fix watchdog

iaai_scraper/scraper.py

@@ -11,6 +11,7 @@ from concurrent.futures import ThreadPoolExecutor, as_completed
 from concurrent.futures import TimeoutError as FuturesTimeoutError
 from datetime import datetime, timezone
 from pathlib import Path
+from threading import Event, Thread
 from typing import Any, Callable
 from urllib.parse import urlsplit, urlunsplit
 from urllib.request import Request, urlopen
@@ -97,6 +98,33 @@ class _PageOpWatchdog:
         return False
 
 
+class _ProgressHeartbeat:
+    """Периодически пульсует progress во время долгой навигации."""
+
+    def __init__(self, report_progress: Callable[[str, Any], None], stage: str, interval_s: float = 15.0, **meta: Any) -> None:
+        self._report_progress = report_progress
+        self._stage = stage
+        self._interval_s = max(5.0, float(interval_s))
+        self._meta = meta
+        self._stop = Event()
+        self._thread: Thread | None = None
+
+    def _run(self) -> None:
+        while not self._stop.wait(self._interval_s):
+            self._report_progress(self._stage, **self._meta)
+
+    def __enter__(self):
+        self._thread = Thread(target=self._run, name=f"progress-heartbeat-{self._stage}", daemon=True)
+        self._thread.start()
+        return self
+
+    def __exit__(self, exc_type, exc, tb):
+        self._stop.set()
+        if self._thread is not None:
+            self._thread.join(timeout=1.0)
+        return False
+
+
 class IAAIScraper:
 
     @staticmethod
@@ -453,16 +481,23 @@ class IAAIScraper:
         )
 
         for expected_page in range(2, target_page_number + 1):
-            # Heartbeat для watchdog: при deep-resume (100+ кликов) задача может
-            # идти несколько минут без batch_upserted, поэтому пульсуем прогресс.
-            if expected_page == 2 or expected_page == target_page_number or expected_page % 5 == 0:
-                self._report_progress(
-                    "listing_resume_progress",
-                    current_page=expected_page - 1,
-                    target_page=target_page_number,
-                )
+            self._report_progress(
+                "listing_resume_progress",
+                current_page=expected_page - 1,
+                target_page=target_page_number,
+                next_page_number=expected_page,
+            )
 
-            if not self.listing_collector.go_to_next_page(page, expected_page_number=expected_page):
+            with _ProgressHeartbeat(
+                self._report_progress,
+                "listing_resume_progress",
+                current_page=expected_page - 1,
+                target_page=target_page_number,
+                next_page_number=expected_page,
+            ):
+                next_ok = self.listing_collector.go_to_next_page(page, expected_page_number=expected_page)
+
+            if not next_ok:
                 self._report_progress(
                     "listing_resume_failed",
                     current_page=expected_page - 1,
@@ -471,6 +506,13 @@ class IAAIScraper:
                 page.close()
                 raise ListingResumeError(f"Failed to resume listing at page {target_page_number}")
 
+            self._report_progress(
+                "listing_resume_progress",
+                current_page=expected_page,
+                target_page=target_page_number,
+                next_page_number=min(target_page_number, expected_page + 1),
+            )
+
         self._report_progress(
             "listing_resume_completed",
             current_page=target_page_number,
@@ -973,8 +1015,17 @@ class IAAIScraper:
                     cars_failed=cars_failed,
                 )
                 try:
-                    with _PageOpWatchdog(_PAGE_NEXT_TIMEOUT_S, f"next page {page_number + 1}"):
-                        next_ok = self.listing_collector.go_to_next_page(page, expected_page_number=page_number + 1)
+                    with _ProgressHeartbeat(
+                        self._report_progress,
+                        "listing_next_page_started",
+                        page_number=page_number,
+                        next_page_number=page_number + 1,
+                        pending_urls=len(pending_urls),
+                        cars_upserted=cars_upserted,
+                        cars_failed=cars_failed,
+                    ):
+                        with _PageOpWatchdog(_PAGE_NEXT_TIMEOUT_S, f"next page {page_number + 1}"):
+                            next_ok = self.listing_collector.go_to_next_page(page, expected_page_number=page_number + 1)
                 except (PageOperationTimeoutError, PlaywrightError) as nav_exc:
                     logger.warning(
                         "go_to_next_page stalled/failed at page %d (%s) — will reopen",