ntohidi f6ab207e25 fix: remove shared LOCK contention in monitor to prevent pod deadlock (#1754) ...

The monitor's update_timeline(), get_health_summary(), and
get_browser_list() all acquired the crawler pool's global LOCK to read
pool stats. That same lock is held during slow browser start/close
operations (get_crawler, janitor, close_all), causing the monitor to
block indefinitely and the pod to become unresponsive after sustained
crawling.

Replaced all three lock acquisitions in monitor.py with a lock-free
get_pool_snapshot() in crawler_pool.py that returns shallow dict copies.
Under CPython's GIL, dict.copy() and len() are atomic — safe for
read-only monitoring with at most slightly stale counts.

2026-03-13 12:17:52 +08:00

..

docker

fix: remove shared LOCK contention in monitor to prevent pod deadlock (#1754)

2026-03-13 12:17:52 +08:00