feat(cleanup): 解耦 request_candidates 与 provider_api_keys 生命周期

- 移除 request_candidates.key_id 对 provider_api_keys 的外键约束(含迁移脚本)
- 删除 Key 时不再级联删除候选记录,改为独立按保留天数定时清理
- 新增 request_candidates_retention_days / request_candidates_cleanup_batch_size 配置项
- batch_delete_task 增加 lock_timeout 及超时自动降批重试机制
- cleanup_key_references 提取阶段化清理流程,移除 RequestCandidate 联动删除
- 前端 CleanupPolicySection 新增候选记录保留天数和清理批次配置

Closes #227

Co-authored-by: Entropy-Xu <entropy.xu@cloudhabitatsh.com>
This commit is contained in:
fawney19
2026-03-14 01:31:39 +08:00
parent bdfe4adc98
commit 776dd2f8ea
14 changed files with 516 additions and 70 deletions

View File

@@ -122,10 +122,12 @@ def test_sync_delete_reports_progress_after_each_batch(
self.rowcounts = [2, 1]
self.commits = 0
self.closed = False
self.text_statements: list[str] = []
def execute(self, _statement: object) -> SimpleNamespace:
# SET LOCAL statement_timeout 不消耗 rowcount
# SET LOCAL statement_timeout / lock_timeout 不消耗 rowcount
if hasattr(_statement, "text"):
self.text_statements.append(_statement.text)
return SimpleNamespace(rowcount=0)
return SimpleNamespace(rowcount=self.rowcounts.pop(0))
@@ -146,7 +148,7 @@ def test_sync_delete_reports_progress_after_each_batch(
monkeypatch.setattr(
taskmod,
"cleanup_key_references",
lambda _db, batch: cleanup_batches.append(list(batch)),
lambda _db, batch, **_kwargs: cleanup_batches.append(list(batch)),
)
monkeypatch.setattr("src.database.create_session", lambda: session)
monkeypatch.setattr(
@@ -165,4 +167,85 @@ def test_sync_delete_reports_progress_after_each_batch(
assert progress_updates == [2, 3]
assert cleanup_batches == [["key-1", "key-2"], ["key-3"]]
assert session.commits == 2
assert session.text_statements == [
"SET LOCAL statement_timeout = '30000'",
"SET LOCAL lock_timeout = '5000'",
"SET LOCAL statement_timeout = '30000'",
"SET LOCAL lock_timeout = '5000'",
]
assert session.closed is True
def test_sync_delete_retries_with_smaller_batches_after_timeout(
monkeypatch: pytest.MonkeyPatch,
) -> None:
class _FakeColumn:
def __eq__(self, other: object) -> tuple[str, object]: # type: ignore[override]
return ("eq", other)
def in_(self, values: list[str]) -> tuple[str, tuple[str, ...]]:
return ("in", tuple(values))
class _FakeProviderAPIKey:
provider_id = _FakeColumn()
id = _FakeColumn()
class _FakeDeleteStatement:
def where(self, *_conditions: object) -> "_FakeDeleteStatement":
return self
class _FakeSession:
def __init__(self) -> None:
self.rowcounts = [2, 2]
self.commits = 0
self.rollbacks = 0
self.closed = False
def execute(self, _statement: object) -> SimpleNamespace:
if hasattr(_statement, "text"):
return SimpleNamespace(rowcount=0)
return SimpleNamespace(rowcount=self.rowcounts.pop(0))
def commit(self) -> None:
self.commits += 1
def rollback(self) -> None:
self.rollbacks += 1
def close(self) -> None:
self.closed = True
session = _FakeSession()
progress_updates: list[int] = []
cleanup_batches: list[list[str]] = []
def fake_cleanup(_db: object, batch: list[str], **_kwargs: object) -> None:
cleanup_batches.append(list(batch))
if len(batch) >= 4:
raise RuntimeError(
"(psycopg2.errors.QueryCanceled) canceling statement due to statement timeout"
)
monkeypatch.setattr(taskmod, "_CLEANUP_BATCH_SIZE", 4)
monkeypatch.setattr(taskmod, "_MIN_RETRY_BATCH_SIZE", 1)
monkeypatch.setattr(taskmod, "cleanup_key_references", fake_cleanup)
monkeypatch.setattr("src.database.create_session", lambda: session)
monkeypatch.setattr("src.models.database.ProviderAPIKey", _FakeProviderAPIKey)
monkeypatch.setattr(taskmod, "sa_delete", lambda _model: _FakeDeleteStatement())
affected = taskmod._sync_delete(
"provider-1",
["key-1", "key-2", "key-3", "key-4"],
progress_updates.append,
)
assert affected == 4
assert progress_updates == [4]
assert cleanup_batches == [
["key-1", "key-2", "key-3", "key-4"],
["key-1", "key-2"],
["key-3", "key-4"],
]
assert session.commits == 2
assert session.rollbacks == 1
assert session.closed is True

