fix(migrations): enforce schema ownership safety

This commit is contained in:
caoxiaozhu
2026-07-14 09:23:34 +08:00
parent 1347366b95
commit 11275e4ba6
18 changed files with 755 additions and 57 deletions

View File

@@ -0,0 +1,364 @@
from __future__ import annotations
import os
import re
from collections.abc import Iterator
from pathlib import Path
import pytest
from alembic.config import Config
from sqlalchemy import create_engine, inspect, text
from sqlalchemy.engine import Engine, make_url
from sqlalchemy.pool import NullPool
from alembic import command
from app.core.config import get_settings
from app.db.migration_preflight import MigrationPreflightError, validate_migration_state
from app.db.schema_ownership import MIGRATION_OWNED_TABLES
MIGRATION_TEST_DATABASE_URL = os.getenv("MIGRATION_TEST_DATABASE_URL", "").strip()
LEGACY_PROBE_TABLE = "legacy_migration_probe_records"
HEAD_REVISION = "20260713_0002"
SERVER_DIR = Path(__file__).resolve().parents[1]
ALEMBIC_INI_PATH = SERVER_DIR / "alembic.ini"
def _normalize_probe_component(value: str) -> str:
return re.sub(r"[^a-z0-9]+", "-", value.lower()).strip("-")
def _is_disposable_probe_host(value: str) -> bool:
markers = ("migration-probe", "disposable-probe")
return value in markers or any(
value.startswith(f"{marker}-") or value.startswith(f"x-financial-{marker}-")
for marker in markers
)
def _is_disposable_probe_database(value: str) -> bool:
markers = ("migration-probe", "disposable-probe")
return value in markers or any(value.startswith(f"{marker}-") for marker in markers)
def _require_disposable_probe_url(raw_url: str) -> str:
try:
parsed = make_url(raw_url)
except Exception as exc: # pragma: no cover - SQLAlchemy 提供具体解析异常
raise RuntimeError("MIGRATION_TEST_DATABASE_URL 不是有效的数据库 URL") from exc
if parsed.get_backend_name() != "postgresql":
raise RuntimeError("迁移测试只允许连接 PostgreSQL 一次性数据库")
host = _normalize_probe_component(parsed.host or "")
database = _normalize_probe_component(parsed.database or "")
if not _is_disposable_probe_host(host):
raise RuntimeError(
"迁移测试数据库主机名必须使用 migration-probe 或 disposable-probe 前缀"
)
if not _is_disposable_probe_database(database):
raise RuntimeError(
"迁移测试数据库名必须使用 migration-probe 或 disposable-probe 前缀"
)
return raw_url
@pytest.fixture(scope="module")
def migration_database_url() -> Iterator[str]:
if not MIGRATION_TEST_DATABASE_URL:
pytest.skip("仅在显式配置 MIGRATION_TEST_DATABASE_URL 时运行一次性 PostgreSQL 迁移测试")
database_url = _require_disposable_probe_url(MIGRATION_TEST_DATABASE_URL)
previous_database_url = os.environ.get("DATABASE_URL")
os.environ["DATABASE_URL"] = database_url
get_settings.cache_clear()
try:
resolved_url = get_settings().resolved_database_url
if make_url(resolved_url) != make_url(database_url):
raise RuntimeError("运行时数据库 URL 未解析到 MIGRATION_TEST_DATABASE_URL")
yield database_url
finally:
if previous_database_url is None:
os.environ.pop("DATABASE_URL", None)
else:
os.environ["DATABASE_URL"] = previous_database_url
get_settings.cache_clear()
def _alembic_config(database_url: str) -> Config:
config = Config(str(ALEMBIC_INI_PATH))
config.set_main_option("sqlalchemy.url", database_url.replace("%", "%%"))
return config
def _upgrade_head(database_url: str) -> None:
get_settings.cache_clear()
command.upgrade(_alembic_config(database_url), "head")
def _downgrade_base(database_url: str) -> None:
get_settings.cache_clear()
command.downgrade(_alembic_config(database_url), "base")
def _table_names(engine: Engine) -> set[str]:
