refactor(agent): serve /metrics through the shared metrics-endpoint crate

[chet]

The agent stood up its own axum server for /metrics -- a spawn_prometheus_metrics_server wrapper plus a get_metrics_router/export_metrics pair that hand-encoded the registry. That is exactly what the shared metrics-endpoint crate already does for the rest of the fleet, so this hands the agent's existing Prometheus registry to metrics_endpoint::run_metrics_endpoint and deletes the bespoke server.

• The metrics listener now builds a MetricsEndpointConfig over get_prometheus_registry() and serves it, keeping the same bind address and the same fire-and-forget task.
• spawn_prometheus_metrics_server, get_metrics_router, and export_metrics are gone, along with their now-unused axum and encoder imports.
• The meter setup -- the InstrumentationSingleton and its custom network and retry views -- is untouched, so the exposed series are unchanged.

The endpoint serves the same registry through the same TextEncoder, so /metrics stays byte-identical; it also answers /health, /ready, and / on the same port, matching the other migrated services.

The previous server offloaded the encode to spawn_blocking; the shared endpoint encodes inline, as it does for every other consumer. That's the shared crate's established behavior, not a regression here -- moving the encode off the async worker for all consumers is tracked in #3524.

This supports #3180

[chet]

@coderabbitai full_review, thanks!

[coderabbitai]

@chet, you're welcome! I'll perform a full review of all changes in this PR.

(=^･ω･^=)

✅ Action performed

Full review finished.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: fa8f1254-5bb7-4f34-be9e-85a802680de0

📥 Commits

Reviewing files that changed from the base of the PR and between 8abff94 and 2fd77c7.

⛔ Files ignored due to path filters (1)

• Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (4)

• crates/agent/Cargo.toml
• crates/agent/src/instrumentation.rs
• crates/agent/src/main_loop.rs
• crates/agent/src/metadata_service.rs

💤 Files with no reviewable changes (1)

• crates/agent/src/instrumentation.rs

---

Summary by CodeRabbit

• New Features
  • Added request tracing for the agent’s HTTP services, capturing request counts and response latency.
  • Exposes Prometheus metrics via a dedicated /metrics endpoint using the configured telemetry metrics address.
  • Split responsibilities so the metadata service and metrics endpoint are handled independently.
• Bug Fixes
  • Detects invalid metrics endpoint addresses and reports a warning without blocking the agent from starting.

Walkthrough

The agent adds shared metrics-endpoint integration, removes its local Prometheus server implementation, and adds Axum tracing middleware that records HTTP request counts and latency.

Changes

Agent metrics and tracing

Layer / File(s)	Summary
HTTP tracing instrumentation crates/agent/src/instrumentation.rs	Adds WithTracingLayer for Axum routers, recording HTTP counters and latency while logging request and response details.
Shared metrics endpoint integration crates/agent/Cargo.toml, crates/agent/src/main_loop.rs, crates/agent/src/metadata_service.rs	Starts metrics_endpoint::run_metrics_endpoint with the shared Prometheus registry, skips startup for invalid addresses, and removes the metadata service’s local metrics server.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• NVIDIA/infra-controller#3474 — Introduces the shared metrics endpoint configuration used by this migration.
• NVIDIA/infra-controller#3523 — Makes a parallel change from local Prometheus routing to metrics_endpoint::run_metrics_endpoint.

Sequence Diagram(s)

sequenceDiagram
  participant AgentSetup
  participant PrometheusRegistry
  participant MetricsEndpoint
  AgentSetup->>PrometheusRegistry: get_prometheus_registry()
  AgentSetup->>MetricsEndpoint: run_metrics_endpoint(config, registry)
  MetricsEndpoint-->>AgentSetup: Serve metrics endpoint

Loading

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: moving the agent's /metrics serving to the shared metrics-endpoint crate.
Description check	✅ Passed	The description is clearly aligned with the change and explains the migration from bespoke metrics serving to the shared endpoint.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[github-actions]

🔍 Container Scan Summary

Service	Total	Critical	High	Medium	Low	Other
boot-artifacts-aarch64	3	0	0	3	0	0
boot-artifacts-x86_64	3	0	0	3	0	0
forge-admin-cli-x86_64	259	13	30	78	7	131
machine-validation-runner	807	40	234	288	36	209
machine_validation	807	40	234	288	36	209
machine_validation-aarch64	807	40	234	288	36	209
nvmetal-carbide	807	40	234	288	36	209
TOTAL	3493	173	966	1236	151	967

Per-CVE detail lives in the per-service grype-* artifacts (JSON + SARIF). Severity counts only — no CVE IDs published here.