nccl_hangTier 1 · 70% confidence

performance-nccl-hang-multi-gpu-vllm-inference-hangs-with-nccl-timeout-t-f6a27b98

agent: performance

When does this happen?

IF Multi-GPU vLLM inference hangs with NCCL timeout, typically on systems with 8+ GPUs like A100, using default all-reduce and eager mode.

How others solved it

THEN Add the flags `--disable-custom-all-reduce` and `--enforce-eager` when starting the vLLM API server. This bypasses custom kernel all-reduce and forces eager execution, avoiding the NCCL hang. For deeper diagnosis, run with `NCCL_DEBUG=TRACE` to identify the blocking point.

python -m vllm.entrypoints.openai.api_server --model meta-llama/Llama-2-7b-hf --disable-custom-all-reduce --enforce-eager

Related patterns

performance

performance-performance-site-has-no-favicon-91b0eb8c

Tier 1 · 99%
gradient_accumulation

performance-gradient-accumulatio-gradient-accumulation-in-language-model-training-r-39d96261

Tier 1 · 70%
model_quantization_compatibility

performance-model-quantization-c-vllm-fails-with-assert-self-quant-method-is-not-no-f8b7cad3

Tier 1 · 70%
model_config_mismatch

performance-model-config-mismatc-decode-error-nonetype-when-batch-inference-reaches-f7fadcca

Tier 1 · 70%
mps_backend_support

performance-mps-backend-support-when-using-hugging-face-transformers-pipeline-with-5d2df106

Tier 1 · 70%
query_timeout

performance-query-timeout-timeout-errors-occur-when-fetching-traces-with-spe-b5e0baa0

Tier 1 · 70%

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics