
return milvus_pb2.VectorIds(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
vector_id_array=_ids)
@mark_grpc_method
def Search(self, request, context):
table_name = request.table_name
topk = request.topk
nprobe = request.nprobe
logger.info('Search {}: topk={} nprobe={}'.format(
table_name, topk, nprobe))
metadata = {'resp_class': milvus_pb2.TopKQueryResult}
if nprobe > self.MAX_NPROBE or nprobe <= 0:
raise exceptions.InvalidArgumentError(
message='Invalid nprobe: {}'.format(nprobe), metadata=metadata)
if topk > self.MAX_TOPK or topk <= 0:
raise exceptions.InvalidTopKError(
message='Invalid topk: {}'.format(topk), metadata=metadata)
table_meta = self.table_meta.get(table_name, None)
if not table_meta:
status, info = self.router.connection(
metadata=metadata).describe_table(table_name)
if not status.OK():
raise exceptions.TableNotFoundError(table_name,
metadata=metadata)
self.table_meta[table_name] = info
table_meta = info
start = time.time()
query_record_array = []
for query_record in request.query_record_array:
query_record_array.append(list(query_record.vector_data))
query_range_array = []
for query_range in request.query_range_array:
query_range_array.append(
Range(query_range.start_value, query_range.end_value))
status, id_results, dis_results = self._do_query(context,
table_name,
table_meta,
query_record_array,
topk,
nprobe,
query_range_array,
partition_tags=getattr(request, "partition_tag_array", []),
metadata=metadata)
now = time.time()
# logger.info('SearchVector takes: {}'.format(now - start))
topk_result_list = milvus_pb2.TopKQueryResult(
status=status_pb2.Status(error_code=status.error_code,
reason=status.reason),
row_num=len(request.query_record_array) if len(id_results) else 0,
ids=id_results,
distances=dis_results)
return topk_result_list
@mark_grpc_method
def SearchInFiles(self, request, context):
raise NotImplemented()
def _describe_table(self, table_name, metadata=None):
return self.router.connection(metadata=metadata).describe_table(table_name)
@mark_grpc_method
def DescribeTable(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return milvus_pb2.TableSchema(status=status_pb2.Status(
error_code=_status.code, reason=_status.message), )
metadata = {'resp_class': milvus_pb2.TableSchema}
logger.info('DescribeTable {}'.format(_table_name))
_status, _table = self._describe_table(metadata=metadata,
table_name=_table_name)
if _status.OK():
return milvus_pb2.TableSchema(
table_name=_table_name,
index_file_size=_table.index_file_size,
dimension=_table.dimension,
metric_type=_table.metric_type,
status=status_pb2.Status(error_code=_status.code,
reason=_status.message),
)
return milvus_pb2.TableSchema(
table_name=_table_name,
status=status_pb2.Status(error_code=_status.code,
reason=_status.message),
)
def _count_table(self, table_name, metadata=None):
return self.router.connection(
metadata=metadata).get_table_row_count(table_name)
@mark_grpc_method
def CountTable(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
status = status_pb2.Status(error_code=_status.code,
reason=_status.message)
return milvus_pb2.TableRowCount(status=status)
logger.info('CountTable {}'.format(_table_name))
metadata = {'resp_class': milvus_pb2.TableRowCount}
_status, _count = self._count_table(_table_name, metadata=metadata)
return milvus_pb2.TableRowCount(
status=status_pb2.Status(error_code=_status.code,
reason=_status.message),
table_row_count=_count if isinstance(_count, int) else -1)
def _get_server_version(self, metadata=None):
return self.router.connection(metadata=metadata).server_version()
@mark_grpc_method
def Cmd(self, request, context):
_status, _cmd = Parser.parse_proto_Command(request)
logger.info('Cmd: {}'.format(_cmd))
if not _status.OK():
return milvus_pb2.StringReply(status=status_pb2.Status(
error_code=_status.code, reason=_status.message))
metadata = {'resp_class': milvus_pb2.StringReply}
if _cmd == 'version':
_status, _reply = self._get_server_version(metadata=metadata)
else:
_status, _reply = self.router.connection(
metadata=metadata).server_status()
return milvus_pb2.StringReply(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
string_reply=_reply)
def _show_tables(self, metadata=None):
return self.router.connection(metadata=metadata).show_tables()
@mark_grpc_method
def ShowTables(self, request, context):
logger.info('ShowTables')
metadata = {'resp_class': milvus_pb2.TableName}
_status, _results = self._show_tables(metadata=metadata)
return milvus_pb2.TableNameList(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
table_names=_results)
def _delete_by_range(self, table_name, start_date, end_date):
return self.router.connection().delete_vectors_by_range(table_name,
start_date,
end_date)
@mark_grpc_method
def DeleteByRange(self, request, context):
_status, unpacks = \
Parser.parse_proto_DeleteByRangeParam(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
_table_name, _start_date, _end_date = unpacks
logger.info('DeleteByRange {}: {} {}'.format(_table_name, _start_date,
_end_date))
_status = self._delete_by_range(_table_name, _start_date, _end_date)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
def _preload_table(self, table_name):
return self.router.connection().preload_table(table_name)
@mark_grpc_method
def PreloadTable(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
logger.info('PreloadTable {}'.format(_table_name))
_status = self._preload_table(_table_name)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
def _describe_index(self, table_name, metadata=None):
return self.router.connection(metadata=metadata).describe_index(table_name)
@mark_grpc_method
def DescribeIndex(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return milvus_pb2.IndexParam(status=status_pb2.Status(
error_code=_status.code, reason=_status.message))
metadata = {'resp_class': milvus_pb2.IndexParam}
logger.info('DescribeIndex {}'.format(_table_name))
_status, _index_param = self._describe_index(table_name=_table_name,
metadata=metadata)
if not _index_param:
return milvus_pb2.IndexParam(status=status_pb2.Status(
error_code=_status.code, reason=_status.message))
_index = milvus_pb2.Index(index_type=_index_param._index_type,
nlist=_index_param._nlist)
return milvus_pb2.IndexParam(status=status_pb2.Status(
error_code=_status.code, reason=_status.message),
table_name=_table_name,
index=_index)
def _drop_index(self, table_name):
return self.router.connection().drop_index(table_name)
@mark_grpc_method
def DropIndex(self, request, context):
_status, _table_name = Parser.parse_proto_TableName(request)
if not _status.OK():
return status_pb2.Status(error_code=_status.code,
reason=_status.message)
logger.info('DropIndex {}'.format(_table_name))
_status = self._drop_index(_table_name)
return status_pb2.Status(error_code=_status.code,
reason=_status.message)