wiremind-grafana-dashboards

Imported from wiremind/wiremind-grafana-dashboards (AGENTS.md). Install upstream with npx skills add wiremind/wiremind-grafana-dashboards. Copyright stays with the author.

Dashboard conventions

Grafana dashboard JSONs, one kebab-case file per dashboard at the repo root. Dashboards are provisioned onto the Wiremind clusters by kube-prometheus-stack (wiremind-services-configuration): the Grafana sidecar downloads the raw JSON from master, so merging here deploys everywhere the dashboard is referenced.

• Classic v1 schema only (schemaVersion + panels/gridPos). Grafana's file provisioner rejects the v2 resource format (elements/layout, apiVersion: dashboard.grafana.app/v2). When exporting from the Grafana UI, pick the Classic export option, and avoid v2-only features (tabs, conditional panels).
• Never hardcode a datasource: declare a datasource template variable (type: datasource, query: prometheus) and reference ${datasource} everywhere. Enable the cleanup hook: git config core.hooksPath .githooks.
• Multi-cluster by default: queries go through Thanos, so add a cluster query variable (label_values(<metric>, cluster), multi + include-all) and filter every query with cluster=~"$cluster". Add namespace (chained on $cluster) when it makes sense.
• Give each dashboard a stable uid and tags; keep the uid unchanged when updating, it is part of dashboard URLs.
• On stat/gauge/bar-gauge panels, set the field default displayName: ${__series.name} if the series name must stay visible when a query returns a single series (Grafana hides it by default).
• To provision a new dashboard, add a dashboards: entry pointing at the raw master URL in helmfile/services/monitoring/kube-prometheus-stack/values/base/kube-prometheus-stack.chart.yaml.gotmpl in wiremind-services-configuration.