[TaylorMutch]

Summary

Use metrics.enabled to enable or disable the gateway's dedicated Prometheus endpoint independently of its port. Metrics remain enabled on port 9090 by default; disabling the endpoint retains the configured port and omits the runtime listener, container port, and Service port.

Related Issue

Closes #5474.

Changes

• Add boolean metrics.enabled, defaulting to true, and use it consistently across runtime and Kubernetes resources.
• Keep service.metricsPort between 1 and 65535; check listener collisions only while metrics are enabled.
• Reject legacy metricsPort: 0 and false values with instructions to set metrics.enabled=false and retain a valid port. Requesting metrics TLS while disabled also fails rendering.
• Cover default, custom, disabled, TLS, and migration configurations with chart tests and Rust-loader validation.
• Update the generated chart README, published gateway configuration docs, and related operating skills.

Testing

• mise run helm:test: 470 gateway and 7 workspace chart tests, generated gateway TOML validation through the Rust loader, and resource coherence checks
• Helm lint and template rendering across maintained CI/dev overlays
• Generated Helm README consistency
• Fern documentation and navigation validation
• Markdown lint, ShellCheck, and git diff whitespace checks
• Live Kubernetes deployment smoke test

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)
• User-facing configuration documentation and relevant skills updated

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pr-5478.docs.buildwithfern.com/openshell

[github-actions]

Compatibility review (advisory)

New or updated breaking changes; see the full report for all findings.

1. helm — No explicit risk acceptance found

service.metricsPort: 0 previously disabled metrics but now fails chart validation, even with metrics.enabled=false.

• Impact: Existing Helm values containing service.metricsPort: 0 block rendering and upgrades until operators set metrics.enabled=false and replace the zero port with a valid port.
• Action: Retain service.metricsPort: 0 as a backward-compatible disabled-metrics alias on this no-break train rather than requiring existing installations to migrate.
• PR: feat(helm): separate metrics enablement from listener port #5478

Review details

• Execution: complete
• Candidate: PR #5478 (786cb907fbfd415428dc885cf71603124de882c3)
• Baseline: 4ad3ce9db2700da6a3d32f8179f68062b7aaa6c3 (4ad3ce9db2700da6a3d32f8179f68062b7aaa6c3)
• Reviewed: rust-sdk, python-sdk, go-sdk, typescript-sdk, api-behavior, cli, configuration, policy, helm, state

Advisory review; publication gates are unchanged.

Branch check run

[TaylorMutch]

This fixes unclear behavior in the metrics port enablement so compatibility is not a concern.