Dashboards
Dashboards
Composable charts over the same query language, with a shared time range.
A dashboard is a set of LSQL queries laid out on a grid, sharing one time range and one set of variables. Every element is a query you could have run in Explore.
Errors by service
1h
Building one
- 1
Start from a query
Get the result right in Explore, then Add to dashboard. Charts built from a working query beat charts built from a blank element.
- 2
Pick the element type
A time column plus a number gives a time series. Two dimensions give a table or a heatmap. One number gives a statistic.
- 3
Add variables
Anything you would otherwise hard-code — service, environment, region — becomes a variable.
- 4
Set the default range
Match it to the audience. An on-call board wants 1h; a weekly review wants 7d.
Dashboards people actually read
- One question per dashboard. "Is checkout healthy?" beats "everything about checkout".
- Put the answer at the top left. That is where eyes land, and it is where the on-call engineer stops reading.
- Six to eight elements. Past that, nobody scrolls.
- Annotate deploys — see annotations. Half of all dashboard questions are answered by a vertical line.