304 lines
12 KiB
Python
304 lines
12 KiB
Python
"""CK 巡检作业(P1a)。
|
||
|
||
只做一件事:把「持有 CK 但久未检测」的账号挑出来,调现成的
|
||
``check_douyu_cookie()`` 检测,把结果写回 ``ck_check_*``。
|
||
|
||
**本作业不修改 CK 本身**:不写 ``cookie`` / ``status`` / ``message``,也不碰
|
||
``attempt_*``。P1a 上线后人工照旧可以点检测和重登,自动重登是 P1b 的事。
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
from concurrent.futures import as_completed
|
||
from datetime import UTC, datetime, timedelta
|
||
from typing import Any
|
||
|
||
from sqlalchemy import and_, case, func, or_
|
||
from sqlalchemy.orm import Session
|
||
|
||
from ....models import Account, AutomationJob
|
||
from ...account_lifecycle import MANUAL_STATES, WORK_LOGIN_RELOGIN
|
||
from ...account_protection_service import load_protection_snapshot
|
||
from ...account_value import (
|
||
HIGH_VALUE_TIER_MAX,
|
||
high_value_recheck_seconds,
|
||
value_tier_expr,
|
||
)
|
||
from ...account_work_queue import enqueue_account_work
|
||
from ...cookie_check_service import (
|
||
check_douyu_cookie,
|
||
)
|
||
from ...cookie_check_service import (
|
||
classify_douyu_cookie_result as classify_result, # noqa: F401 - 兼容诊断入口
|
||
)
|
||
from ...douyu_account_probe import (
|
||
persist_probe_result,
|
||
probe_session,
|
||
)
|
||
from ...douyu_service import cookie_holding_filter, cookie_operable_filter
|
||
from ...task_runtime import TaskLane, submit
|
||
|
||
JOB_KEY = "ck_health"
|
||
FLUSH_EVERY = 25
|
||
|
||
# 非鉴权类失败(网络超时、响应异常、被判风控)占比超过它就自动暂停作业。
|
||
# 这类失败说明「我们查不到」,不是「这个 CK 失效了」;如果不加区分地写
|
||
# invalid,斗鱼接口抖动一小时就能把上千个好账号标成失效,客服会误判成
|
||
# CK 大面积挂掉。
|
||
ERROR_PAUSE_RATIO = 0.3
|
||
ERROR_PAUSE_MIN_SCANNED = 20
|
||
|
||
|
||
def _candidates_query(db: Session, job: AutomationJob, now: datetime):
|
||
"""本轮「待检测账号」的完整候选集(**不带 LIMIT**,供计数复用)。
|
||
|
||
口径与人工点检测一致(``cookie_operable_filter``:账号有 CK 且没有在途尝试),
|
||
再加三条:距上次检测超过间隔、账号允许自动管理、**未被停用/归档**。
|
||
第 1、2 档用 6 小时间隔,其余用作业上配置的间隔。
|
||
|
||
排序刻意不用 ``NULLS FIRST``:那是 PostgreSQL/SQLite 语法,**MySQL 会报
|
||
1064 语法错误**(生产踩过一次)。改用 ``ck_checked_at IS NULL DESC``,
|
||
三个方言下语义一致且都是合法 SQL。
|
||
"""
|
||
interval = max(int(job.interval_seconds or 0), 60)
|
||
cutoff = now - timedelta(seconds=interval)
|
||
# 第 1、2 档 6 小时查一次(作业间隔更短时跟着更短的走),其余用作业间隔。
|
||
# 到点的高价值号先占本轮名额,剩下的再按「最久没查」填。
|
||
hv_cutoff = now - timedelta(seconds=high_value_recheck_seconds(interval))
|
||
tier = value_tier_expr()
|
||
protection = load_protection_snapshot(
|
||
db,
|
||
job_config=getattr(job, "config", None),
|
||
)
|
||
protected_ids = protection.monitored_ids
|
||
due_filter = or_(
|
||
Account.ck_checked_at.is_(None),
|
||
and_(tier <= HIGH_VALUE_TIER_MAX, Account.ck_checked_at < hv_cutoff),
|
||
and_(tier > HIGH_VALUE_TIER_MAX, Account.ck_checked_at < cutoff),
|
||
)
|
||
if protected_ids:
|
||
protected_due = []
|
||
for group_key, group in protection.groups.items():
|
||
group_ids = protection.group_account_ids.get(group_key, frozenset())
|
||
if not group["enabled"] or not group_ids:
|
||
continue
|
||
group_cutoff = now - timedelta(
|
||
seconds=int(group["probe_interval_minutes"]) * 60
|
||
)
|
||
protected_due.append(
|
||
and_(
|
||
Account.id.in_(group_ids),
|
||
or_(
|
||
Account.ck_checked_at.is_(None),
|
||
