fix(self-managed): configure LLM router worker address

[mikeyrcamp]

TL;DR

Backport the reviewed self-managed LLM request-router worker-address contract to the v0.7 release branch. Colocated deployments use the cluster-local service by default, while split deployments can provide a worker-reachable override.

Additional Details

Why

The v0.7 stack did not populate the LLM request-router worker address consumed by generated LLM workers. Compute agents can reject LLM launches without that address. The release needs a safe colocated default and a topology-owned override for split deployments.

What changed

• Added global.workerEndpoints.llmRequestRouterAddress as an optional topology override.
• Defaulted empty or omitted values to llm-request-router.nvcf.svc.cluster.local:50071.
• Derived the cluster-local default and request-router chart service from the same configurable gRPC port.
• Validated explicit DNS or IPv4 host:port and bracketed IPv6 host:port overrides, including port bounds.
• Mapped the effective address through api.remoteConfig.configData.nvcf.llm-request-router.worker-address.
• Kept disabled LLM deployments from receiving a staged worker address.
• Removed redundant local fixture values so local rendering and BDD wiring exercise the default path.
• Updated installation, enablement, local-development, and topology guidance.

Customer Release Notes

Self-managed v0.7 LLM deployments with colocated workers now use llm-request-router.nvcf.svc.cluster.local:50071 automatically. Split deployments must override global.workerEndpoints.llmRequestRouterAddress with a request-router host and port reachable from worker pods.

Plan Summary

Configuration-only backport. No Kubernetes resources, chart versions, release pins, or replica counts are changed.

Usage

Colocated deployments require no additional worker-address configuration. The cluster-local address follows addons.llm.requestRouter.service.grpcPort when a non-default service port is configured. For a split deployment, configure an override before enabling the LLM add-on:

global:
  workerEndpoints:
    llmRequestRouterAddress: router.example.com:50071

Testing

• make -C deploy/stacks/self-managed test
• go test -short ./... from tests/bdd
• ./tools/ci/check-docs
• Bash syntax validation
• git diff --check

Notes

Split-cluster router network exposure is tracked separately in #689.

The v0.7 branch retains its existing stack and chart version pins. No API chart bump is required because v0.7 already pins a chart with the nested remote-config path.

References

• Relates to Self-managed LLM deployments omit the request router worker address #676

Related Pull Requests

• Mainline fix: fix(self-managed): configure LLM router worker address #678

Dependencies

None. No third-party dependency, license, or NOTICE changes.

For the Reviewer

Please compare the default and explicit-override behavior with #678 and confirm the release branch retains its existing v0.7 version pins.

For QA

QA is recommended for one colocated and one split-cluster v0.7 LLM function launch. The focused render tests and non-live BDD suite are green.

Issues

Relates to #676

Checklist

• I am familiar with the Contributing Guidelines.
• I have signed off my commits for Developer Certificate of Origin compliance.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

🗂️ Base branches to auto review (1)

• main

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 0f85688e-b83b-4726-b833-8c17e0b18972

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

🌿 Preview your docs: https://nvidia-preview-agent-backport-v0-7-llm-router-worker-address.docs.buildwithfern.com/nvcf