return set(inspect(engine).get_table_names(schema="public"))
def _assert_unique_constraint(
engine: Engine,
table_name: str,
constraint_name: str,
expected_columns: tuple[str, ...],
) -> None:
constraints = {
str(item["name"]): tuple(item["column_names"])
for item in inspect(engine).get_unique_constraints(table_name, schema="public")
}
assert constraints.get(constraint_name) == expected_columns
def _assert_indexes(
engine: Engine,
table_name: str,
expected_indexes: dict[str, tuple[str, ...]],
) -> None:
indexes = {
str(item["name"]): tuple(item["column_names"])
for item in inspect(engine).get_indexes(table_name, schema="public")
}
for index_name, expected_columns in expected_indexes.items():
assert indexes.get(index_name) == expected_columns
def _assert_cascade_foreign_key(engine: Engine, table_name: str) -> None:
foreign_keys = inspect(engine).get_foreign_keys(table_name, schema="public")
matching = [
item
for item in foreign_keys
if item["constrained_columns"] == ["expense_case_id"]
and item["referred_table"] == "expense_cases"
and item["referred_columns"] == ["id"]
]
assert len(matching) == 1
assert str(matching[0].get("options", {}).get("ondelete", "")).upper() == "CASCADE"
def _assert_head_schema(engine: Engine) -> None:
names = _table_names(engine)
assert MIGRATION_OWNED_TABLES.issubset(names)
assert "alembic_version" in names
with engine.connect() as connection:
assert connection.scalar(text("SELECT version_num FROM alembic_version")) == HEAD_REVISION
_assert_unique_constraint(
engine,
"expense_cases",
"uq_expense_cases_tenant_case_no",
("tenant_id", "case_no"),
)
_assert_unique_constraint(
engine,
"expense_case_links",
"uq_expense_case_links_resource",
("resource_type", "resource_id"),
)
_assert_unique_constraint(
engine,
"business_events",
"uq_business_event_idempotency",
("tenant_id", "aggregate_type", "aggregate_id", "event_type", "idempotency_key"),
)
_assert_unique_constraint(
engine,
"auth_sessions",
"uq_auth_sessions_token_hash",
("token_hash",),
)
_assert_indexes(
engine,
"expense_cases",
{
"ix_expense_cases_tenant_stage": ("tenant_id", "current_stage"),
"ix_expense_cases_tenant_status": ("tenant_id", "status"),
},
)
_assert_indexes(
engine,
"expense_case_links",
{"ix_expense_case_links_tenant_case": ("tenant_id", "expense_case_id")},
)
_assert_indexes(
engine,
"business_events",
{
"ix_business_events_aggregate": ("aggregate_type", "aggregate_id"),
"ix_business_events_outbox": ("delivery_status", "occurred_at"),
"ix_business_events_tenant_case_time": (
"tenant_id",
"expense_case_id",
"occurred_at",
),
},
)
_assert_indexes(
engine,
"auth_sessions",
{
"ix_auth_sessions_principal_active": (
"principal_type",
"revoked_at",
"expires_at",
),
"ix_auth_sessions_tenant_username": ("tenant_id", "username"),
},
)
_assert_cascade_foreign_key(engine, "expense_case_links")
_assert_cascade_foreign_key(engine, "business_events")
def _assert_runtime_cascade(engine: Engine) -> None:
with engine.begin() as connection:
connection.execute(
text(
"""
INSERT INTO expense_cases (
id, tenant_id, case_no, scene_code, title, current_stage, status
) VALUES (
'migration-probe-case', 'migration-probe', 'CASE-MIGRATION-PROBE',
'reimbursement', '迁移级联验证', 'claiming', 'active'
)
"""
)
)
connection.execute(
text(
"""
INSERT INTO expense_case_links (
id, tenant_id, expense_case_id, resource_type, resource_id, relation_type
) VALUES (
'migration-probe-link', 'migration-probe', 'migration-probe-case',
