fix(scrapers): ban/stuck-ротация для cian+yandex (provider-aware _rotate_proxy_ip)
All checks were successful
CI / changes (pull_request) Successful in 6s
CI / backend-tests (pull_request) Has been skipped
CI / frontend-tests (pull_request) Has been skipped
CI / openapi-codegen-check (pull_request) Has been skipped

This commit is contained in:
bot-backend 2026-06-26 19:10:26 +03:00
parent 197068c515
commit c1cdf3969c
3 changed files with 371 additions and 18 deletions

View file

@ -359,6 +359,9 @@ class Settings(BaseSettings):
# changeip-ссылка для Cian-прокси (ротация IP при бане/таймауте). Если не задан —
# fallback на avito_proxy_rotate_url. ENV: CIAN_PROXY_ROTATE_URL.
cian_proxy_rotate_url: str | None = None
# Максимум IP-ротаций для Cian на один sweep-прогон. Аналог avito_proxy_max_rotations.
# ENV: CIAN_PROXY_MAX_ROTATIONS.
cian_proxy_max_rotations: int = 4
@property
def cian_proxy_url(self) -> str | None:
@ -373,6 +376,9 @@ class Settings(BaseSettings):
# changeip-ссылка для Yandex-прокси (ротация IP при капче/таймауте). Если не задан —
# fallback на avito_proxy_rotate_url. ENV: YANDEX_PROXY_ROTATE_URL.
yandex_proxy_rotate_url: str | None = None
# Максимум IP-ротаций для Yandex на один sweep-прогон. Аналог avito_proxy_max_rotations.
# ENV: YANDEX_PROXY_MAX_ROTATIONS.
yandex_proxy_max_rotations: int = 4
@property
def yandex_proxy_url(self) -> str | None:

View file

@ -53,15 +53,29 @@ from app.services.yandex_price_history import record_yandex_price_history
logger = logging.getLogger(__name__)
async def _rotate_proxy_ip(*, reason: str, rotations_done: int) -> bool:
"""Вызвать changeip-ссылку mobileproxy и подождать settle (#1790).
async def _rotate_proxy_ip(*, reason: str, rotations_done: int, source: str = "avito") -> bool:
"""Вызвать changeip-ссылку mobileproxy и подождать settle (#1790/#1848).
Provider-aware: выбирает rotate_url и max_rotations по параметру `source`
(avito / cian / yandex). Fallback-цепочка для rotate_url:
avito: avito_proxy_rotate_url
cian: cian_proxy_rotate_url avito_proxy_rotate_url
yandex: yandex_proxy_rotate_url avito_proxy_rotate_url
Аналог AvitoScraper._rotate_ip, но на уровне pipeline используется в
enrichment-фазах (houses + detail) где нет доступа к экземпляру scraper'а.
enrichment-фазах где нет доступа к экземпляру scraper'а.
Returns True при успешной смене IP, False если rotate_url не задан или ошибка.
Логирует каждую ротацию (причина, порядковый номер, остаток лимита).
Логирует каждую ротацию (причина, порядковый номер, source, остаток лимита).
