search

vllm-project / vllm

A high-throughput and memory-efficient inference and serving engine for LLMs
https://docs.vllm.ai
Apache License 2.0
27k stars 3.96k forks source link

[Bug]: No available block found in 60 second in shm #6614

Open wjj19950828 opened 2 months ago

wjj19950828 commented 2 months ago

Your current environment

Collecting environment information... PyTorch version: 2.3.0+cu121 Is debug build: False CUDA used to build PyTorch: 12.1 ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04.3 LTS (x86_64) GCC version: (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0 Clang version: 14.0.0-1ubuntu1.1 CMake version: version 3.24.4 Libc version: glibc-2.35

Python version: 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] (64-bit runtime) Python platform: Linux-5.10.112-005.ali5000.alios7.x86_64-x86_64-with-glibc2.35 Is CUDA available: True CUDA runtime version: 12.3.107 CUDA_MODULE_LOADING set to: LAZY GPU models and configuration: GPU 0: NVIDIA A100-SXM4-80GB GPU 1: NVIDIA A100-SXM4-80GB GPU 2: NVIDIA A100-SXM4-80GB GPU 3: NVIDIA A100-SXM4-80GB GPU 4: NVIDIA A100-SXM4-80GB GPU 5: NVIDIA A100-SXM4-80GB GPU 6: NVIDIA A100-SXM4-80GB GPU 7: NVIDIA A100-SXM4-80GB

Nvidia driver version: 515.105.01 cuDNN version: Probably one of the following: /usr/lib/x86_64-linux-gnu/libcudnn.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.9.4 /usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.9.4 HIP runtime version: N/A MIOpen runtime version: N/A Is XNNPACK available: True

CPU: Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 46 bits physical, 57 bits virtual Byte Order: Little Endian CPU(s): 128 On-line CPU(s) list: 0-127 Vendor ID: GenuineIntel BIOS Vendor ID: Intel(R) Corporation Model name: Intel(R) Xeon(R) Platinum 8369B CPU @ 2.90GHz BIOS Model name: Intel(R) Xeon(R) Platinum 8369B CPU @ 2.90GHz CPU family: 6 Model: 106 Thread(s) per core: 2 Core(s) per socket: 32 Socket(s): 2 Stepping: 6 CPU max MHz: 3500.0000 CPU min MHz: 800.0000 BogoMIPS: 5800.00 Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 invpcid_single intel_ppin ssbd mba ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm rdt_a avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb intel_pt avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local split_lock_detect wbnoinvd dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req avx512vbmi umip pku ospke avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg tme avx512_vpopcntdq rdpid fsrm md_clear pconfig flush_l1d arch_capabilities Virtualization: VT-x L1d cache: 3 MiB (64 instances) L1i cache: 2 MiB (64 instances) L2 cache: 80 MiB (64 instances) L3 cache: 96 MiB (2 instances) NUMA node(s): 1 NUMA node0 CPU(s): 0-127 Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization Vulnerability Spectre v2: Vulnerable: eIBRS with unprivileged eBPF Vulnerability Srbds: Not affected Vulnerability Tsx async abort: Not affected

Versions of relevant libraries: [pip3] mypy==1.9.0 [pip3] mypy-extensions==1.0.0 [pip3] numpy==1.26.2 [pip3] nvidia-nccl-cu12==2.20.5 [pip3] onnx==1.16.0 [pip3] onnx-graphsurgeon==0.3.27 [pip3] onnxruntime==1.16.3 [pip3] torch==2.3.0 [pip3] triton==2.3.0 [pip3] tritonclient==2.44.0 [conda] Could not collectROCM Version: Could not collect Neuron SDK Version: N/A vLLM Version: 0.5.2 vLLM Build Flags: CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled GPU Topology: GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 NIC0 NIC1 NIC2 NIC3 CPU Affinity NUMA Affinity GPU0 X NV12 NV12 NV12 NV12 NV12 NV12 NV12 PXB SYS SYS SYS 0-127 N/A GPU1 NV12 X NV12 NV12 NV12 NV12 NV12 NV12 PXB SYS SYS SYS 0-127 N/A GPU2 NV12 NV12 X NV12 NV12 NV12 NV12 NV12 SYS PXB SYS SYS 0-127 N/A GPU3 NV12 NV12 NV12 X NV12 NV12 NV12 NV12 SYS PXB SYS SYS 0-127 N/A GPU4 NV12 NV12 NV12 NV12 X NV12 NV12 NV12 SYS SYS PXB SYS 0-127 N/A GPU5 NV12 NV12 NV12 NV12 NV12 X NV12 NV12 SYS SYS PXB SYS 0-127 N/A GPU6 NV12 NV12 NV12 NV12 NV12 NV12 X NV12 SYS SYS SYS PXB 0-127 N/A GPU7 NV12 NV12 NV12 NV12 NV12 NV12 NV12 X SYS SYS SYS PXB 0-127 N/A NIC0 PXB PXB SYS SYS SYS SYS SYS SYS X SYS SYS SYS NIC1 SYS SYS PXB PXB SYS SYS SYS SYS SYS X SYS SYS NIC2 SYS SYS SYS SYS PXB PXB SYS SYS SYS SYS X SYS NIC3 SYS SYS SYS SYS SYS SYS PXB PXB SYS SYS SYS X

