Dashboard elements

Every chart type and the result shape it needs.

The element type is a rendering decision. The query decides what is possible.

Element	Needs	Good for
`timeseries`	A `bin()` time column plus a number	Rates and latencies over time
`statistic`	A single row, single number	The one number at the top left
`gauge`	One number, plus min and max	Utilisation against a known ceiling
`table`	Anything	Top-N lists
`heatmap`	Two dimensions plus a number	Latency distribution, error density by region
`pie`	One dimension plus a number	Share of total. Use sparingly, and never above six slices
`log-stream`	Raw events, no aggregation	Recent errors, in context
`note`	Nothing	Markdown: runbook links, what to do at 3am

A time series element

read top to bottom

1['api-gateway-prod']
2| summarize requests = count(), errors = countif(status >= 500)
3  by bin_auto(_time), service
4| order by _time asc

Use bin_auto in dashboards

A fixed bin(_time, 1m) draws 10,080 points when someone switches the range to 7 days, and the browser stops responding. bin_auto follows the range — see time and bins.

The note element is underrated

A markdown note on the dashboard, saying what the charts mean and what to do when they go red, is worth more than a ninth chart. It is the difference between a dashboard and a runbook.

PreviousDashboards NextTemplate variables