"""
if source == "cian":
rotate_url = settings.cian_proxy_rotate_url or settings.avito_proxy_rotate_url
max_rot = settings.cian_proxy_max_rotations
elif source == "yandex":
rotate_url = settings.yandex_proxy_rotate_url or settings.avito_proxy_rotate_url
max_rot = settings.yandex_proxy_max_rotations
else:
# avito (default) — backward compat
rotate_url = settings.avito_proxy_rotate_url
max_rot = settings.avito_proxy_max_rotations
if not rotate_url:
@ -78,7 +92,9 @@ async def _rotate_proxy_ip(*, reason: str, rotations_done: int) -> bool:
pass
await asyncio.sleep(settings.avito_proxy_rotate_settle_s)
logger.info(
"pipeline: IP rotated via changeip — reason=%s rotation=#%d remaining=%d new_ip=%s",
"pipeline: IP rotated via changeip — source=%s reason=%s "
"rotation=#%d remaining=%d new_ip=%s",
source,
reason,
rotations_done + 1,
max_rot - rotations_done - 1,
@ -87,7 +103,8 @@ async def _rotate_proxy_ip(*, reason: str, rotations_done: int) -> bool:
return True
except Exception:
logger.warning(
"pipeline: IP rotation failed (reason=%s, rotation=#%d)",
"pipeline: IP rotation failed (source=%s reason=%s rotation=#%d)",
source,
reason,
rotations_done + 1,
exc_info=True,
@ -759,9 +776,7 @@ async def run_avito_city_sweep(
scraper._browser = shared_bf
# Shared-browser режим: _cffi=None → curl_cffi-fallback на
# firewall пропускается (by design). Логируем для observability.
logger.debug(
"avito pipeline-mode: shared browser only, no cffi fallback"
)
logger.debug("avito pipeline-mode: shared browser only, no cffi fallback")
else:
scraper._cffi = session
@ -1482,6 +1497,7 @@ async def run_yandex_city_sweep(
enrich_delay = request_delay_sec if request_delay_sec is not None else 3.0
_resolved_delay = request_delay_sec if request_delay_sec is not None else 9.0
consecutive_failures = 0
yandex_rotations_done = 0 # #1848: бюджет IP-ротаций на весь sweep
# Вычисляем watchdog-таймаут для combos-режима (центр, anchors=None).
# В combos-режиме один "anchor" выполняет num_segments × num_combos × max_pages
@ -1565,7 +1581,7 @@ async def run_yandex_city_sweep(
каждого (room×price) combo, сохраняет порцию сразу в БД + обновляет heartbeat.
anchor_lots накапливается (через _al) для address-enrich и price-history.
"""
nonlocal consecutive_failures
nonlocal consecutive_failures, yandex_rotations_done
# ── Phase 1+2: SERP + инкрементальный save per-combo ──────
def _on_combo(
@ -1675,19 +1691,25 @@ async def run_yandex_city_sweep(
if items:
from curl_cffi.requests import AsyncSession as _AsyncSession
# #1848: rotation tracking для Yandex address-enrich (stuck-proxy)
_yandex_enrich_consec_fail = 0
_yandex_enrich_abort = 3 # abort address-enrich после N подряд
async with _AsyncSession(
impersonate="chrome120",
timeout=30.0,
proxies=_proxies,
headers={"Accept-Language": "ru-RU,ru;q=0.9,en;q=0.8"},
) as enrich_session:
for eidx, item in enumerate(items):
eidx = 0
while eidx < len(items):
item = items[eidx]
lid: int = item["id"]
old_addr: str = item["address"]
url: str = item["source_url"]
try:
resp = await enrich_session.get(url, allow_redirects=True)
if resp.status_code != 200:
if resp.status_code not in {200}:
logger.warning(
"yandex-sweep run_id=%d address-enrich:"
" HTTP %d listing_id=%d",
@ -1696,6 +1718,11 @@ async def run_yandex_city_sweep(
lid,
)
counters.address_failed += 1
# 503/429 = прокси заблокирован → считаем как failure
if resp.status_code in {429, 503}:
_yandex_enrich_consec_fail += 1
else:
_yandex_enrich_consec_fail = 0
else:
new_addr = _extract_address_from_title(resp.text)
if (
@ -1741,17 +1768,53 @@ async def run_yandex_city_sweep(
db.rollback()
except Exception:
pass
_yandex_enrich_consec_fail = 0
except Exception as fetch_exc:
counters.address_failed += 1
_yandex_enrich_consec_fail += 1
logger.warning(
