fix(collector): reduce critical section duration

By copying the deployment dict the lock is released sooner while still
allowing the watcher to process events without interfering with the
collector run.

Author: michelesr

schedule_scaling/main.py

@@ -300,8 +300,11 @@ def collect_scaling_jobs(cls, ds: DeploymentStore, queue: Queue) -> None:
 
         while not shutdown:
             with ds.lock:
-                for deployment, schedule_action in ds.deployments.items():
-                    process_deployment(deployment, schedule_action, queue)
+                # work on a copy so that we can release the lock sooner
+                deployments = ds.deployments.copy()
+
+            for deployment, schedule_action in deployments.items():
+                process_deployment(deployment, schedule_action, queue)
             logging.debug(f"queue items: {list(queue.queue)}")
             # wait until next minute but wake up if you have to shutdown
             with cls.condition: