Add Windrose metrics exporter

[carl-shipstuff]

Summary

• add stdlib Prometheus exporter in metrics.py, runnable as a standalone sidecar or imported by server.py for opt-in /metrics
• wire opt-in Helm and Compose support for the metrics sidecar, ServiceMonitor, annotation-based Prometheus scraping, and packaged Grafana dashboard
• add tests and CI coverage for exporter output, HTTP handlers, dashboard JSON, Helm metrics render, and hostNetwork metrics/UI port conflict validation

Validation

• python3 -m py_compile server.py metrics.py scripts/patch-idle-cpu.py tests/test_metrics.py
• python3 tests/test_metrics.py
• python3 tests/test_retention.py
• python3 tests/test_restore.py
• python3 tests/test_auto_backup.py
• python3 tests/test_server_control.py
• python3 tests/test_schema.py
• python3 tests/test_http.py
• bash tests/test_install_env_merge.sh
• bash tests/test_engine_ini_reconcile.sh
• bash tests/test_shadow_stamp.sh
• docker compose config --quiet
• docker compose --profile metrics config --quiet
• helm lint ./helm/windrose
• Helm templates for default, managed, mutable, and metrics-enabled paths
• verified default Helm render has no metrics resources
• verified hostNetwork port conflict fails when metrics.port == service.port

Canary dogfood

• built/pushed split-persistence-compatible image registry.local:30500/games/windrose-server:metrics-split-5554b8d
• took pinned backup manual-20260505T020754Z
• live-patched canary only because its current StatefulSet has storage drift that this main-based chart cannot safely represent
• canary is 4/4 Running, zero restarts, with windrose-metrics on 28082
• direct scrape returns windrose_exporter_scrape_success 1 and build labels for Steam build 23065343
• Prometheus query windrose_exporter_scrape_success{job="windrose-canary"} returns one successful series
• Grafana imported dashboard Windrose Server at /d/windrose-server/windrose-server

Note: https://windrose-canary.shipstuff.fun/api/status timed out from this environment; current canary Helm values have ingress.enabled=false, so validation used in-cluster service/API.

[seslly]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c639463e7e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Source game limits from the game container, not sidecar

When .Values.resources.game.limits.* are unset, this sidecar gets empty WINDROSE_GAME_*_LIMIT values and metrics.py falls back to cgroup detection. In the windrose-metrics container, /sys/fs/cgroup/* reflects the sidecar’s own cgroup (for example its default 128Mi memory cap), not the game container’s, so windrose_cpu_limit_millicores / windrose_memory_limit_bytes can be wrong and dashboards/alerts will compute incorrect utilization. This is especially visible for installs that do not set explicit game CPU/memory limits in Helm values.

Useful? React with 👍 / 👎.