Grafana Platform Dashboard

Quick Start

1. Confirm dashboard target:

oc --context <ctx> get grafanadashboard -A | rg -i '<dashboard-name-or-theme>'

1. Export dashboard and JSON:

skills/grafana-platform-dashboard/scripts/grafanadashboard_roundtrip.sh export \
  --context <ctx> \
  --namespace <ns> \
  --name <grafanadashboard-name> \
  --out-dir /tmp/<workspace>

1. Edit the JSON and validate all PromQL:

skills/grafana-platform-dashboard/scripts/promql_scan_thanos.sh \
  --context <ctx> \
  --dashboard-json /tmp/<workspace>/<name>.json

1. Apply live safely:

skills/grafana-platform-dashboard/scripts/grafanadashboard_roundtrip.sh apply \
  --context <ctx> \
  --namespace <ns> \
  --name <grafanadashboard-name> \
  --json /tmp/<workspace>/<name>.json

Workflow

1) Lock Scope From Platform Contracts

1. Keep L1 command view constrained to critical pre-tenant-impact signals.
2. Use gate-aligned components first (critical CO gate, nodes, MCP, core API/etcd/ingress).
3. Keep service-specific sections (Crossplane, Keycloak) below L1.

2) Enforce Information Architecture

1. L1: critical-only, immediate action, minimal panel budget.
2. L2: platform services by dependency domain.
3. L3: deep dives (for example future GPU dashboard), not in L1.

3) Build Queries From Known Library

1. Start from known-good queries and adapt labels minimally.
2. Prefer counts and action tables over decorative charts.
3. Filter alert noise explicitly (for example ArgoCD/GitOps) when requested.

4) Validate Before Apply

skills/grafana-platform-dashboard/scripts/promql_scan_thanos.sh \
  --context <ctx> \
  --dashboard-json <file.json> \
  --output <scan.tsv>

success

5) Apply and Verify Sync

skills/grafana-platform-dashboard/scripts/grafanadashboard_roundtrip.sh apply ...
oc --context <ctx> -n <ns> get grafanadashboard <name> \
  -o jsonpath='{.status.conditions[?(@.type=="DashboardSynchronized")].status}{"|"}{.status.conditions[?(@.type=="DashboardSynchronized")].reason}{"\n"}'

6) Close With Operator-Focused Summary

1. What changed (panel names and intent).
2. Validation result (query count and failures).
3. Sync status and any residual risk.
4. Next step: promote live changes into GitOps-managed source.

Design Rules

1. Put critical tenant-impact predictors first.
2. Every red panel must imply an action path.
3. Avoid ambiguous panel names (for example replace “platform pods” with concrete namespace scope).
4. Keep L1 low-noise; move detail below or to dedicated dashboards.
5. Keep GPU deep diagnostics in a dedicated GPU dashboard, not mixed into L1.

References

1. Platform Contract
2. PromQL Panel Library
3. Layout Guidelines