vllm - 💡(How to fix) Fix [Bug]: vllm-glm5.1 with llmcache with glm-5-fp8 failed

Your current environment

ubuntu22.04 vllm/vllm-openai:glm51

🐛 Describe the bug

[Bug]: vllm-glm5.1 with llmcache with glm-5-fp8 failed [31;20m[2026-05-16 13:54:01,727] LMCache ERROR: LMCacheEngine marked as init failed: Traceback (most recent call last): (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/manager.py", line 208, in post_init (Worker_TP0 pid=120) self._lmcache_engine.post_init(async_lookup_server=async_lookup_server) (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/cache_engine.py", line 301, in post_init (Worker_TP0 pid=120) self.storage_manager = StorageManager( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/storage_manager.py", line 246, in init (Worker_TP0 pid=120) self.create_backends() (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/storage_manager.py", line 1224, in create_backends (Worker_TP0 pid=120) new_backends = CreateStorageBackends( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/init.py", line 160, in CreateStorageBackends (Worker_TP0 pid=120) local_cpu_backend = LocalCPUBackend( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/local_cpu_backend.py", line 67, in init (Worker_TP0 pid=120) self.initialize_allocator(config, metadata) # type: ignore (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/local_cpu_backend.py", line 434, in initialize_allocator (Worker_TP0 pid=120) return MixedMemoryAllocator( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/memory_management.py", line 2070, in init (Worker_TP0 pid=120) self.buffer = _allocate_cpu_memory(size, self.numa_mapping, self.shm_name) (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/memory_management.py", line 428, in _allocate_cpu_memory (Worker_TP0 pid=120) ptr = alloc_fn(size, *alloc_args) (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) RuntimeError: cudaHostAlloc failed: 2 (Worker_TP0 pid=120) (cache_engine.py:277:lmcache.v1.cache_engine) (Worker_TP0 pid=120) [2026-05-16 13:54:01,728] LMCache ERROR: Failed during post_init: Traceback (most recent call last): (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/manager.py", line 208, in post_init (Worker_TP0 pid=120) self._lmcache_engine.post_init(async_lookup_server=async_lookup_server) (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/cache_engine.py", line 301, in post_init (Worker_TP0 pid=120) self.storage_manager = StorageManager( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/storage_manager.py", line 246, in init (Worker_TP0 pid=120) self.create_backends() (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/storage_manager.py", line 1224, in create_backends (Worker_TP0 pid=120) new_backends = CreateStorageBackends( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/init.py", line 160, in CreateStorageBackends (Worker_TP0 pid=120) local_cpu_backend = LocalCPUBackend( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/local_cpu_backend.py", line 67, in init (Worker_TP0 pid=120) self.initialize_allocator(config, metadata) # type: ignore (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/storage_backend/local_cpu_backend.py", line 434, in initialize_allocator (Worker_TP0 pid=120) return MixedMemoryAllocator( (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/memory_management.py", line 2070, in init (Worker_TP0 pid=120) self.buffer = _allocate_cpu_memory(size, self.numa_mapping, self.shm_name) (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) File "/usr/local/lib/python3.12/dist-packages/lmcache/v1/memory_management.py", line 428, in _allocate_cpu_memory (Worker_TP0 pid=120) ptr = alloc_fn(size, *alloc_args) (Worker_TP0 pid=120) ^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Worker_TP0 pid=120) RuntimeError: cudaHostAlloc failed: 2 (Worker_TP0 pid=120) . System will operate in degraded mode (recompute). (manager.py:235:lmcache.v1.manager) (Worker_TP0 pid=120) ERROR 05-16 13:54:01 [multiproc_executor.py:949] WorkerProc hit an exception.

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.