matrixorigin
[Bug]: vector index:create index which lists=1000 reported file 018e79d2-44ae-714c-9acf-15798c49c41b_00000 is not found
Is there an existing issue for the same bug?
- [X] I have checked the existing issues.
Branch Name
main
Commit ID
b85d7579eec4ca041b63ecf8cf9ceef325f59071
Other Environment Information
- Hardware parameters:
- OS type:
- Others:
Actual Behavior
gist 960dim 100M create index lists=1000 reported error file 018e79d2-44ae-714c-9acf-15798c49c41b_00000 is not found
Expected Behavior
No response
Steps to Reproduce
1.create table t3(a int, b vecf32(960));
1.use benchmark tool load gist 960dim 100M data
2.create index idx3 using ivfflat on t3(b) lists=1000 op_type "vector_l2_ops";
Additional information
No response
Internally "CREATE INDEX" is running only 3 SQL queries.
I think it is similar to this issue: https://github.com/matrixorigin/matrixone/issues/14529
@triump2020 could you kindly have a look at this?
- Create the database and table
create database a;
use a;
create table t3(a int, b vecf32(960));
-
Download GIST dataset and unzip http://corpus-texmex.irisa.fr/
-
Run the python code to insert GIST960 data. Replace
/Users/arjunsunilkumar/Downloads/benchmark/1million/gist/gist_base.fvecs
import binascii
import time
import numpy as np
from sqlalchemy import create_engine, text
from sqlalchemy.orm import sessionmaker
def to_db_binary(value):
if value is None:
return value
value = np.asarray(value, dtype='<f4')
if value.ndim != 1:
raise ValueError('expected ndim to be 1')
return binascii.b2a_hex(value.tobytes()).decode('utf-8')
def fvecs_read(filename, c_contiguous=True):
fv = np.fromfile(filename, dtype=np.float32)
if fv.size == 0:
return np.zeros((0, 0))
dim = fv.view(np.int32)[0]
assert dim > 0
fv = fv.reshape(-1, 1 + dim)
fv = fv[:, 1:]
if c_contiguous:
fv = np.ascontiguousarray(fv)
return fv
def run():
# use pymysql
engine = create_engine("mysql+pymysql://root:[email protected]:6001/a")
Session = sessionmaker(bind=engine)
session = Session()
sql_insert = text("INSERT INTO t5 (a, b) VALUES (:id, decode(:data,'hex'));")
start = time.time()
vecList = fvecs_read("/Users/arjunsunilkumar/Downloads/benchmark/1million/gist/gist_base.fvecs")
binVecList = []
for i in range(0, len(vecList)):
binVecList.append(to_db_binary(vecList[i]))
print(f"binary duration={time.time() - start}")
for i in range(0, len(binVecList)):
session.execute(sql_insert, {"id": i, "data": binVecList[i]})
if i % 1000 == 0:
print(f"inserted {i}")
# commit last
session.commit()
def main():
start = time.time()
run()
duration = time.time() - start
print(f"duration={duration}")
if __name__ == "__main__":
main()
- Create index using this query
create index idx3 using ivfflat on t3(b) lists=1000 op_type "vector_l2_ops";
reporduced on local, the not found is caused by GC delete. and this object was merged before that deletes.
grep "018f4cd6-8533-76db-9005-bb00a47c7cb7_00000" matrixone/mo-service.log
{"level":"INFO","time":"2024/05/06 15:59:36.115022 +0800","caller":"merge/executor.go:231","msg":"[Mergeblocks] Scheduled 272505-t3 [t1224|on2,bn2|1.99MB,11.95MB], merged(538): 269(3cfb75799ece) 269(bb00a47c7cb7)"}
{"level":"INFO","time":"2024/05/06 15:59:36.115169 +0800","caller":"mergesort/task.go:107","msg":"[Start] Mergeblocks","table":"t3","txn-start-ts":"1714982376115110051-0","from-objs":"3cfb75799ece,bb00a47c7cb7,"}
{"level":"INFO","time":"2024/05/06 17:02:41.595889 +0800","caller":"gc/delete.go:70","msg":"[DB GC] disableGC: false, files to delete: [018f4cd6-90ab-7017-9561-9a6d307b84d5_00000 018f4cd6-8533-76db-9005-bb00a47c7cb7_00000 018f4ce8-ff67-75dd-a631-89144e5f221f_00000 018f4ce8-f0b5-7482-a28d-eaff168ce745_00000 ...
{"level":"ERROR","time":"2024/05/06 17:04:02.164434 +0800","caller":"fileservice/local_fs.go:435","msg":"error: file 018f4cd6-8533-76db-9005-bb00a47c7cb7_00000 is not found"}
{"level":"ERROR","time":"2024/05/06 17:28:23.583578 +0800","name":"log-service.frontend","caller":"frontend/mysql_cmd_executor.go:3202","msg":"file 018f4cd6-8533-76db-9005-bb00a47c7cb7_00000 is not found","uuid":"7c4dccb4-4d3c-41f8-b482-5251dc7a41bf","session_info":"connectionId 1001|127.0.0.1:16722|{account sys:dump:moadmin -- 0:1:0}|goRoutineId 580147|018f4ce9-443a-7ec4-8fcb-bb310673af9e","session_id":"018f4ce9-443a-7ec4-8fcb-bb310673af9e","statement_id":"018f4ce9-bf37-7471-bce3-02a5dd3d7fb9","txn_id":"018f4ce9bf37786a969e56e99819076a/Active/S:1714982338357691859-1"}
the log mo-service (1).log.zip
the ts of that aborted txn is smaller than the merge ts, so this not found err is not expected.
1714982338357691859-1
1714982376115110051-0 merge ts
this created index txn lasted too long, exceeded 1h.
