feat(db): SQLite 全量迁移至 MySQL(config.yaml 配置化 + 迁移脚本 + 一致性校验)

- config.yaml database.mysql:host/port/db/user/charset 明文可提交;密码经 password_env
  引用 .env 的 MYSQL_PASSWORD(AGENT.md §33 密钥不进 git)
- config.py _build_mysql_url:URL 优先级 DATABASE_URL env > database.mysql 段 > sqlite 兜底
- pyproject 引入 pymysql>=1.1
- tests/conftest.py 强制每进程 /tmp SQLite(测试绝不触 MySQL 开发库);test_config 覆盖
  mysql 组装/密码可选/sqlite 兜底分支
- scripts/migrate_sqlite_to_mysql.py:sqlite→mysql 一次性迁移工具(keyset 分页 + chunk
  多值 INSERT + 攒批 commit + 幂等续传 + 低配 MySQL 节流 --throttle-sec + --verify-only
  行数与抽样一致性校验);已用于 data/quant.db 约 1605 万行迁移并经校验一致
This commit is contained in:
Simon
2026-09-08 23:58:48 +08:00
parent a3fabf9ae9
commit 6c2f198261
8 changed files with 409 additions and 20 deletions
+25 -1
View File
@@ -10,6 +10,7 @@ import os
from dataclasses import dataclass
from functools import lru_cache
from pathlib import Path
from urllib.parse import quote_plus
import yaml
@@ -99,6 +100,26 @@ def _normalize_sqlite_url(url: str) -> str:
return f"{_SQLITE_PREFIX}{(PROJECT_ROOT / rest).resolve()}"
def _build_mysql_url(mysql: dict | None) -> str | None:
"""由 config.yaml database.mysql 段组装 mysql+pymysql URL。
规则:仅当 enabled 且 host/user/db 齐全时返回 URL;密码经 password_env
指定的环境变量读取(AGENT.md §33 密钥只放 .env),未设置则按空密码处理。
"""
if not mysql or not mysql.get("enabled"):
return None
host = mysql.get("host")
user = mysql.get("user")
db = mysql.get("db")
if not (host and user and db):
return None
port = mysql.get("port") or 3306
charset = mysql.get("charset") or "utf8mb4"
password = os.environ.get(mysql.get("password_env") or "MYSQL_PASSWORD", "")
auth = f"{quote_plus(user)}:{quote_plus(password)}" if password else quote_plus(user)
return f"mysql+pymysql://{auth}@{host}:{port}/{db}?charset={charset}"
@lru_cache
def get_settings() -> Settings:
"""加载并缓存 Settings。config / env 路径可通过参数覆盖以便测试。"""
@@ -115,7 +136,10 @@ def get_settings() -> Settings:
llm_model_env = agent_llm.get("model_env") or "LLM_MODEL"
default_url = "sqlite:///./data/quant.db"
database_url = os.environ.get(url_env) or default_url
# URL 优先级:环境变量(DATABASE_URL) > config.yaml database.mysql 段 > SQLite 兜底
database_url = os.environ.get(url_env) or _build_mysql_url(
_deep(cfg, "database.mysql") or {}
) or default_url
migrations_rel = _deep(cfg, "database.migrations_dir") or (
"app/infrastructure/persistence/migrations"