Account.ck_checked_at < group_cutoff,
|
||
),
|
||
)
|
||
)
|
||
due_filter = or_(
|
||
*protected_due,
|
||
due_filter,
|
||
)
|
||
|
||
query = (
|
||
db.query(Account)
|
||
.filter(cookie_operable_filter())
|
||
.filter(Account.auto_managed.is_(True))
|
||
# invalid 已进入自动重登队列,manual 已停止自动处理;普通 CK 探针
|
||
# 再扫它们只会重复消耗官方请求,并掩盖真正需要发现的新失效。
|
||
.filter(Account.ck_check_status.notin_(("invalid", "manual")))
|
||
# 生命周期状态机:停用/归档的账号不参与任何自动化(见 account_lifecycle)。
|
||
.filter(Account.status.notin_(MANUAL_STATES))
|
||
.filter(due_filter)
|
||
)
|
||
if protected_ids:
|
||
query = query.order_by(
|
||
case((Account.id.in_(protected_ids), 0), else_=1),
|
||
tier.asc(),
|
||
Account.ck_checked_at.is_(None).desc(),
|
||
Account.ck_checked_at.asc(),
|
||
Account.id.asc(),
|
||
)
|
||
else:
|
||
query = query.order_by(
|
||
tier.asc(),
|
||
Account.ck_checked_at.is_(None).desc(),
|
||
Account.ck_checked_at.asc(),
|
||
Account.id.asc(),
|
||
)
|
||
return query
|
||
|
||
|
||
def build_select_query(db: Session, job: AutomationJob, now: datetime):
|
||
"""构造本轮要检测**账号**的查询(与取行分开,便于按方言编译做回归)。
|
||
|
||
解耦后**按账号选行**(2026-10-04):CK 与 ``ck_check_*`` 都在 ``accounts`` 上,
|
||
以前是 ``db.query(LoginTask).join(Account)`` —— 那样「没有登录历史的账号」
|
||
永远进不了巡检队列,而且一个账号有多行时会被重复检测。
|
||
"""
|
||
return _candidates_query(db, job, now).limit(max(int(job.batch_size or 0), 1))
|
||
|
||
|
||
def _select_accounts(db: Session, job: AutomationJob, now: datetime) -> list[Account]:
|
||
return build_select_query(db, job, now).all()
|
||
|
||
|
||
def count_remaining(db: Session, job: AutomationJob, now: datetime) -> int:
|
||
"""剩余待检账号数(给页面提示用)。
|
||
|
||
刻意**不用** ``build_select_query`` —— 那个带 ``LIMIT batch_size``,
|
||
取 count 只会得到批次上限,完全看不出积压。
|
||
"""
|
||
return int(
|
||
_candidates_query(db, job, now).order_by(None).with_entities(Account.id).count()
|
||
or 0
|
||
)
|
||
|
||
|
||
def _detect(account: Account) -> dict[str, Any]:
|
||
try:
|
||
return probe_session(
|
||
account.cookie or "",
|
||
checker=check_douyu_cookie,
|
||
)
|
||
except Exception as exc: # noqa: BLE001 单个账号失败不影响整轮
|
||
return {
|
||
"status": "error",
|
||
"message": f"检测异常: {exc}",
|
||
"checked_at": datetime.now(UTC).isoformat(),
|
||
"detail": {},
|
||
}
|
||
|
||
|
||
def run_ck_health(db: Session, job: AutomationJob) -> dict[str, Any]:
|
||
now = datetime.now(UTC)
|
||
protection = load_protection_snapshot(
|
||
db,
|
||
job_config=getattr(job, "config", None),
|
||
)
|
||
monitored_ids = protection.monitored_ids
|
||
relogin_ids = protection.relogin_ids
|
||
accounts = _select_accounts(db, job, now)
|
||
if not accounts:
|
||
earliest_checked = (
|
||
db.query(func.min(Account.ck_checked_at))
|
||
.filter(cookie_holding_filter())
|
||
.scalar()
|
||
)
|
||
msg = "全库暂无到期待检账号"
|
||
if earliest_checked:
|
||
if earliest_checked.tzinfo is None:
|
||
earliest_checked = earliest_checked.replace(tzinfo=UTC)
|
||
hours_ago = (now - earliest_checked).total_seconds() / 3600
|
||
recheck_hours = max(int(job.interval_seconds or 0), 60) / 3600
|
||
msg = f"全库暂无到期待检账号(最久前检测于 {hours_ago:.1f} 小时前,重检间隔为 {recheck_hours:.1f} 小时)"
|
||
return {"scanned": 0, "valid": 0, "invalid": 0, "skipped": 0, "message": msg}
|
||
|
