Pin autoscaling requester (ray-project#42254)

Pin the Ray Data autoscaling requester to the head node so that the actor cannot prevent down-scaling of an idle worker node. Signed-off-by: Stephanie Wang <[email protected]>
c21 · Jan 11, 2024 · 52e71fb · 52e71fb
1 parent cf7e4c0
commit 52e71fb
Showing 1 changed file with 1 addition and 2 deletions.
diff --git a/python/ray/data/_internal/execution/autoscaling_requester.py b/python/ray/data/_internal/execution/autoscaling_requester.py
@@ -119,8 +119,7 @@ def get_or_create_autoscaling_requester_actor():
  # point to the head node.
  scheduling_strategy = NodeAffinitySchedulingStrategy(
  ray.get_runtime_context().get_node_id(),
- soft=True,
- _spill_on_unavailable=True,
+ soft=False,
  )
  with _autoscaling_requester_lock:
  return AutoscalingRequester.options(