Readiness probes for prefill and decode pods needed for accurate status

[namasl]

Component

Helm Chart

Desired use case or feature

When standing up decode and prefill pods with quickstart/llmd-installer.sh, they lack an appropriate readiness probe. Large models take multiple minutes beyond "ready" before they can actually accept requests.

Proposed solution

Add readiness probes that ensure vLLM is ready to accept requests (like httpGet probe to /health), along with probes for any sidecars.

Alternatives

No response

Additional context or screenshots

No response

[nerdalert]

Hi @namasl is this something you are willing to contribute? If so, I will assign, no pressure though, Ty!