Dashboards and notebooks

1. Designing #

One question per dashboard.

Structure #

Good practice: Organize by golden signals (latency, traffic, errors, saturation) at the top, technical details below, and $env and $service template variables so that a single dashboard serves all services.

2. Formulas and functions #

Computing in the widget.

Examples #

# Taux d'erreur en %
(sum:trace.http.request.errors{$env,$service}.as_count() / sum:trace.http.request.hits{$env,$service}.as_count()) * 100

# Comparaison avec la semaine précédente
week_before(avg:system.load.1{$env})

# Prévision et anomalies
forecast(avg:system.disk.in_use{$env} by {host}, 'linear', 1)
anomalies(sum:orders.created{$env}.as_count(), 'agile', 2)

3. Notebooks and sharing #

Investigating, documenting.

Uses #

• Notebooks: graphs, log queries and text mixed together, for a post-mortem or a reproducible investigation.
• Powerpacks: groups of widgets reusable across dashboards (for example the company's standard « golden signals » block).
• Sharing via a public link or embedding in a NOC screen, with fixed template variables.
• Dashboards in Terraform (datadog_dashboard_json) or exported as JSON and versioned.