View File

@@ -287,13 +287,12 @@ def test_cleanup_key_references_preserves_usage_and_video_tasks(
side_effects_module.cleanup_key_references(cast(Any, db), ["key-1", "key-2"])
assert [(stmt.kind, stmt.model.__name__) for stmt in db.statements] == [
("delete", "RequestCandidate"),
("delete", "GeminiFileMapping"),
("update", "Usage"),
("update", "VideoTask"),
]
assert db.statements[2].values_dict == {"provider_api_key_id": None}
assert db.statements[3].values_dict == {"key_id": None}
assert db.statements[1].values_dict == {"provider_api_key_id": None}
assert db.statements[2].values_dict == {"key_id": None}
@pytest.mark.asyncio

View File

@@ -2,6 +2,7 @@ from __future__ import annotations
import inspect
from types import SimpleNamespace
from unittest.mock import MagicMock
import pytest
@@ -41,3 +42,76 @@ async def test_maintenance_scheduler_start_skips_startup_task_when_disabled(
await scheduler.start()
assert created is False
@pytest.mark.asyncio
async def test_candidate_cleanup_uses_dedicated_retention_and_batch_settings(
monkeypatch: pytest.MonkeyPatch,
) -> None:
scheduler = MaintenanceScheduler()
class _FakeLoop:
async def run_in_executor(self, _executor, func): # type: ignore[no-untyped-def]
return func()
class _ConfigSession:
def close(self) -> None:
return None
class _BatchSession:
def __init__(self, ids: list[str]) -> None:
self.ids = ids
self.closed = False
self.committed = False
self.query_obj = MagicMock()
filtered = self.query_obj.filter.return_value
filtered.order_by.return_value.limit.return_value.all.return_value = [
SimpleNamespace(id=value) for value in ids
]
def query(self, _model): # type: ignore[no-untyped-def]
return self.query_obj
def execute(self, _statement): # type: ignore[no-untyped-def]
return SimpleNamespace(rowcount=len(self.ids))
def commit(self) -> None:
self.committed = True
def rollback(self) -> None:
raise AssertionError("rollback should not be called")
def close(self) -> None:
self.closed = True
config_session = _ConfigSession()
batch_one = _BatchSession(["candidate-1", "candidate-2"])
batch_two = _BatchSession([])
sessions = iter([config_session, batch_one, batch_two])
def fake_create_session(): # type: ignore[no-untyped-def]
return next(sessions)
config_values = {
"enable_auto_cleanup": True,
"request_candidates_retention_days": 21,
"request_candidates_cleanup_batch_size": 2,
}
monkeypatch.setattr(maintenance_scheduler_module, "create_session", fake_create_session)
monkeypatch.setattr(
maintenance_scheduler_module.SystemConfigService,
"get_config",
lambda _db, key, default=None: config_values.get(key, default),
)
monkeypatch.setattr(
maintenance_scheduler_module.asyncio, "get_running_loop", lambda: _FakeLoop()
)
await scheduler._perform_candidate_cleanup()
batch_one.query_obj.filter.return_value.order_by.return_value.limit.return_value.all.assert_called_once()
batch_one.query_obj.filter.return_value.order_by.return_value.limit.assert_called_once_with(2)
assert batch_one.committed is True
assert batch_one.closed is True
assert batch_two.closed is True