[Bug]: Curl failed: Received HTTP/0.9 when not allowed

[GreatBryan]

Your current environment

The output of python collect_env.py

Your output of `python collect_env.py` here

🐛 Describe the bug

Issue Description:
In a data-parallel scenario, when the --data-parallel-rpc-port argument value conflicts with the vLLM HTTP service port, the service starts without any error logs despite the port collision.

Actual Behavior:
When clients send requests, they only receive the error message: "Received HTTP/0.9 when not allowed".

Expected Behavior:
The vLLM service should either:

Detect the port conflict during startup and fail with a clear error message, or
Dynamically handle the conflict by selecting an alternative port.

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.

[GreatBryan]

Error screenshot attached below:

[yankay]

Error screenshot attached below:

HI @GreatBryan

Would you please share the commands to reproduce the issue :-)

[GreatBryan]

Error screenshot attached below:

HI @GreatBryan

Would you please share the commands to reproduce the issue :-)
Hi @yankay
vllm start comman:
master node
vllm serve --model="/data/mindspore/deepseek_r1_w8a8_safetensor" --trust_remote_code --max-num-seqs=128 --max_model_len=2048 --max-num-batched-tokens=128 --block-size=128 --gpu-memory-utilization=0.9 --tensor-parallel-size 1 --data-parallel-size 16 --data-parallel-size-local 8 --data-parallel-start-rank 0 --data-parallel-address 10.155.205.167 --data-parallel-rpc-port 12370 --enable-expert-parallel > log 2>&1 &

sub node
vllm serve --model="/data/mindspore/deepseek_r1_w8a8_safetensor" --trust_remote_code --max-num-seqs=128 --max_model_len=2048 --max-num-batched-tokens=128 --block-size=128 --gpu-memory-utilization=0.9 --tensor-parallel-size 1 --data-parallel-size 16 --data-parallel-size-local 8 --data-parallel-start-rank 0 --data-parallel-address 10.155.205.167 --data-parallel-rpc-port 12370 --enable-expert-parallel > log 2>&1 &

curl commad:
curl http://10.155.205.167:8000/v1/completions -H "Content-Type: application/json" -d '{"model": "/data/mindspore/deepseek_r1_w8a8_safetensor", "prompt": "You are a helpful assistant.<｜User｜>我去市场买了10个苹果。我给了邻居2个苹果和修理工2个苹果。然后我去买了5个苹果并吃了1个。我还剩下多少苹果？<｜Assistant｜><think>\n", "max_tokens": 800, "temperature": 0, "top_p": 1.0, "top_k": 1, "repetition_penalty":1.0}'