Record has changed since last read in table study_system_attributes [mariadb] #6065
Description
Expected behavior
When running an optimization using NSGAII and VSBXCrossover, I expect it to run without database errors. The NSGAII initialization is as follows:
sampler = optuna.samplers.NSGAIISampler(
population_size=150,
mutation_prob=0.08,
crossover_prob=0.85,
swapping_prob=0.5,
seed=_seed,
crossover=optuna.samplers.nsgaii.VSBXCrossover(eta=15),
constraints_func=self.constraints_opti if self.constraints else None
)and database configuration below:
storage = optuna.storages.RDBStorage(url=self.db_url, heartbeat_interval=80, grace_period=160,
failed_trial_callback=RetryFailedTrialCallback(max_retry=5))Environment
- Optuna version: 4.2.1
- Python version: 3.13.3
- OS: Linux-6.12.21-amd64-x86_64-with-glibc2.41 (Debian 12)
Error messages, stack traces, or logs
ile "/home/jyck/.pyenv/versions/miniforge3-latest/envs/py313/lib/python3.13/site-packages/pymysql/connections.py", line 1199, in read
first_packet = self.connection._read_packet()
File "/home/jyck/.pyenv/versions/miniforge3-latest/envs/py313/lib/python3.13/site-packages/pymysql/connections.py", line 775, in _read_packet
packet.raise_for_error()
~~~~~~~~~~~~~~~~~~~~~~^^
File "/home/jyck/.pyenv/versions/miniforge3-latest/envs/py313/lib/python3.13/site-packages/pymysql/protocol.py", line 219, in raise_for_error
err.raise_mysql_exception(self._data)
~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^
File "/home/jyck/.pyenv/versions/miniforge3-latest/envs/py313/lib/python3.13/site-packages/pymysql/err.py", line 150, in raise_mysql_exception
raise errorclass(errno, errval)
sqlalchemy.exc.OperationalError: (pymysql.err.OperationalError) (1020, "Record has changed since last read in table 'study_system_attributes'")
[SQL: UPDATE study_system_attributes SET value_json=%(value_json)s WHERE study_system_attributes.study_system_attribute_id = %(study_system_attributes_study_system_attribute_id)s]
[parameters: {'value_json': '[5, [837, 877, 888, 943, 710, 871, 880, 908, 936, 937, 953, 357, 346, 853, 856, 869, 890, 934, 720, 802, 538, 652, 815, 513, 673, 688, 826, 20, 833, ... (439 characters truncated) ... 64, 462, 460, 454, 453, 450, 448, 444, 443, 439, 435, 433, 431, 430, 429, 428, 424, 423, 421, 418, 416, 413, 408, 407, 404, 403, 402, 398, 395, 392]]', 'study_system_attributes_study_system_attribute_id': 47167}]
(Background on this error at: https://sqlalche.me/e/20/e3q8)Steps to reproduce
- Just run my optimization rutine on 20 different terminals locally, and after some trials (not spsecifics steps to reproduce), some of the executions stopped with the error message provided.
Additional context
I have tried to use the GRPC proxy, but the problem happens similarly, it is even more common under it. It seems that there is an update to the DB at the same record on multiple threads and one of them fails. Tried different configurations for retry errors and heartbeat/grace periods but the issue remains.
I am using mariadb locally, version: mariadb from 11.8.1-MariaDB, client 15.2 for debian-linux-gnu (x86_64) using EditLine wrapper
I have also tried Optuna 4.3.0 but it raises another error with NSGAII which will raise once this is solved to not have crossed issues.