'expense_claim', 'migration-probe-claim', 'claim'
)
"""
)
)
connection.execute(
text(
"""
INSERT INTO business_events (
id, tenant_id, expense_case_id, aggregate_type, aggregate_id,
event_type, event_version, idempotency_key, correlation_id,
actor_id, actor_type, payload_json, delivery_status, delivery_attempts
) VALUES (
'migration-probe-event', 'migration-probe', 'migration-probe-case',
'expense_claim', 'migration-probe-claim', 'claim_draft_created', 1,
'migration-probe-idempotency', 'migration-probe-correlation',
'migration-probe-user', 'user', '{}', 'pending', 0
)
"""
)
)
connection.execute(
text("DELETE FROM expense_cases WHERE id = 'migration-probe-case'")
)
assert connection.scalar(
text("SELECT COUNT(*) FROM expense_case_links WHERE id = 'migration-probe-link'")
) == 0
assert connection.scalar(
text("SELECT COUNT(*) FROM business_events WHERE id = 'migration-probe-event'")
) == 0
def _create_legacy_sentinel(engine: Engine) -> None:
with engine.begin() as connection:
connection.execute(
text(
f"""
CREATE TABLE {LEGACY_PROBE_TABLE} (
id VARCHAR(64) PRIMARY KEY,
payload VARCHAR(255) NOT NULL
)
"""
)
)
connection.execute(
text(
f"""
INSERT INTO {LEGACY_PROBE_TABLE} (id, payload)
VALUES ('legacy-sentinel', 'must-survive-migration-cycle')
"""
)
)
def _assert_legacy_sentinel(engine: Engine) -> None:
assert LEGACY_PROBE_TABLE in _table_names(engine)
with engine.connect() as connection:
payload = connection.scalar(
text(
f"SELECT payload FROM {LEGACY_PROBE_TABLE} "
"WHERE id = 'legacy-sentinel'"
)
)
assert payload == "must-survive-migration-cycle"
def _assert_base_schema(engine: Engine) -> None:
names = _table_names(engine)
assert not MIGRATION_OWNED_TABLES.intersection(names)
assert "alembic_version" in names
with engine.connect() as connection:
assert connection.scalar(text("SELECT COUNT(*) FROM alembic_version")) == 0
@pytest.mark.parametrize(
"database_url",
[
"sqlite+pysqlite:///:memory:",
"postgresql+psycopg://postgres:postgres@x-financial-local-postgres:5432/x_financial",
"postgresql+psycopg://probe:probe@migration-probe-123:5432/x_financial",
],
)
def test_disposable_database_guard_rejects_unsafe_urls(database_url: str) -> None:
with pytest.raises(RuntimeError):
_require_disposable_probe_url(database_url)
def test_alembic_migration_cycle_on_disposable_postgres(
migration_database_url: str,
) -> None:
engine = create_engine(migration_database_url, poolclass=NullPool)
try:
assert _table_names(engine) == set(), "迁移测试必须从全新空库开始"
assert validate_migration_state(engine).revision is None
_upgrade_head(migration_database_url)
_assert_head_schema(engine)
assert validate_migration_state(engine).revision == HEAD_REVISION
_upgrade_head(migration_database_url)
_assert_head_schema(engine)
_assert_runtime_cascade(engine)
_create_legacy_sentinel(engine)
_downgrade_base(migration_database_url)
_assert_base_schema(engine)
_assert_legacy_sentinel(engine)
assert validate_migration_state(engine).revision is None
with engine.begin() as connection:
connection.execute(text("CREATE TABLE expense_cases (id VARCHAR(36) PRIMARY KEY)"))
with pytest.raises(MigrationPreflightError, match="migration-owned tables exist"):
validate_migration_state(engine)
with engine.begin() as connection:
connection.execute(text("DROP TABLE expense_cases"))
assert validate_migration_state(engine).revision is None
_upgrade_head(migration_database_url)
_assert_head_schema(engine)
_assert_legacy_sentinel(engine)
finally:
engine.dispose()

