574 lines
26 KiB
Python
574 lines
26 KiB
Python
"""
|
||
跨库数据迁移脚本（MySQL → PostgreSQL / SQL Server / SQLite / Oracle 等）
|
||
=============================================================
|
||
|
||
把「源库」的数据整体搬到「目标库」：
|
||
- 目标库结构用模型 metadata 生成（db.metadata.create_all，方言可移植，
|
||
模型即 head，与 src/models_registry.py / setup_db.py 同源）；
|
||
- 数据按拓扑序逐表 SELECT ... ORDER BY id，显式带 id 批量插入，保住
|
||
无外键裸列引用关系（user.dept_id、role_menu.role_id/menu_id 等）；
|
||
- 最后按方言重置自增序列（PG setval / MSSQL IDENTITY_INSERT + DBCC
|
||
CHECKIDENT / Oracle 补 sequence + BEFORE INSERT 触发器；
|
||
MySQL / SQLite 显式插 id 后自动跟随，无需处理）。
|
||
|
||
与 init_db.py / setup_db.py 的分工：
|
||
- init_db.py / setup_db.py 单库初始化（当前 .env 目标库 create_all + 种子数据）
|
||
- migrate_db.py 双库迁移（源库 → 目标库，CLI 显式参数，不依赖 .env 连接串）
|
||
|
||
用法：
|
||
python scripts/migrate_db.py \
|
||
--src-driver=mysql --src-host=127.0.0.1 --src-port=3306 \
|
||
--src-db=djangoadmin.flask.naivevue --src-user=root --src-pass=root \
|
||
--dst-driver=postgresql --dst-host=127.0.0.1 --dst-port=5432 \
|
||
--dst-db=djangoadmin.flask.naivevue --dst-user=postgres --dst-pass=123456 \
|
||
--drop-target-first
|
||
|
||
先干跑核对计划：加 --dry-run（只连源、逐表数行数、打印计划，不碰目标库）。
|
||
"""
|
||
|
||
import argparse
|
||
import os
|
||
import sys
|
||
import time
|
||
from contextlib import contextmanager
|
||
from pathlib import Path
|
||
from urllib.parse import quote_plus
|
||
|
||
# 将项目根加入 sys.path（脚本以 `python scripts/xxx.py` 运行时 cwd 是 scripts/）
|
||
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
||
if ROOT not in sys.path:
|
||
sys.path.insert(0, ROOT)
|
||
|
||
# Windows 控制台默认 GBK(cp936)：✓ 等字符无法编码会崩溃、中文按终端编码显示成乱码。
|
||
# 统一把 stdout/stderr 切到 UTF-8，Git Bash / 重定向均正常；无 reconfigure 的环境跳过。
|
||
if sys.stdout and hasattr(sys.stdout, 'reconfigure'):
|
||
sys.stdout.reconfigure(encoding='utf-8')
|
||
sys.stderr.reconfigure(encoding='utf-8')
|
||
|
||
from sqlalchemy import create_engine, inspect, select, text # noqa: E402
|
||
from sqlalchemy.pool import NullPool # noqa: E402
|
||
|
||
from config import DB_PREFIX # noqa: E402
|
||
from config.database import _normalize_driver, _sqlite_url # noqa: E402
|
||
import src.core.database as _dbm # noqa: E402
|
||
import src.models_registry as _models_registry # noqa: E402,F401 # 模型注册中心：填充 db.metadata
|
||
db = _dbm.db
|
||
|
||
# 各驱动默认端口（sqlite 无端口概念）
|
||
_DEFAULT_PORTS = {"mysql": 3306, "postgresql": 5432, "mssql": 1433, "oracle": 1521, "sqlite": None}
|
||
|
||
# 迁移顺序：基础/被引用表在前，依赖表在后。项目无外键，表间关联是裸 Integer 列
|
||
# 由应用层维护，显式带 id 插入即保住引用关系；保序只是锦上添花（父先子后）。
|
||
_BASE_TABLES = [
|
||
"dept", "menu", "role", "level", "position", "city", "category",
|
||
"dict", "config", "file_template", "param", "job", "link",
|
||
]
|
||
_DEP_TABLES = [
|
||
"role_menu", "dict_item", "config_item", "job_log",
|
||
"user", "user_role", "article", "notice", "operation_log", "login_log",
|
||
]
|
||
|
||
|
||
def build_parser() -> argparse.ArgumentParser:
|
||
"""CLI 参数：源/目标两端连接显式给出，不依赖 .env。"""
|
||
p = argparse.ArgumentParser(description="跨库数据迁移：源库 → 目标库")
|
||
|
||
def add_side(side: str, default_user: str):
|
||
p.add_argument(f"--{side}-driver", choices=("mysql", "postgresql", "mssql", "oracle", "sqlite"), default="mysql",
|
||
help="驱动（oracle 时 --{side}-db 即 service_name；sqlite 时即数据库文件路径）")
|
||
p.add_argument(f"--{side}-host", default="127.0.0.1")
|
||
p.add_argument(f"--{side}-port", type=int, default=None,
|
||
help="端口，缺省按驱动默认（mysql 3306 / pg 5432 / mssql 1433 / oracle 1521）")
|
||
p.add_argument(f"--{side}-db", required=True, help="数据库名（sqlite 时为文件路径）")
|
||
p.add_argument(f"--{side}-user", default=default_user)
|
||
p.add_argument(f"--{side}-pass", default="",
|
||
help=f"口令，留空回落环境变量 {side.upper()}_DB_PASS（避免进 shell 历史）")
|
||
|
||
add_side("src", "root")
|
||
add_side("dst", "postgres")
|
||
|
||
p.add_argument("--drop-target-first", action="store_true", help="先删目标库再重建（默认关）")
|
||
p.add_argument("--dry-run", action="store_true", help="只连源、逐表数行数、打印计划，不碰目标库")
|
||
p.add_argument("--chunk-size", type=int, default=1000, help="每批插入行数（默认 1000）")
|
||
p.add_argument("--only-tables", default="", help="逗号分隔的表名子集（全名或去掉前缀的短名），续传用")
|
||
p.add_argument("--skip-existing", action="store_true", help="目标已等于源行数的表跳过")
|
||
p.add_argument("--skip-verify", action="store_true", help="跳过迁移后的逐表行数校验")
|
||
p.add_argument("--single-transaction", action="store_true", help="全部表包进一个大事务（默认每表一个事务）")
|
||
return p
|
||
|
||
|
||
# ============================================================
|
||
# URL / 引擎
|
||
# ============================================================
|
||
|
||
def make_url(driver: str, host: str, port: int, database: str, username: str, password: str) -> str:
|
||
"""带显式参数的连接串构建器，URL 形态与 config.database._build_database_url 逐字等价
|
||
（后者读 .env 全局，脚本端需要显式参数版）。"""
|
||
drv = _normalize_driver(driver)
|
||
if drv == 'sqlite':
|
||
return _sqlite_url(database) # 此时 database 即文件路径，忽略 host/port/user/pass
|
||
base = quote_plus(username) + ':' + quote_plus(password) + '@' + host + ':' + str(port)
|
||
if drv == 'oracle':
|
||
# Oracle 的"库"即连接的服务名（service_name），schema 是连接用户
|
||
return 'oracle+oracledb://' + base + '/?service_name=' + quote_plus(database)
|
||
base = base + '/' + database
|
||
if drv == 'mysql':
|
||
return 'mysql+pymysql://' + base + '?charset=utf8mb4'
|
||
if drv == 'postgresql':
|
||
return 'postgresql+psycopg2://' + base
|
||
return 'mssql+pymssql://' + base + '?charset=utf8'
|
||
|
||
|
||
def make_engine(driver: str, host: str, port: int, database: str, username: str, password: str):
|
||
"""迁移专用引擎：NullPool，避免连接池把目标/源的会话状态串起来。"""
|
||
return create_engine(make_url(driver, host, port, database, username, password), poolclass=NullPool)
|
||
|
||
|
||
def _resolve_port(driver: str, port: int | None) -> int | None:
|
||
return port if port is not None else _DEFAULT_PORTS[driver]
|
||
|
||
|
||
def _fmt_duration(sec: float) -> str:
|
||
"""耗时格式化：<1s 显示毫秒，<60s 显示秒，否则分+秒。"""
|
||
if sec < 1:
|
||
return f"{sec * 1000:.0f}ms"
|
||
if sec < 60:
|
||
return f"{sec:.1f}s"
|
||
return f"{int(sec // 60)}m{int(sec % 60)}s"
|
||
|
||
|
||
def _busy(msg: str) -> None:
|
||
"""缓冲提示：耗时操作开始前打印一行，让用户感知进程仍在工作（而非卡死）。"""
|
||
print(f" · {msg}...", flush=True)
|
||
|
||
|
||
@contextmanager
|
||
def step(title: str):
|
||
"""耗时步骤缓冲包装：进入打印"正在...（请勿中断）"，退出打印"✓ 完成（耗时）"。
|
||
|
||
用于删库/建库/建表/大表迁移等可能耗时较长的操作，避免用户误以为进程卡死。
|
||
通过 `with step(...) as h: h['detail'] = ...` 可给完成行追加明细（如行数/失败数）。
|
||
"""
|
||
t0 = time.perf_counter()
|
||
print(f"[·] {title}...（可能耗时较长，请勿中断）", flush=True)
|
||
holder: dict = {}
|
||
try:
|
||
yield holder
|
||
finally:
|
||
detail = f"，{holder['detail']}" if holder.get('detail') else ""
|
||
print(f"[✓] {title} 完成（耗时 {_fmt_duration(time.perf_counter() - t0)}）{detail}", flush=True)
|
||
|
||
|
||
# ============================================================
|
||
# 建目标库（镜像 init_db.py 的分支）
|
||
# ============================================================
|
||
|
||
def create_database_if_missing(driver: str, host: str, port: int, database: str,
|
||
username: str, password: str, drop_first: bool = False,
|
||
tables: tuple = ()):
|
||
"""按方言连维护库建目标库（不存在才建）。sqlite 无建库概念，只确保父目录并可选删旧文件。
|
||
|
||
- mysql: 连无库连接 CREATE DATABASE ... utf8mb4
|
||
- postgresql: 连 postgres 维护库（autocommit），库名含点号必须双引号
|
||
- mssql: 连 master 维护库，标识符用方括号
|
||
- oracle: 服务由 DBA 管理，schema 即连接用户，无需/无法"建库"；仅校验连通性，
|
||
drop_first 时尽力 DROP TABLE PURGE 清掉 tables 里已存在的项目表
|
||
"""
|
||
if driver == 'sqlite':
|
||
file = Path(database).expanduser()
|
||
if file.parent.as_posix() != '.':
|
||
file.parent.mkdir(parents=True, exist_ok=True)
|
||
if drop_first and file.exists():
|
||
_busy(f"正在删除旧目标文件 {file}")
|
||
file.unlink()
|
||
print(f"removed existing sqlite file {file}")
|
||
return
|
||
|
||
if driver == 'mysql':
|
||
import pymysql
|
||
_busy("正在连接 MySQL（维护实例）")
|
||
conn = pymysql.connect(host=host, port=int(port), user=username, password=password, autocommit=True)
|
||
cur = conn.cursor()
|
||
if drop_first:
|
||
_busy(f"正在删除旧目标库 `{database}`")
|
||
cur.execute(f"DROP DATABASE IF EXISTS `{database}`")
|
||
_busy(f"正在创建目标库 `{database}`（utf8mb4）")
|
||
cur.execute(
|
||
f"CREATE DATABASE IF NOT EXISTS `{database}` "
|
||
f"DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci"
|
||
)
|
||
cur.close()
|
||
conn.close()
|
||
return
|
||
|
||
if driver == 'postgresql':
|
||
import psycopg2
|
||
_busy("正在连接 PostgreSQL（postgres 维护库）")
|
||
# 连 postgres 维护库创建目标库；DB_DATABASE 含点号，标识符必须双引号
|
||
conn = psycopg2.connect(
|
||
host=host, port=int(port), user=username, password=password, dbname="postgres",
|
||
)
|
||
conn.autocommit = True
|
||
with conn.cursor() as cur:
|
||
cur.execute("SELECT 1 FROM pg_database WHERE datname = %s", (database,))
|
||
exists = cur.fetchone() is not None
|
||
if drop_first and exists:
|
||
_busy(f'正在删除旧目标库 "{database}"（DROP ... WITH FORCE）')
|
||
cur.execute(f'DROP DATABASE "{database}" WITH (FORCE)')
|
||
exists = False
|
||
if not exists:
|
||
_busy(f'正在创建目标库 "{database}"（UTF8）')
|
||
# 强制 UTF8 编码（TEMPLATE template0 避免继承模板库非 UTF8 编码导致 CJK 乱码）
|
||
cur.execute(f'CREATE DATABASE "{database}" ENCODING \'UTF8\' TEMPLATE template0')
|
||
conn.close()
|
||
return
|
||
|
||
if driver == 'oracle':
|
||
# 服务由 DBA 管理，schema 就是连接用户；连上即视为"目标库就绪"
|
||
import oracledb
|
||
_busy(f"正在连接 Oracle 服务 {database}")
|
||
conn = oracledb.connect(user=username, password=password, dsn=f"{host}:{port}/{database}")
|
||
cur = conn.cursor()
|
||
cur.execute("SELECT 1 FROM dual")
|
||
cur.fetchone()
|
||
if drop_first:
|
||
_busy(f"正在清理 {len(tables)} 张旧项目表")
|
||
for t in tables:
|
||
try:
|
||
cur.execute(f"DROP TABLE {t} PURGE")
|
||
except Exception:
|
||
pass # 表不存在则跳过
|
||
conn.close()
|
||
print(f"[OK] Oracle 服务 {database} 连接成功（schema={username}）")
|
||
if drop_first:
|
||
print(f" drop-first：已尽力清理 {len(tables)} 张项目表")
|
||
return
|
||
|
||
# mssql：连 master 维护库创建目标库
|
||
import pymssql
|
||
_busy("正在连接 SQL Server（master 维护库）")
|
||
conn = pymssql.connect(
|
||
server=host, user=username, password=password,
|
||
port=int(port), database="master", autocommit=True,
|
||
)
|
||
cur = conn.cursor()
|
||
if drop_first:
|
||
_busy(f"正在删除旧目标库 [{database}]（SINGLE_USER + DROP）")
|
||
cur.execute(
|
||
f"IF DB_ID(N'{database}') IS NOT NULL "
|
||
f"ALTER DATABASE [{database}] SET SINGLE_USER WITH ROLLBACK IMMEDIATE"
|
||
)
|
||
cur.execute(f"IF DB_ID(N'{database}') IS NOT NULL DROP DATABASE [{database}]")
|
||
_busy(f"正在创建目标库 [{database}]")
|
||
cur.execute(f"IF DB_ID(N'{database}') IS NULL CREATE DATABASE [{database}]")
|
||
cur.close()
|
||
conn.close()
|
||
|
||
|
||
# ============================================================
|
||
# 迁移顺序
|
||
# ============================================================
|
||
|
||
def ordered_tables():
|
||
"""按拓扑序返回要迁移的 Table 对象（base 先、dep 后；metadata 中未列出的追加末尾）。"""
|
||
out, seen = [], set()
|
||
for name in _BASE_TABLES + _DEP_TABLES:
|
||
tbl = db.metadata.tables.get(DB_PREFIX + name)
|
||
if tbl is not None:
|
||
out.append(tbl)
|
||
seen.add(tbl.name)
|
||
for tbl in db.metadata.tables.values():
|
||
if tbl.name not in seen:
|
||
out.append(tbl)
|
||
return out
|
||
|
||
|
||
# ============================================================
|
||
# 泵数据
|
||
# ============================================================
|
||
|
||
def sanitize(payload: list[dict], table) -> list[dict]:
|
||
"""把驱动读回的零日期/非法日期字符串统一置 None（防御性兜底）。
|
||
|
||
PyMySQL 默认已把 MySQL 的 '0000-00-00 00:00:00' 转成 None；此处兜底处理
|
||
个别驱动返回原始字符串的情况，避免 PostgreSQL/MSSQL 拒绝非法日期。
|
||
本项目时间列全部 nullable，置 None 安全。
|
||
"""
|
||
for row in payload:
|
||
for col in table.columns:
|
||
v = row.get(col.name)
|
||
if isinstance(v, str) and v.startswith('0000-00-00'):
|
||
row[col.name] = None
|
||
return payload
|
||
|
||
|
||
def pump_table(src, dst, table, chunk_size: int, driver: str) -> int:
|
||
"""逐 chunk 批量插入一张表，显式带 id 以保住裸列引用关系。
|
||
|
||
src/dst 为已就绪的 Connection；dst 处于调用方开启的事务内
|
||
（默认每表一个事务；--single-transaction 时全库一个大事务）。
|
||
mssql 必须 SET IDENTITY_INSERT ON 才能显式插自增 id。
|
||
"""
|
||
pk = next(iter(table.primary_key.columns))
|
||
sel = select(*table.columns).order_by(pk) # ORDER BY id：自引用父先子后
|
||
use_identity = (driver == 'mssql')
|
||
total = 0
|
||
t0 = time.perf_counter()
|
||
if use_identity:
|
||
dst.execute(text(f"SET IDENTITY_INSERT [{table.name}] ON"))
|
||
try:
|
||
result = src.execute(sel)
|
||
while True:
|
||
rows = result.fetchmany(chunk_size)
|
||
if not rows:
|
||
break
|
||
payload = sanitize([dict(r._mapping) for r in rows], table)
|
||
dst.execute(table.insert(), payload)
|
||
total += len(payload)
|
||
print(f" [{table.name}] +{len(payload)}（累计 {total}）")
|
||
finally:
|
||
if use_identity:
|
||
dst.execute(text(f"SET IDENTITY_INSERT [{table.name}] OFF"))
|
||
print(f" [✓ {table.name}] 共 {total} 行（耗时 {_fmt_duration(time.perf_counter() - t0)}）")
|
||
return total
|
||
|
||
|
||
# ============================================================
|
||
# 自增序列重置
|
||
# ============================================================
|
||
|
||
def pg_reset_sql(table_name: str) -> str:
|
||
"""PostgreSQL：把 SERIAL/IDENTITY 序列推进到 MAX(id)，否则显式插 id 后新插入会主键冲突。
|
||
|
||
- 非空表：setval(seq, MAX(id))，is_called=true → 下一 id = MAX(id)+1
|
||
- 空表： setval(seq, 1, false)，is_called=false → 下一 id = 1
|
||
（不能 setval(seq, 0)：PG 序列最小值是 1，会抛 NumericValueOutOfRange）
|
||
"""
|
||
return (
|
||
f"SELECT setval(pg_get_serial_sequence('{table_name}', 'id'), "
|
||
f"COALESCE((SELECT MAX(id) FROM \"{table_name}\"), 1), "
|
||
f"(SELECT MAX(id) FROM \"{table_name}\") IS NOT NULL)"
|
||
)
|
||
|
||
|
||
def mssql_reset_sql(table_name: str, max_id) -> str:
|
||
"""SQL Server：RESEED 必须是字面量，先查 MAX(id) 再拼。空表 RESEED 0 → 下一 id=1。"""
|
||
return f"DBCC CHECKIDENT ('{table_name}', RESEED, {max_id or 0})"
|
||
|
||
|
||
def oracle_sequence_sql(table_name: str, start_with: int) -> str:
|
||
"""Oracle：为无自增机制的 INTEGER 主键建 sequence，START WITH 决定下一 id。"""
|
||
return f"CREATE SEQUENCE {table_name}_id_seq START WITH {start_with} INCREMENT BY 1 NOCACHE"
|
||
|
||
|
||
def oracle_trigger_sql(table_name: str) -> str:
|
||
"""Oracle：BEFORE INSERT 触发器，id 为 NULL 时从 sequence 取 NEXTVAL。"""
|
||
seq = f"{table_name}_id_seq"
|
||
trg = f"{table_name}_id_trg"
|
||
return (
|
||
f"CREATE OR REPLACE TRIGGER {trg}\n"
|
||
f"BEFORE INSERT ON {table_name} FOR EACH ROW\n"
|
||
f"WHEN (NEW.id IS NULL)\n"
|
||
f"BEGIN SELECT {seq}.NEXTVAL INTO :NEW.id FROM dual; END;"
|
||
)
|
||
|
||
|
||
def ensure_oracle_autoincrement(conn, table_name: str):
|
||
"""Oracle 迁移后补自增机制：重建 sequence（START WITH MAX(id)+1）+ BEFORE INSERT 触发器。
|
||
|
||
Oracle 的 create_all 对 Integer 自增主键只生成 INTEGER NOT NULL（无 identity/sequence），
|
||
不补的话应用 save() 拿不到新 id。空表 MAX(id) 为 NULL → COALESCE 0 → START WITH 1。
|
||
Oracle 无 DROP SEQUENCE IF EXISTS，先查 user_sequences 再删。
|
||
"""
|
||
max_id = conn.execute(text(f"SELECT COALESCE(MAX(id), 0) FROM {table_name}")).scalar()
|
||
seq = f"{table_name}_id_seq"
|
||
exists = conn.execute(text(
|
||
"SELECT COUNT(*) FROM user_sequences WHERE sequence_name = UPPER(:n)"
|
||
), {"n": seq}).scalar()
|
||
if exists:
|
||
conn.execute(text(f"DROP SEQUENCE {seq}"))
|
||
conn.execute(text(oracle_sequence_sql(table_name, max_id + 1)))
|
||
conn.execute(text(oracle_trigger_sql(table_name)))
|
||
|
||
|
||
def reset_sequence(driver: str, conn, table):
|
||
"""迁移后重置自增序列。MySQL / SQLite 显式插 id 后 auto_increment/rowid 自动
|
||
跟随 MAX(id)，无需处理；PostgreSQL / SQL Server / Oracle 需要。"""
|
||
if driver == 'postgresql':
|
||
conn.execute(text(pg_reset_sql(table.name)))
|
||
elif driver == 'mssql':
|
||
max_id = conn.execute(text(f"SELECT MAX(id) FROM [{table.name}]")).scalar()
|
||
conn.execute(text(mssql_reset_sql(table.name, max_id)))
|
||
elif driver == 'oracle':
|
||
ensure_oracle_autoincrement(conn, table.name)
|
||
|
||
|
||
# ============================================================
|
||
# 校验
|
||
# ============================================================
|
||
|
||
def verify_counts(src, dst, tables) -> list[tuple[str, int, int]]:
|
||
"""逐表 count 对比，返回不一致的表列表（(表名, 源行数, 目标行数)）。"""
|
||
failed = []
|
||
for t in tables:
|
||
s = src.execute(text(f"SELECT COUNT(*) FROM {t.name}")).scalar()
|
||
d = dst.execute(text(f"SELECT COUNT(*) FROM {t.name}")).scalar()
|
||
mark = "PASS" if s == d else "FAIL"
|
||
print(f" [{t.name}] source={s} target={d} {mark}")
