[MRv2] Default to MRv1 when a connector is present (#42955)

Signed-off-by: NickLucche <nlucches@redhat.com>

tests/test_config.py

@@ -148,6 +148,13 @@ def test_is_default_v2_model_runner_model(model_config, expected):
     assert VllmConfig._is_default_v2_model_runner_model(config) is expected
 
 
+def test_use_v2_model_runner_defaults_to_v1_when_kv_connector_present():
+    config = SimpleNamespace(kv_transfer_config=object())
+    with patch.object(envs, "VLLM_USE_V2_MODEL_RUNNER", None):
+        result = VllmConfig.use_v2_model_runner.fget(config)
+    assert result is False
+
+
 @pytest.mark.skip_global_cleanup
 def test_with_hf_config_populates_missing_architectures_from_causal_lm_mapping(
     monkeypatch,

vllm/config/vllm.py

@@ -498,6 +498,10 @@ class VllmConfig:
         if use_v2_model_runner is not None:
             return use_v2_model_runner
 
+        # KVCache layout changes are breaking, let's stick with v1 for now (see #42846)
+        if self.kv_transfer_config is not None:
+            return False
+
         if not self._is_default_v2_model_runner_model():
             return False