View File

@@ -0,0 +1,135 @@
from __future__ import annotations
from pathlib import Path
import pytest
from sqlalchemy import Column, Integer, MetaData, Table, create_engine, text
from sqlalchemy.engine import Engine
from app.db.migration_preflight import (
MIGRATION_OWNED_TABLES_BY_REVISION,
MigrationPreflightError,
validate_migration_state,
)
@pytest.fixture
def engine() -> Engine:
database = create_engine("sqlite+pysqlite:///:memory:")
try:
yield database
finally:
database.dispose()
def _create_tables(engine: Engine, table_names: set[str] | frozenset[str]) -> None:
metadata = MetaData()
for table_name in table_names:
Table(table_name, metadata, Column("id", Integer, primary_key=True))
metadata.create_all(engine)
def _create_version_table(engine: Engine, *revisions: str) -> None:
with engine.begin() as connection:
connection.execute(text("CREATE TABLE alembic_version (version_num VARCHAR(32) NOT NULL)"))
for revision in revisions:
connection.execute(
text("INSERT INTO alembic_version (version_num) VALUES (:revision)"),
{"revision": revision},
)
def test_unversioned_database_without_migration_owned_tables_is_safe(engine: Engine) -> None:
state = validate_migration_state(engine)
assert state.revision is None
assert state.owned_tables == frozenset()
@pytest.mark.parametrize(
"owned_table",
sorted(MIGRATION_OWNED_TABLES_BY_REVISION["20260713_0002"]),
)
def test_unversioned_database_with_any_migration_owned_table_is_rejected(
engine: Engine,
owned_table: str,
) -> None:
_create_tables(engine, {owned_table})
with pytest.raises(MigrationPreflightError, match="unversioned database contains"):
validate_migration_state(engine)
@pytest.mark.parametrize(
("revision", "expected_tables"),
list(MIGRATION_OWNED_TABLES_BY_REVISION.items()),
)
def test_known_revision_requires_and_accepts_its_exact_owned_table_set(
engine: Engine,
revision: str,
expected_tables: frozenset[str],
) -> None:
_create_tables(engine, expected_tables)
_create_version_table(engine, revision)
state = validate_migration_state(engine)
assert state.revision == revision
assert state.owned_tables == expected_tables
@pytest.mark.parametrize(
("revision", "actual_tables"),
[
("20260713_0001", frozenset({"expense_cases", "expense_case_links"})),
("20260713_0001", MIGRATION_OWNED_TABLES_BY_REVISION["20260713_0002"]),
(
"20260713_0002",
MIGRATION_OWNED_TABLES_BY_REVISION["20260713_0002"] - {"auth_sessions"},
),
],
)
def test_known_revision_with_missing_or_unexpected_owned_tables_is_rejected(
engine: Engine,
revision: str,
actual_tables: frozenset[str],
) -> None:
_create_tables(engine, actual_tables)
_create_version_table(engine, revision)
with pytest.raises(MigrationPreflightError, match="does not match revision"):
validate_migration_state(engine)
def test_unknown_revision_is_rejected(engine: Engine) -> None:
_create_version_table(engine, "20990101_unknown")
with pytest.raises(MigrationPreflightError, match="unknown Alembic revision"):
validate_migration_state(engine)
def test_multiple_revisions_are_rejected(engine: Engine) -> None:
_create_version_table(engine, "20260713_0001", "20260713_0002")
with pytest.raises(MigrationPreflightError, match="multiple Alembic revisions"):
validate_migration_state(engine)
def test_empty_version_table_is_safe_only_when_owned_tables_are_absent(engine: Engine) -> None:
_create_version_table(engine)
assert validate_migration_state(engine).revision is None
_create_tables(engine, {"expense_cases"})
with pytest.raises(MigrationPreflightError, match="no recorded revision"):
validate_migration_state(engine)
def test_server_start_runs_preflight_before_alembic_upgrade() -> None:
script_path = Path(__file__).resolve().parents[1] / "server_start.sh"
script = script_path.read_text(encoding="utf-8")
preflight = '"$PYTHON_BIN" -m app.db.migration_preflight'
upgrade = '"$PYTHON_BIN" -m alembic -c "$SCRIPT_DIR/alembic.ini" upgrade head'
assert 'PYTHONPATH="$SCRIPT_DIR/src${PYTHONPATH:+:$PYTHONPATH}"' in script
assert script.index(preflight) < script.index(upgrade)

View File

@@ -0,0 +1,26 @@
from __future__ import annotations
from sqlalchemy import create_engine, inspect
from app.db.schema_ownership import MIGRATION_OWNED_TABLES, create_legacy_schema
def test_create_legacy_schema_never_creates_migration_owned_tables() -> None:
engine = create_engine("sqlite+pysqlite:///:memory:")
try:
create_legacy_schema(engine)
table_names = set(inspect(engine).get_table_names())
finally:
engine.dispose()
assert MIGRATION_OWNED_TABLES == frozenset(
{
"auth_sessions",
"business_events",
"expense_case_links",
"expense_cases",
}
)
assert table_names
assert table_names.isdisjoint(MIGRATION_OWNED_TABLES)