Files
live-hub-py/web/backend/routers/automation.py
T
yml2213 f745a61df1 feat(automation): CK 自动巡检(P1a),只检测不改 CK
后台原本靠人工逐批点「CK 检测」发现失效账号,1.2 万个账号根本盯不过来。
P1a 把「哪些账号失效了」自动化:自动挑久未检测的账号、调现成检测接口、
把结果写回 ck_check_*。**本作业不替换任何 CK**,重登仍由人工点,符合规划里
「先做只读/只提示,确认无误再做自动动作」的渐进上线原则。

实现:
- 迁移 20261002_0037:automation_jobs / automation_runs 两张表,
  login_tasks 加 relogin_attempts / relogin_next_at / relogin_last_error
  (为 P1b 备好),accounts 加 auto_managed(特殊账号不参与自动化);
- services/automation/scheduler.py:进程内调度线程,60s tick,状态全部落库
  (重启不丢调度进度、不重复跑轮);AUTOMATION_ENABLED 默认关闭,作业开关
  也默认关闭;ensure_jobs 只补缺失行、不覆盖运维改过的配置;
  作业异常只记 failed,不拖垮调度线程;单容器 --workers 1,无需分布式锁;
- services/automation/jobs/ck_health.py:选行口径与人工检测一致
  (cookie_holding_filter:持有 CK 且无在途尝试),叠加久未检测、
  auto_managed、batch_size,按 ck_checked_at 升序(最久没查的优先);
  只写 ck_check_status / ck_check_result / ck_checked_at;
- routers/automation.py:GET /jobs、PATCH /jobs/{key}、POST /jobs/{key}/run、
  GET /runs;新增权限点 automation:view(运营可看)与 automation:manage
  (仅超管),改配置与手动触发都写审计;
- 顺带补 POST /api/login/batch 的审计(此前批次是谁发起、多少账号无据可查);
- 前端 /automation 页:作业卡片(开关/周期/单轮条数/并发)、运行历史、
  「立即跑一次」;总开关未打开、调度线程未运行时给明确提示。

验证:
- 22 条新测试覆盖调度器行为(只跑到点的、异常不抛出、下轮时间、配置不被重置)、
  选行条件(在途尝试/auto_managed/间隔/batch_size)、接口与权限,其中一条
  专门断言巡检后 cookie/status/message/attempt_status 一字未动;
- 全量 257 条通过;前端 tsc 与构建通过;
- 端到端用 3 个真实账号跑真实斗鱼接口:巡检结果与直接调 check_douyu_cookie
  逐条一致(3/3 valid),CK 与尝试状态全程未变。

顺带修掉一处会炸生产的坑:DB 时间列读出来是 naive UTC,直接与
datetime.now(UTC) 比较会 TypeError,现统一走 _as_utc() 补时区。
2026-10-02 19:47:41 +08:00

192 lines
6.1 KiB
Python

"""自动化作业接口。
设计上「先能看、再能开」:查看需要 ``automation:view``(运营也有),改配置与
手动触发需要 ``automation:manage``(仅超管)。作业默认关闭,打开后由进程内
调度器按间隔执行。
"""
from __future__ import annotations
from datetime import UTC, datetime
from typing import Any
from fastapi import APIRouter, Depends, HTTPException, Query
from pydantic import BaseModel, Field
from sqlalchemy.orm import Session
from ..database import get_db
from ..deps import require_permission
from ..models import AutomationJob, AutomationRun, User
from ..services.audit_service import record_audit
from ..services.automation import (
ensure_jobs,
registered_jobs,
run_job_now,
scheduler_status,
)
router = APIRouter(prefix="/api/automation", tags=["自动化"])
MAX_BATCH_SIZE = 2000
MIN_INTERVAL_SECONDS = 300
class JobUpdate(BaseModel):
"""作业配置更新;只传要改的字段。"""
enabled: bool | None = None
interval_seconds: int | None = Field(default=None, ge=MIN_INTERVAL_SECONDS)
batch_size: int | None = Field(default=None, ge=1, le=MAX_BATCH_SIZE)
concurrency: int | None = Field(default=None, ge=1, le=16)
def _job_out(job: AutomationJob) -> dict[str, Any]:
return {
"key": job.key,
"title": job.title,
"enabled": bool(job.enabled),
"interval_seconds": int(job.interval_seconds or 0),
"batch_size": int(job.batch_size or 0),
"concurrency": int(job.concurrency or 0),
"last_run_at": job.last_run_at.isoformat() if job.last_run_at else None,
"next_run_at": job.next_run_at.isoformat() if job.next_run_at else None,
"last_status": job.last_status or "",
"last_message": job.last_message or "",
"registered": job.key in registered_jobs(),
}
def _run_out(run: AutomationRun) -> dict[str, Any]:
return {
"id": run.id,
"job_key": run.job_key,
"started_at": run.started_at.isoformat() if run.started_at else None,
"finished_at": run.finished_at.isoformat() if run.finished_at else None,
"status": run.status or "",
"scanned": int(run.scanned or 0),
"valid": int(run.valid or 0),
"invalid": int(run.invalid or 0),
"skipped": int(run.skipped or 0),
"message": run.message or "",
"detail": run.detail or None,
}
@router.get("/jobs")
def list_jobs(
db: Session = Depends(get_db),
current: User = Depends(require_permission("automation:view")),
):
"""列出作业配置、最近一次结果与调度器状态。"""
ensure_jobs(db)
jobs = db.query(AutomationJob).order_by(AutomationJob.key.asc()).all()
latest: dict[str, AutomationRun] = {}
for run in (
db.query(AutomationRun).order_by(AutomationRun.id.desc()).limit(200).all()
):
latest.setdefault(run.job_key, run)
return {
"scheduler": scheduler_status(),
"jobs": [
{
**_job_out(job),
"latest_run": _run_out(latest[job.key]) if job.key in latest else None,
}
for job in jobs
],
}
@router.patch("/jobs/{key}")
def update_job(
key: str,
payload: JobUpdate,
db: Session = Depends(get_db),
current: User = Depends(require_permission("automation:manage")),
):
"""改开关 / 间隔 / 批量 / 并发。打开开关时排定下一轮时间。"""
ensure_jobs(db)
job = db.query(AutomationJob).filter(AutomationJob.key == key).first()
if job is None:
raise HTTPException(status_code=404, detail="作业不存在")
changed: dict[str, Any] = {}
if payload.enabled is not None and bool(payload.enabled) != bool(job.enabled):
job.enabled = bool(payload.enabled)
changed["enabled"] = job.enabled
# 刚打开时排一轮:不早于一个 tick,避免打开瞬间就压满 CPU。
job.next_run_at = datetime.now(UTC) + _tick_delta() if job.enabled else None
if payload.interval_seconds is not None:
changed["interval_seconds"] = payload.interval_seconds
job.interval_seconds = payload.interval_seconds
if payload.batch_size is not None:
changed["batch_size"] = payload.batch_size
job.batch_size = payload.batch_size
if payload.concurrency is not None:
changed["concurrency"] = payload.concurrency
job.concurrency = payload.concurrency
if changed:
job.updated_at = datetime.now(UTC)
record_audit(
db,
current,
action="automation:update",
target=key,
detail=changed,
)
db.commit()
return _job_out(job)
@router.post("/jobs/{key}/run")
def run_job(
key: str,
db: Session = Depends(get_db),
current: User = Depends(require_permission("automation:manage")),
):
"""立刻跑一轮(不看 enabled)。用于正式打开开关前先试跑。"""
ensure_jobs(db)
try:
outcome = run_job_now(db, key)
except KeyError:
raise HTTPException(status_code=404, detail="作业不存在") from None
record_audit(
db,
current,
action="automation:run",
target=key,
detail={
"status": outcome.get("status"),
"scanned": outcome.get("scanned"),
"valid": outcome.get("valid"),
"invalid": outcome.get("invalid"),
"message": outcome.get("message"),
},
)
db.commit()
return outcome
@router.get("/runs")
def list_runs(
job_key: str | None = Query(default=None),
limit: int = Query(default=50, ge=1, le=200),
db: Session = Depends(get_db),
current: User = Depends(require_permission("automation:view")),
):
"""运行历史,最近的在前。"""
query = db.query(AutomationRun)
if job_key:
query = query.filter(AutomationRun.job_key == job_key)
runs = query.order_by(AutomationRun.id.desc()).limit(limit).all()
return {"items": [_run_out(run) for run in runs]}
def _tick_delta():
from datetime import timedelta
from ..services.automation.scheduler import TICK_SECONDS
return timedelta(seconds=TICK_SECONDS)