What happened + What you expected to happen

I am starting Ray with a small object store memory size, while preventing object spilling. Then, I do ray.put on a large numpy array.

I expect this to fail, since the array is much larger than the object store. However, it works fine.

If I look at the Ray processes, I can see that the data has been written to the disk at some point.

However, I don't see anything in /tmp/ray/session... like when spilling is enabled, so I'm not really sure what is happening.

Versions / Dependencies

Ray 2.46.0
Python 3.12.3
Ubuntu

Reproduction script

import ray
import numpy as np
import time

ray.init(object_store_memory=100 * 1024 * 1024, _system_config={'automatic_object_spilling_enabled': False})

ref = ray.put(np.random.random((200, 1024, 1024)))

time.sleep(1000)

This is the method suggested in #21998

Issue Severity

Medium: It is a significant difficulty but I can work around it.