vllm - 💡(How to fix) Fix [Usage]: Missing vllm:num_requests_running and other metrics in vLLM v0.18 when deploying Qwen3.5 27B [1 participants]

TL;DR

The missing metrics in the /metrics endpoint may be due to a configuration or version issue with vLLM v0.18.

Guidance

• Verify that the vllm:num_requests_running metric is supported in vLLM v0.18 by checking the documentation or release notes.
• Check the configuration of the Qwen3.5 27B model deployment to ensure that metrics collection is enabled.
• Review the output of python collect_env.py to identify any potential environment or configuration issues that may be contributing to the missing metrics.
• Consider checking the vLLM documentation or community resources for known issues or workarounds related to metrics collection in v0.18.

Notes

The solution may depend on the specific configuration and deployment details of the Qwen3.5 27B model, which are not provided in the issue.

Recommendation

Apply workaround: Check the vLLM documentation and community resources for known issues or workarounds related to metrics collection in v0.18, as upgrading to a fixed version is not explicitly implied in the issue.