|
||
if s != d:
|
||
failed.append((t.name, s, d))
|
||
return failed
|
||
|
||
|
||
def _existing_equal(src, dst, table) -> bool:
|
||
s = src.execute(text(f"SELECT COUNT(*) FROM {table.name}")).scalar()
|
||
d = dst.execute(text(f"SELECT COUNT(*) FROM {table.name}")).scalar()
|
||
return s == d
|
||
|
||
|
||
# ============================================================
|
||
# 主流程
|
||
# ============================================================
|
||
|
||
def _dry_run(src_engine, tables, dst_driver: str, dst_db: str, t_start: float) -> int:
|
||
print("=" * 62)
|
||
print(f"[DRY-RUN 开始] 计划迁移 {len(tables)} 张表 → {dst_driver}:{dst_db}")
|
||
print("=" * 62)
|
||
with src_engine.connect() as src:
|
||
for t in tables:
|
||
n = src.execute(text(f"SELECT COUNT(*) FROM {t.name}")).scalar()
|
||
print(f" - {t.name}: {n} 行")
|
||
print(f"[DRY-RUN 完成] {len(tables)} 张表核对完毕（耗时 {_fmt_duration(time.perf_counter() - t_start)}），未触碰目标库")
|
||
return 0
|
||
|
||
|
||
def run(args) -> int:
|
||
"""执行一次迁移。返回进程退出码（0 成功 / 1 校验失败）。"""
|
||
src_driver = _normalize_driver(args.src_driver)
|
||
dst_driver = _normalize_driver(args.dst_driver)
|
||
|
||
src_engine = make_engine(
|
||
src_driver, args.src_host, _resolve_port(src_driver, args.src_port),
|
||
args.src_db, args.src_user, args.src_pass or os.environ.get("SRC_DB_PASS", ""),
|
||
)
|
||
dst_engine = make_engine(
|
||
dst_driver, args.dst_host, _resolve_port(dst_driver, args.dst_port),
|
||
args.dst_db, args.dst_user, args.dst_pass or os.environ.get("DST_DB_PASS", ""),
|
||
)
|
||
|
||
tables = ordered_tables()
|
||
if args.only_tables:
|
||
wanted = {x.strip() for x in args.only_tables.split(",") if x.strip()}
|
||
tables = [t for t in tables if t.name in wanted or t.name[len(DB_PREFIX):] in wanted]
|
||
|
||
try:
|
||
# 源端缺表容错（老库可能缺整表）
|
||
src_insp = inspect(src_engine)
|
||
present, missing = [], []
|
||
for t in tables:
|
||
if src_insp.has_table(t.name):
|
||
present.append(t)
|
||
else:
|
||
missing.append(t)
|
||
for t in missing:
|
||
print(f"[SKIP] 源库不存在表 {t.name}，跳过")
|
||
|
||
t_start = time.perf_counter()
|
||
if args.dry_run:
|
||
return _dry_run(src_engine, present, dst_driver, args.dst_db, t_start)
|
||
|
||
print("=" * 62)
|
||
print(f"[迁移开始] 源: {src_driver}:{args.src_db} → 目标: {dst_driver}:{args.dst_db}")
|
||
note = "" if not missing else f"，源端缺失 {len(missing)} 张已跳过"
|
||
print(f"[迁移开始] 待迁移 {len(present)} 张表{note}")
|
||
print("=" * 62)
|
||
|
||
# 1) 建目标库（删库/建库可能耗时较长，step() 会给出缓冲提示）
|
||
with step("[1/5] 建目标库"):
|
||
create_database_if_missing(
|
||
dst_driver, args.dst_host, _resolve_port(dst_driver, args.dst_port),
|
||
args.dst_db, args.dst_user, args.dst_pass or os.environ.get("DST_DB_PASS", ""),
|
||
drop_first=args.drop_target_first,
|
||
tables=tuple(t.name for t in present), # Oracle drop_first 时清表用
|
||
)
|
||
|
||
# 2) 建结构（模型即 head，方言可移植）
|
||
with step(f"[2/5] 建表结构（{len(present)} 张表 create_all）") as h:
|
||
db.metadata.create_all(dst_engine)
|
||
h['detail'] = f"{len(present)} 张表"
|
||
|
||
# 3) 泵数据
|
||
total_rows = 0
|
||
with step("[3/5] 逐表迁移数据") as h:
|
||
if args.single_transaction:
|
||
with dst_engine.begin() as dst, src_engine.connect() as src:
|
||
for idx, t in enumerate(present, 1):
|
||
if args.skip_existing and _existing_equal(src, dst, t):
|
||
print(f" [skip] {t.name}（源=目标）")
|
||
continue
|
||
_busy(f"[{idx}/{len(present)}] 正在迁移 {t.name}")
|
||
total_rows += pump_table(src, dst, t, args.chunk_size, dst_driver)
|
||
else:
|
||
with src_engine.connect() as src:
|
||
for idx, t in enumerate(present, 1):
|
||
with dst_engine.begin() as dst:
|
||
if args.skip_existing and _existing_equal(src, dst, t):
|
||
print(f" [skip] {t.name}（源=目标）")
|
||
continue
|
||
_busy(f"[{idx}/{len(present)}] 正在迁移 {t.name}")
|
||
total_rows += pump_table(src, dst, t, args.chunk_size, dst_driver)
|
||
h['detail'] = f"共 {total_rows} 行"
|
||
|
||
# 4) 重置自增序列（PG/MSSQL/Oracle；MySQL/SQLite 自动跟随）
|
||
# 单表失败不中断其余表，避免一张空表序列异常连累后续全部序列
|
||
reset_errors = []
|
||
if dst_driver in ("postgresql", "mssql", "oracle"):
|
||
with step("[4/5] 重置自增序列") as h:
|
||
for t in present:
|
||
try:
|
||
# 每表独立事务：Oracle 的 CREATE/DROP SEQUENCE / TRIGGER 是 DDL 会隐式提交，
|
||
# 不能与其它表共享一个大事务
|
||
with dst_engine.begin() as dst:
|
||
reset_sequence(dst_driver, dst, t)
|
||
except Exception as e:
|
||
reset_errors.append((t.name, str(e)))
|
||
print(f" [WARN] 序列重置失败 {t.name}: {e}")
|
||
if reset_errors:
|
||
h['detail'] = f"{len(reset_errors)} 张失败"
|
||
else:
|
||
print("[4/5 重置序列] MySQL/SQLite 自动跟随，无需处理")
|
||
|
||
# 5) 校验
|
||
if not args.skip_verify:
|
||
with step("[5/5] 逐表校验行数") as h:
|
||
with src_engine.connect() as src, dst_engine.connect() as dst:
|
||
failed = verify_counts(src, dst, present)
|
||
h['detail'] = f"{len(failed)} 张表不一致" if failed else f"{len(present)} 张表全部 PASS"
|
||
if failed:
|
||
for name, s, d in failed:
|
||
print(f" - {name}: source={s} target={d}")
|
||
return 1
|
||
else:
|
||
print("[5/5 校验] 已跳过（--skip-verify）")
|
||
|
||
# 汇总
|
||
print("=" * 62)
|
||
print(f"[迁移完成] 成功：{total_rows} 行 / {len(present)} 张表 / 总耗时 {_fmt_duration(time.perf_counter() - t_start)}")
|
||
print("=" * 62)
|
||
return 0
|
||
finally:
|
||
src_engine.dispose()
|
||
dst_engine.dispose()
|
||
|
||
|
||
def main():
|
||
args = build_parser().parse_args()
|
||
try:
|
||
sys.exit(run(args))
|
||
except KeyboardInterrupt:
|
||
sys.exit(130)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|