Legend:

X = Self SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI) NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU) PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge) PIX = Connection traversing at most a single PCIe bridge NV# = Connection traversing a bonded set of # NVLinks

NIC Legend:

NIC0: mlx5_bond_0 NIC1: mlx5_bond_1 NIC2: mlx5_bond_2 NIC3: mlx5_bond_3

🐛 Describe the bug

Now because it involves shm communication, are there any current requirements for the size of shm?

Currently, In v0.5.2, there will be a random timeout phenomenon. The specific reason is that shm

image

The shm size in the container is 90G. Is there a recommended shm size? Or is it due to other reasons?

youkaichao commented 1 month ago

I think the root cause should be the engine iteration timeout error.

[Bug]: No available block found in 60 second

this is just a warning.

DarkLight1337 commented 1 month ago

Please report this error under the tracking issue #5901 so we can get more data on how to fix it.

G-z-w commented 1 month ago

I also encountered this problem, I hope it can be solved. I printed out the rank information at the warning and found that one of the GPUs was stuck (a total of 4 GPUs for tensor parallelism). This bug is highly reproducible, especially when running models above 70B and encountering a large number of requests.

image

image

youkaichao commented 1 month ago

@G-z-w please report to #5901 .

wjj19950828 commented 1 month ago

@youkaichao Please solve this problem as soon as possible. At present, it seems that the latest version of vLLM always has some distributed errors~

DarkLight1337 commented 1 month ago

Reopening so that it's not seen as being resolved by #5901 just yet.

wjj19950828 commented 1 month ago

@DarkLight1337 @youkaichao Is there any progress on this issue? What is the specific cause?

I debugged and found that the stack is stuck at the CA stage. What is the specific cause?

Thanks～

youkaichao commented 1 month ago

What do you mean by CA?

wjj19950828 commented 1 month ago

What do you mean by CA?

Custom Allreduce, stuck at self.all_reduce_unreg

youkaichao commented 1 month ago

If you want a quick fix, you can directly disable custom allreduce by --disable-custom-allreduce

Debugging that part is quite difficult, and I cannot give any timeline for it.

hengxinCheung commented 1 month ago

I also encountered this problem, I hope it can be solved. I printed out the rank information at the warning and found that one of the GPUs was stuck (a total of 4 GPUs for tensor parallelism). This bug is highly reproducible, especially when running models above 70B and encountering a large number of requests.

image

image

It can not always blamed on shm_broadcast. AsyncEngine start workers in advance, and shm_broadcast will not receive inputs forever if current step is stuck.

G-z-w commented 1 month ago

I also encountered this problem, I hope it can be solved. I printed out the rank information at the warning and found that one of the GPUs was stuck (a total of 4 GPUs for tensor parallelism). This bug is highly reproducible, especially when running models above 70B and encountering a large number of requests. image image

It can not always blamed on shm_broadcast. AsyncEngine start workers in advance, and shm_broadcast will not receive inputs forever if current step is stuck.

Thanks, I ran into the same issue when I replaced AsyncEngine with LLM_Engine.

carljones3000 commented 1 month ago

Seems like a large prompt can put the vllm server into a state where it just doesn't work anymore. Even small prompts stop working once this error (below) happens.