"yandex-sweep run_id=%d: address-enrich "
"fetch failed listing_id=%d: %s",
"fetch failed listing_id=%d "
"(consecutive=%d): %s",
run_id,
lid,
_yandex_enrich_consec_fail,
fetch_exc,
)
# Ротация Yandex-прокси при stuck/ban (#1848)
if _yandex_enrich_consec_fail >= _yandex_enrich_abort:
if yandex_rotations_done < settings.yandex_proxy_max_rotations:
rotated = await _rotate_proxy_ip(
reason=(
f"yandex enrich consecutive="
f"{_yandex_enrich_consec_fail}"
),
rotations_done=yandex_rotations_done,
source="yandex",
)
yandex_rotations_done += 1
if rotated:
_yandex_enrich_consec_fail = 0
logger.info(
"yandex-sweep run_id=%d: enrich ROTATE — "
"IP changed, reset consecutive "
"(rotation #%d/%d)",
run_id,
yandex_rotations_done,
settings.yandex_proxy_max_rotations,
)
# retry текущего item после ротации
continue
logger.error(
"yandex-sweep run_id=%d: address-enrich ABORT — "
"%d consecutive failures (proxy likely stuck/banned)",
run_id,
_yandex_enrich_consec_fail,
)
break
if eidx < len(items) - 1:
await asyncio.sleep(enrich_delay)
eidx += 1
logger.info(
"yandex-sweep run_id=%d anchor %s: address enrich=%d/%d failed=%d",
@ -1954,6 +2017,7 @@ async def run_cian_city_sweep(
_anchors = anchors if anchors is not None else EKB_ANCHORS
counters = CianCitySweepCounters(anchors_total=len(_anchors))
consecutive_failures = 0
cian_rotations_done = 0 # #1848: бюджет IP-ротаций на весь sweep
try:
for idx, (lat, lon, name) in enumerate(_anchors, start=1):
@ -1994,7 +2058,7 @@ async def run_cian_city_sweep(
_a_name: str = _c_name,
) -> None:
"""Все фазы одного cian anchor'а (SERP + detail + houses)."""
nonlocal anchor_lots, consecutive_failures
nonlocal anchor_lots, consecutive_failures, cian_rotations_done
from sqlalchemy import text as _text
# ── Phase 1+2: SERP + save ─────────────────────────────────
@ -2051,7 +2115,12 @@ async def run_cian_city_sweep(
.mappings()
.all()
)
for didx, row in enumerate(priority_rows):
# #1848: Cian bans/stuck-proxy tracking + rotation (зеркало avito)
_cian_detail_consec_failures = 0
_cian_detail_abort = 3 # abort detail-фазы после N подряд
didx = 0
while didx < len(priority_rows):
row = priority_rows[didx]
listing_id: int = row["id"]
source_url: str = row["source_url"]
counters.detail_attempted += 1
@ -2069,18 +2138,54 @@ async def run_cian_city_sweep(
listing_id,
source_url,
)
_cian_detail_consec_failures = 0
except Exception as exc:
counters.detail_failed += 1
counters.errors_count += 1
_cian_detail_consec_failures += 1
logger.warning(
"cian-sweep run_id=%d: detail failed listing_id=%d: %s",
"cian-sweep run_id=%d: detail failed listing_id=%d "
"(consecutive=%d): %s",
run_id,
listing_id,
_cian_detail_consec_failures,
exc,
)
if _cian_detail_consec_failures >= _cian_detail_abort:
# Попытка ротации Cian-прокси перед abort'ом (#1848)
if cian_rotations_done < settings.cian_proxy_max_rotations:
rotated = await _rotate_proxy_ip(
reason=(
f"cian detail consecutive="
f"{_cian_detail_consec_failures}"
),
rotations_done=cian_rotations_done,
source="cian",
)
cian_rotations_done += 1
if rotated:
_cian_detail_consec_failures = 0
logger.info(
"cian-sweep run_id=%d: detail ROTATE — "
"IP changed, reset consecutive "
"(rotation #%d/%d)",
run_id,
cian_rotations_done,
settings.cian_proxy_max_rotations,
)
# retry текущего detail (не инкрементируем didx)
continue
logger.error(
"cian-sweep run_id=%d: detail ABORT — "
"%d consecutive failures (proxy likely banned/stuck)",
run_id,
_cian_detail_consec_failures,
)
break
if didx < len(priority_rows) - 1:
jitter = random.uniform(0.8, 1.2)
await asyncio.sleep(request_delay_sec * jitter)
didx += 1
logger.info(
"cian-sweep run_id=%d anchor %s: detail=%d/%d failed=%d",
run_id,

View file

@ -0,0 +1,242 @@
"""Тесты Fix 2: provider-aware _rotate_proxy_ip (#1848).
Тест 1: source='cian' cian_proxy_rotate_url используется.
Тест 2: source='yandex' yandex_proxy_rotate_url используется.
Тест 3: source='avito' (default) avito_proxy_rotate_url (backward compat).
Тест 4: cian без cian url fallback на avito_proxy_rotate_url.
Тест 5: rotate_url=None returns False (no HTTP call).
Тест 6: network error returns False (no crash).
Тест 7: yandex fallback на avito.
"""
from __future__ import annotations
import os
from unittest.mock import AsyncMock, MagicMock, patch
import pytest
os.environ.setdefault("DATABASE_URL", "postgresql+psycopg://test:test@localhost/test_db")
from app.services.scrape_pipeline import _rotate_proxy_ip
# ── Helpers ──────────────────────────────────────────────────────────────────
def _async_session_ctx(resp_json: dict | None = None):
"""Возвращает (async_cm_class, session_mock) для curl_cffi AsyncSession."""
session = AsyncMock()
if resp_json is not None:
mock_resp = MagicMock()
mock_resp.json.return_value = resp_json
session.get = AsyncMock(return_value=mock_resp)
else:
session.get = AsyncMock(side_effect=RuntimeError("network error"))
# curl_cffi AsyncSession используется как `async with AsyncSession(...) as rot:`
# Нам нужен callable класс-замена, вызов которого возвращает async context manager.
class _MockAsyncSession:
def __init__(self, timeout=30):
pass
async def __aenter__(self):
return session
async def __aexit__(self, *_):
pass
return _MockAsyncSession, session
# ── Тест 1: cian_proxy_rotate_url ────────────────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_cian_uses_cian_url() -> None:
"""source='cian' → cian_proxy_rotate_url (не avito_proxy_rotate_url)."""
mock_session_cls, session = _async_session_ctx(resp_json={"new_ip": "1.2.3.4"})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
patch("app.services.scrape_pipeline.asyncio.sleep", AsyncMock()),
):
s.cian_proxy_rotate_url = "http://cian-rotate.test/"
s.avito_proxy_rotate_url = "http://avito-rotate.test/"
s.yandex_proxy_rotate_url = None
s.cian_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.yandex_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
result = await _rotate_proxy_ip(reason="ban", rotations_done=0, source="cian")
assert result is True
called_url: str = session.get.call_args[0][0]
assert "cian-rotate.test" in called_url
assert "avito-rotate.test" not in called_url
# ── Тест 2: yandex_proxy_rotate_url ─────────────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_yandex_uses_yandex_url() -> None:
"""source='yandex' → yandex_proxy_rotate_url."""
mock_session_cls, session = _async_session_ctx(resp_json={"new_ip": "5.6.7.8"})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
patch("app.services.scrape_pipeline.asyncio.sleep", AsyncMock()),
):
s.yandex_proxy_rotate_url = "http://yandex-rotate.test/"
s.avito_proxy_rotate_url = "http://avito-rotate.test/"
s.cian_proxy_rotate_url = None
s.yandex_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.cian_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
result = await _rotate_proxy_ip(reason="timeout", rotations_done=0, source="yandex")
assert result is True
called_url: str = session.get.call_args[0][0]
assert "yandex-rotate.test" in called_url
assert "avito-rotate.test" not in called_url
# ── Тест 3: avito backward compat ────────────────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_avito_backward_compat() -> None:
"""source не передан (дефолт 'avito') → avito_proxy_rotate_url."""
