[Bug][Failing Test]: V1 - NixlConnector `test_prompt_less_than_block_size`

[robertgshaw2-redhat]

Your current environment

pytest -v -x v1/kv_connector/unit/test_nixl_connector.py::test_prompt_less_than_block_size

🐛 Describe the bug

Likely due to changes with hybrid memory allocator

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[markmc]

Failed tests:

FAILED v1/kv_connector/unit/test_nixl_connector.py::test_prompt_less_than_block_size - AssertionError: assert 1 == 0
 +  where 1 = len({'id-1': ReqMeta(local_block_ids=[], remote_block_ids=[], remote_host='my-host', remote_port=1234, remote_engine_id='my-engine-id')})
 +    where {'id-1': ReqMeta(local_block_ids=[], remote_block_ids=[], remote_host='my-host', remote_port=1234, remote_engine_id='my-engine-id')} = <vllm.distributed.kv_transfer.kv_connector.v1.nixl_connector.NixlConnectorMetadata object at 0x7efac57a9700>.requests

Logs:

``` =================================== FAILURES =================================== _______________________ test_prompt_less_than_block_size _______________________

def test_prompt_less_than_block_size():
    """
    Test that we can handle case where prompt is < block.

    In this case, the P worker will send empty remote_block_ids.
    The D worker should not schedule an async read in this case,
    since there is nothing to pull.
    """
    vllm_config = create_vllm_config()
    scheduler = create_scheduler(vllm_config)

    # Half of a block.
    BLOCK_SIZE = vllm_config.cache_config.block_size
    NUM_TOKENS = int(BLOCK_SIZE * 0.5)

    # Request will have 0 remote blocks.
    request = create_request(request_id=1,
                             num_tokens=NUM_TOKENS,
                             do_remote_prefill=True,
                             num_remote_blocks=0)
    scheduler.add_request(request)
    scheduler_output = scheduler.schedule()

    # This request should not have to read async.
    kv_connector_metadata = scheduler_output.kv_connector_metadata
    assert kv_connector_metadata is not None
    assert isinstance(kv_connector_metadata, NixlConnectorMetadata)

  assert len(kv_connector_metadata.requests) == 0

E AssertionError: assert 1 == 0
E + where 1 = len({'id-1': ReqMeta(local_block_ids=[], remote_block_ids=[], remote_host='my-host', remote_port=1234, remote_engine_id='my-engine-id')})
E + where {'id-1': ReqMeta(local_block_ids=[], remote_block_ids=[], remote_host='my-host', remote_port=1234, remote_engine_id='my-engine-id')} = <vllm.distributed.kv_transfer.kv_connector.v1.nixl_connector.NixlConnectorMetadata object at 0x7efac57a9700>.requests

v1/kv_connector/unit/test_nixl_connector.py:70: AssertionError
=============================== warnings summary ===============================

</details>

[DarkLight1337]

Should be fixed by #18429?