vllm-project/vllm · error · RuntimeError
DBO is not yet supported in elastic EP
Error message
DBO is not yet supported in elastic EP
What it means
Error "DBO is not yet supported in elastic EP" thrown in vllm-project/vllm.
Source
Thrown at vllm/distributed/elastic_ep/elastic_execute.py:354
def _commit_staged_moe_quant_methods(self) -> None:
model = self.worker.model_runner.get_model()
moe_modules = [module for module in model.modules() if is_moe_layer(module)]
for module in moe_modules:
staged_quant_method = self._staged_moe_quant_methods.pop(module, None)
if staged_quant_method is None:
continue
assert staged_quant_method.moe_kernel is not None
module._replace_quant_method(staged_quant_method)
staged_quant_method.moe_kernel.prepare_finalize.on_commit()
self._staged_moe_quant_methods.clear()
def _release_cuda_graphs(self) -> None:
if isinstance(self.worker.model_runner.model, CUDAGraphWrapper):
wrapper = self.worker.model_runner.model
wrapper.concrete_cudagraph_entries = {}
elif isinstance(self.worker.model_runner.model, UBatchWrapper):
raise RuntimeError("DBO is not yet supported in elastic EP")
torch.compiler.reset()
with set_current_vllm_config(self.worker.vllm_config):
reset_compile_wrapper(self.worker.model_runner.get_model())
gc.collect()
torch.accelerator.synchronize()
torch.accelerator.empty_cache()
def switch_and_remove(self) -> None:
self._release_cuda_graphs()
_replace_active_groups(world=None, dp=None, ep=None, eplb=None, node_count=None)
def switch_and_prepare(self) -> None:
old_dp_size = get_dp_group().world_size
old_ep_size = get_ep_group().world_size
self._release_cuda_graphs()View on GitHub (pinned to c794754062)
Solutions
- Disable DBO (dual batch overlap) when using elastic EP; it is not supported yet.
When it happens
Trigger: Raised at vllm/distributed/elastic_ep/elastic_execute.py:354 when validation fails: DBO is not yet supported in elastic EP. Typically triggered by an incompatible or incomplete vLLM configuration, an unsupported platform/backend combination, or a runtime resource/dependency that is missing.
Common situations: Commonly encountered at vllm/distributed/elastic_ep/elastic_execute.py:354 during vLLM startup/config validation or runtime setup when: (1) conflicting CLI flags or config fields are combined, (2) the current platform (CUDA/ROCm/CPU/XPU) or installed optional packages do not support the requested feature, or (3) a required value is absent or out of range. Resolve by correcting the configuration as described in the message, or by selecting a supported alternative.
AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14).
Data as JSON: /api/errors/cb2e5a877efd0c12.
Report an issue: GitHub.