No logs in `*-model` pods after it starts up

[vanakema]

Environment

1. OS and Architecture: N/A
2. App or Package Name: vllm, text-embeddings
3. App or Package Version: v11
4. Kubernetes Distribution: N/A
5. Kubernetes Version: N/A

Steps to reproduce

1. Tail the text-embeddings pod logs
2. Add a document to an assistant so it triggers indexing on that document
3. Observe that there are no logs output to indicate that a request came in at all

Expected result

• When a request comes in, I can see some information about it in the pod logs

Actual Result

• There are no logs outputted after the pod has booted up

Additional Context

I believe I investigated this, and found that there are in fact logging statements, but they are not making it to the pod logs. I'm thinking maybe the logger is misconfigured or something.

[justinthelaw]

This only occurs in text-embeddings, correct? vLLM and llama-cpp-python should be logging metrics and params of each generation request.