I'm also seeing the "No available block found in 60 second" which is why I'm adding this on to this thread.

python -m vllm.entrypoints.openai.api_server --model Qwen/Qwen2-72B-Instruct --tensor-parallel-size 8 --max-model-len 12591 --swap-space 1 --disable-custom-all-reduce

I used this "--disable-custom-all-reduce" in hopes that it would be a work around (from comment above) but it did not stop the issue.


INFO 07-28 21:00:40 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:00:50 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:01:00 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:01:10 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:01:20 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:01:30 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
ERROR 07-28 21:01:31 async_llm_engine.py:631] Engine iteration timed out. This should never happen!
ERROR 07-28 21:01:31 async_llm_engine.py:54] Engine background task failed
ERROR 07-28 21:01:31 async_llm_engine.py:54] Traceback (most recent call last):
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 604, in run_engine_loop
ERROR 07-28 21:01:31 async_llm_engine.py:54]     done, _ = await asyncio.wait(
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/asyncio/tasks.py", line 384, in wait
ERROR 07-28 21:01:31 async_llm_engine.py:54]     return await _wait(fs, timeout, return_when, loop)
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/asyncio/tasks.py", line 491, in _wait
ERROR 07-28 21:01:31 async_llm_engine.py:54]     await waiter
ERROR 07-28 21:01:31 async_llm_engine.py:54] asyncio.exceptions.CancelledError
ERROR 07-28 21:01:31 async_llm_engine.py:54] 
ERROR 07-28 21:01:31 async_llm_engine.py:54] During handling of the above exception, another exception occurred:
ERROR 07-28 21:01:31 async_llm_engine.py:54] 
ERROR 07-28 21:01:31 async_llm_engine.py:54] Traceback (most recent call last):
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 44, in _log_task_completion
ERROR 07-28 21:01:31 async_llm_engine.py:54]     return_value = task.result()
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 603, in run_engine_loop
ERROR 07-28 21:01:31 async_llm_engine.py:54]     async with asyncio_timeout(ENGINE_ITERATION_TIMEOUT_S):
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_timeout.py", line 95, in __aexit__
ERROR 07-28 21:01:31 async_llm_engine.py:54]     self._do_exit(exc_type)
ERROR 07-28 21:01:31 async_llm_engine.py:54]   File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_timeout.py", line 178, in _do_exit
ERROR 07-28 21:01:31 async_llm_engine.py:54]     raise asyncio.TimeoutError
ERROR 07-28 21:01:31 async_llm_engine.py:54] asyncio.exceptions.TimeoutError
Exception in callback functools.partial(<function _log_task_completion at 0x782a4a391e10>, error_callback=<bound method AsyncLLMEngine._error_callback of <vllm.engine.async_llm_engine.AsyncLLMEngine object at 0x782a32679060>>)
handle: <Handle functools.partial(<function _log_task_completion at 0x782a4a391e10>, error_callback=<bound method AsyncLLMEngine._error_callback of <vllm.engine.async_llm_engine.AsyncLLMEngine object at 0x782a32679060>>)>
Traceback (most recent call last):
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 604, in run_engine_loop
    done, _ = await asyncio.wait(
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/asyncio/tasks.py", line 384, in wait
    return await _wait(fs, timeout, return_when, loop)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/asyncio/tasks.py", line 491, in _wait
    await waiter
asyncio.exceptions.CancelledError

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 44, in _log_task_completion
    return_value = task.result()
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 603, in run_engine_loop
    async with asyncio_timeout(ENGINE_ITERATION_TIMEOUT_S):
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_timeout.py", line 95, in __aexit__
    self._do_exit(exc_type)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_timeout.py", line 178, in _do_exit
    raise asyncio.TimeoutError
asyncio.exceptions.TimeoutError

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "uvloop/cbhandles.pyx", line 63, in uvloop.loop.Handle._run
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/vllm/engine/async_llm_engine.py", line 56, in _log_task_completion
    raise AsyncEngineDeadError(
vllm.engine.async_llm_engine.AsyncEngineDeadError: Task finished unexpectedly. This should never happen! Please open an issue on Github. See stack trace above for theactual cause.
INFO 07-28 21:01:31 async_llm_engine.py:169] Aborted request cmpl-af3bf93597c44c508f515b4b20851d2b.
ERROR:    Exception in ASGI application
Traceback (most recent call last):
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/responses.py", line 265, in __call__
    await wrap(partial(self.listen_for_disconnect, receive))
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/responses.py", line 261, in wrap
    await func()
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/responses.py", line 238, in listen_for_disconnect
    message = await receive()
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/uvicorn/protocols/http/httptools_impl.py", line 553, in receive
    await self.message_event.wait()
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/asyncio/locks.py", line 214, in wait
    await fut
asyncio.exceptions.CancelledError: Cancelled by cancel scope 782a2c14e590

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/uvicorn/protocols/http/httptools_impl.py", line 399, in run_asgi
    result = await app(  # type: ignore[func-returns-value]
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/uvicorn/middleware/proxy_headers.py", line 70, in __call__
    return await self.app(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/fastapi/applications.py", line 1054, in __call__
    await super().__call__(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/applications.py", line 123, in __call__
    await self.middleware_stack(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/middleware/errors.py", line 186, in __call__
    raise exc
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/middleware/errors.py", line 164, in __call__
    await self.app(scope, receive, _send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/middleware/cors.py", line 85, in __call__
    await self.app(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/middleware/exceptions.py", line 65, in __call__
    await wrap_app_handling_exceptions(self.app, conn)(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
    raise exc
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
    await app(scope, receive, sender)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/routing.py", line 756, in __call__
    await self.middleware_stack(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/routing.py", line 776, in app
    await route.handle(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/routing.py", line 297, in handle
    await self.app(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/routing.py", line 77, in app
    await wrap_app_handling_exceptions(app, request)(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 64, in wrapped_app
    raise exc
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/_exception_handler.py", line 53, in wrapped_app
    await app(scope, receive, sender)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/routing.py", line 75, in app
    await response(scope, receive, send)
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/starlette/responses.py", line 258, in __call__
    async with anyio.create_task_group() as task_group:
  File "/home/user2/software/anaconda3/envs/vllmenv/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 680, in __aexit__
    raise BaseExceptionGroup(
exceptiongroup.ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
(VllmWorkerProcess pid=103446) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103440) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103445) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103442) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103443) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103441) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
(VllmWorkerProcess pid=103444) WARNING 07-28 21:01:32 shm_broadcast.py:394] No available block found in 60 second. 
INFO 07-28 21:01:40 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:01:45 metrics.py:295] Avg prompt throughput: 887.9 tokens/s, Avg generation throughput: 0.2 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:02:00 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:02:10 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation throughput: 0.0 tokens/s, Running: 1 reqs, Swapped: 0 reqs, Pending: 0 reqs, GPU KV cache usage: 9.1%, CPU KV cache usage: 0.0%.
INFO 07-28 21:02:20 metrics.py:295] Avg prompt throughput: 0.0 tokens/s, Avg generation```
SnzFor16Min commented 1 month ago

Same problem with batch_size around 32, and --disable-custom-all-reduce seems to work. But what else does this option affect? Any update on this to work without --disable-custom-all-reduce? (Another similar issue in #6818)

kkk-an commented 4 weeks ago

I meet the same error "No available block found in 60 second" in shm, so far is there any method to avoid this error?

youkaichao commented 4 weeks ago

@kkk-an No available block found in 60 second is just a warning. The root cause is the vllm engine is stuck somewhere. You can try to follow https://docs.vllm.ai/en/latest/getting_started/debugging.html to debug where it is stuck.

njhill commented 4 weeks ago

@kkk-an can you confirm that you’re using the latest version

kkk-an commented 4 weeks ago

@kkk-an No available block found in 60 second is just a warning. The root cause is the vllm engine is stuck somewhere. You can try to follow https://docs.vllm.ai/en/latest/getting_started/debugging.html to debug where it is stuck.

@kkk-an can you confirm that you’re using the latest version

Hello, thanks for your kindly reply. The version I used is vllm==0.5.4, torch==2.4.0. I just ran llama3-70b inference with run_batch, met the same error as the top one "No available block found in 60 second" in shm, and raised the same "TimeoutError", but I found adding '--disable-custom-all-reduce' to my command works for me.

Thank you!

zhaotyer commented 4 weeks ago

I meet the same error "No available block found in 60 second" in shm,The version I used is vllm==0.5.3.post1+cu118, torch==2.3.1+cu118.

heibaidaolx123 commented 13 hours ago

same error, with vllm-openai:v0.6.1.post2 docker image, when serving qwen2-vl-72B on 4*A40. vllm hung after worker on GPU0 died, no error log.