mock_session_cls, session = _async_session_ctx(resp_json={"new_ip": "9.9.9.9"})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
patch("app.services.scrape_pipeline.asyncio.sleep", AsyncMock()),
):
s.avito_proxy_rotate_url = "http://avito-only.test/"
s.cian_proxy_rotate_url = None
s.yandex_proxy_rotate_url = None
s.avito_proxy_max_rotations = 4
s.cian_proxy_max_rotations = 4
s.yandex_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
# Вызов без source= — должен работать как раньше
result = await _rotate_proxy_ip(reason="block", rotations_done=0)
assert result is True
called_url: str = session.get.call_args[0][0]
assert "avito-only.test" in called_url
# ── Тест 4: cian fallback на avito когда нет cian url ─────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_cian_fallback_to_avito() -> None:
"""cian_proxy_rotate_url=None → fallback на avito_proxy_rotate_url."""
mock_session_cls, session = _async_session_ctx(resp_json={})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
patch("app.services.scrape_pipeline.asyncio.sleep", AsyncMock()),
):
s.cian_proxy_rotate_url = None
s.avito_proxy_rotate_url = "http://avito-fallback.test/"
s.yandex_proxy_rotate_url = None
s.cian_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.yandex_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
result = await _rotate_proxy_ip(reason="ban", rotations_done=0, source="cian")
assert result is True
called_url: str = session.get.call_args[0][0]
assert "avito-fallback.test" in called_url
# ── Тест 5: rotate_url=None → returns False ──────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_returns_false_when_no_url() -> None:
"""Никакой URL не задан → False, HTTP-запрос не делается."""
mock_session_cls, session = _async_session_ctx(resp_json={})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
):
s.cian_proxy_rotate_url = None
s.avito_proxy_rotate_url = None
s.yandex_proxy_rotate_url = None
s.cian_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.yandex_proxy_max_rotations = 4
result = await _rotate_proxy_ip(reason="ban", rotations_done=0, source="cian")
assert result is False
# HTTP-запрос не делался
session.get.assert_not_called()
# ── Тест 6: network error → returns False ────────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_returns_false_on_network_error() -> None:
"""Ошибка сети → returns False (не крашит caller)."""
mock_session_cls, _session = _async_session_ctx(resp_json=None) # raises RuntimeError
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
):
s.cian_proxy_rotate_url = "http://cian-rotate.test/"
s.avito_proxy_rotate_url = None
s.yandex_proxy_rotate_url = None
s.cian_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.yandex_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
result = await _rotate_proxy_ip(reason="ban", rotations_done=0, source="cian")
assert result is False
# ── Тест 7: yandex fallback на avito ─────────────────────────────────────────
@pytest.mark.asyncio
async def test_rotate_proxy_ip_yandex_fallback_to_avito() -> None:
"""yandex_proxy_rotate_url=None → fallback на avito_proxy_rotate_url."""
mock_session_cls, session = _async_session_ctx(resp_json={"new_ip": "11.22.33.44"})
with (
patch("app.services.scrape_pipeline.settings") as s,
patch("app.services.scrape_pipeline.AsyncSession", mock_session_cls),
patch("app.services.scrape_pipeline.asyncio.sleep", AsyncMock()),
):
s.yandex_proxy_rotate_url = None
s.avito_proxy_rotate_url = "http://shared-proxy.test/"
s.cian_proxy_rotate_url = None
s.yandex_proxy_max_rotations = 4
s.avito_proxy_max_rotations = 4
s.cian_proxy_max_rotations = 4
s.avito_proxy_rotate_settle_s = 0.0
result = await _rotate_proxy_ip(reason="503", rotations_done=0, source="yandex")
assert result is True
called_url: str = session.get.call_args[0][0]
assert "shared-proxy.test" in called_url