||
results: dict[int, dict[str, Any]] = {}
|
||
concurrency = max(min(int(job.concurrency or 1), 16), 1)
|
||
futures = {}
|
||
for account in accounts:
|
||
future = submit(TaskLane.ACCOUNT_CHECK, _detect, account)
|
||
if future is not None:
|
||
futures[future] = account
|
||
for future in as_completed(futures):
|
||
account = futures[future]
|
||
results[account.id] = future.result()
|
||
|
||
# 只写检测结果三件套;cookie / status / attempt_* 一律不动。
|
||
valid = invalid = errors = protected_scanned = protected_invalid = 0
|
||
for index, account in enumerate(accounts, start=1):
|
||
result = results.get(account.id)
|
||
if result is None:
|
||
continue
|
||
verdict = str(result.get("status") or "error")
|
||
is_protected = int(account.id) in monitored_ids
|
||
if is_protected:
|
||
protected_scanned += 1
|
||
if verdict == "valid":
|
||
valid += 1
|
||
elif verdict == "invalid":
|
||
invalid += 1
|
||
if is_protected:
|
||
protected_invalid += 1
|
||
else:
|
||
errors += 1
|
||
# 自动巡检不额外刷 account_probe 事件;统一落库仍会处理 CK 状态变化
|
||
# 和生命周期事件,避免每 6 小时给所有账号写一条重复查询事件。
|
||
detail = result.get("detail") if isinstance(result.get("detail"), dict) else {}
|
||
persist_probe_result(
|
||
db,
|
||
account,
|
||
{
|
||
"probe_id": f"ck-health:{now.timestamp()}:{account.id}",
|
||
"identity": {
|
||
"uid": account.uid,
|
||
"username": account.username,
|
||
"nickname": detail.get("nickname") or "",
|
||
"nickname_source": "session",
|
||
},
|
||
"session": result,
|
||
"elite": {},
|
||
"eligibility": {},
|
||
"roles": {},
|
||
"gold": {},
|
||
"goods": {},
|
||
"errors": {},
|
||
},
|
||
source="automation.ck_health",
|
||
record_event=False,
|
||
commit=False,
|
||
)
|
||
if verdict == "invalid" and int(account.id) in relogin_ids:
|
||
# 保护组失效账号优先进入重登队列。on_ck_verdict 会登记默认
|
||
# priority=30,这里把已存在的同类型待办降到 10,不重复建行。
|
||
enqueue_account_work(
|
||
db,
|
||
account_id=int(account.id),
|
||
work_type=WORK_LOGIN_RELOGIN,
|
||
priority=10,
|
||
max_attempts=20,
|
||
due_at=now,
|
||
reason="履约保护账号 CK 失效,优先自动重登",
|
||
)
|
||
|
||
if index % FLUSH_EVERY == 0:
|
||
db.commit()
|
||
db.commit()
|
||
|
||
# 剩余待检数量只用于提示,不参与调度。
|
||
remaining = count_remaining(db, job, now)
|
||
error_ratio = errors / len(accounts) if accounts else 0.0
|
||
paused = False
|
||
if errors >= ERROR_PAUSE_MIN_SCANNED and error_ratio > ERROR_PAUSE_RATIO:
|
||
# 大面积「查不到」通常是斗鱼接口抖动或被风控,继续跑只会污染更多账号。
|
||
job.enabled = False
|
||
job.updated_at = datetime.now(UTC)
|
||
paused = True
|
||
|
||
message = f"本轮检测 {len(accounts)} 条,剩余待检 {remaining} 条"
|
||
if errors:
|
||
message += f";非鉴权类失败 {errors} 条({error_ratio:.0%})"
|
||
if protected_scanned:
|
||
message += (
|
||
f";履约保护账号 {protected_scanned} 条"
|
||
f",失效 {protected_invalid} 条"
|
||
)
|
||
if paused:
|
||
message += ";失败占比过高,已自动暂停作业"
|
||
return {
|
||
"scanned": len(accounts),
|
||
"valid": valid,
|
||
"invalid": invalid,
|
||
"skipped": errors,
|
||
"message": message,
|
||
"detail": {
|
||
"remaining": remaining,
|
||
"concurrency": concurrency,
|
||
"errors": errors,
|
||
"error_ratio": round(error_ratio, 4),
|
||
"paused": paused,
|
||
"protected_scanned": protected_scanned,
|
||
"protected_invalid": protected_invalid,
|
||
},
|
||
}
|