feat(helm): package Grafana dashboards

docs/observability/core_metrics.md

@@ -123,6 +123,7 @@ This file contains a list of metrics exported by NVIDIA Infra Controller (NICo).
 <tr><td>carbide_site_explorer_created_power_shelves_count</td><td>gauge</td><td>The amount of Power Shelves that had been created by Site Explorer after being identified</td></tr>
 <tr><td>carbide_site_explorer_enabled</td><td>gauge</td><td>Whether site-explorer is enabled (1) or paused (0)</td></tr>
 <tr><td>carbide_site_explorer_iteration_latency_milliseconds</td><td>histogram</td><td>The time it took to perform one site explorer iteration</td></tr>
+<tr><td>carbide_site_explorer_last_run_status</td><td>gauge</td><td>The status of the latest Site Explorer run</td></tr>
 <tr><td>carbide_site_explorer_phase_latency_milliseconds</td><td>histogram</td><td>The time it took to perform one site explorer iteration phase</td></tr>
 <tr><td>carbide_site_explorer_update_explored_endpoints_count</td><td>gauge</td><td>Counts from the last update_explored_endpoints phase by kind</td></tr>
 <tr><td>carbide_switches_enqueuer_iteration_latency_milliseconds</td><td>histogram</td><td>The overall time it took to enqueue state handling tasks for all carbide_switches in the system</td></tr>

helm/PREREQUISITES.md

@@ -41,6 +41,21 @@ If you want Prometheus metrics collection, install the [Prometheus Operator](htt
 - **nico-hardware-health** also exposes an optional `telemetryServiceMonitor` (disabled by default) that scrapes `/telemetry` for per-machine sensor gauge data (temperature, power, fans, etc.) from the Prometheus sink. Use `serviceMonitor` for `/metrics` operational metrics only.
 - NICo functions normally without the Prometheus Operator installed.
 
+### Grafana Dashboard Sidecar (Optional)
+
+The umbrella chart can install its packaged Grafana dashboards as a ConfigMap
+when `grafanaDashboards.enabled=true`. This requires an existing Grafana
+installation with a dashboard sidecar watching the ConfigMap namespace and
+labels. The default `grafana_dashboard: "1"` label matches the common
+`kube-prometheus-stack` selector. To place dashboards in the configured NICo
+folder, the sidecar must read the `grafana_folder` annotation (or both sides
+must be configured with another annotation key).
+
+Grafana is not installed by the NICo chart. If `grafanaDashboards.namespace`
+targets a namespace other than the NICo release namespace, create that
+namespace first and configure the sidecar to search it. See
+[`README.md`](./README.md#grafana-dashboards) for values and namespace examples.
+
 ---
 
 ## 2. PostgreSQL Database

helm/README.md

@@ -74,6 +74,49 @@ Top-level `global:` values are automatically passed to all subcharts.
 | `global.spiffe.trustDomain` | SPIFFE trust domain for mTLS | `nico.local` |
 | `global.labels` | Common labels applied to all resources | See `values.yaml` |
 
+### Grafana Dashboards
+
+The chart packages three dashboards built from NICo's exported Prometheus
+metrics: a site overview, object lifecycle diagnostics, and API performance.
+They are disabled by default because this chart does not install Grafana.
+The source JSON files live in [`dashboards/`](./dashboards/) and can also be
+imported into Grafana directly.
+
+To expose the dashboards to a Grafana dashboard sidecar in the release
+namespace:
+
+```yaml
+grafanaDashboards:
+  enabled: true
+```
+
+The default `grafana_dashboard: "1"` label matches the dashboard-sidecar
+selector used by `kube-prometheus-stack`. The chart also adds the conventional
+`grafana_folder: NICo` annotation; configure the Grafana sidecar's
+`folderAnnotation` setting if it does not already read that key. If Grafana
+watches a different namespace or selector, configure them explicitly:
+
+```yaml
+grafanaDashboards:
+  enabled: true
+  namespace: monitoring
+  folder: Infrastructure/NICo
+  folderAnnotation: grafana_folder
+  labels:
+    grafana_dashboard: "1"
+  annotations: {}
+```
+
+The target namespace must exist before Helm runs, and the Helm identity must be
+allowed to create ConfigMaps there. The Grafana sidecar must also watch that
+namespace; for `kube-prometheus-stack`, configure
+`grafana.sidecar.dashboards.searchNamespace` accordingly.
+
+Each dashboard provides a Prometheus data-source selector, a NICo scrape-job
+selector, and an editable metric-prefix variable. The prefix defaults to
+`carbide`, which is the prefix currently emitted by NICo. Set it to `nico` (or
+another configured value) when using the `alt_metric_prefix` site setting.
+
 ### Subchart Enable/Disable Flags
 
 Each subchart can be independently enabled or disabled. All core NICo services are enabled by default. Infrastructure services (`unbound`) that may already be provided by the environment are disabled by default.