#Everything red in a group

You want every pipeline that failed in a group and its subgroups since this morning, in one table, with the reason each one failed visible in the row. One node in the browse tree covers every project beneath it, so this is one wait rather than one wait per project.

▍recipe-failed-pipelines120×32

recipe-failed-pipelinesA recording of a group's failed pipelines being collected. Pressing b opens the browse tree; Enter on the platform group scopes the dashboard there and the context bar changes to platform. Tab moves to the pipelines view, which keeps the same scope. Pressing f opens the filter bar and status:failed with updated under 24h is typed. The table redraws with nine failed pipelines from nine different projects in the group tree, each row showing project, ref, status, duration, queued time, source and failure reason. Three of the rows read runner_system_failure or stuck_or_timeout_failure rather than script_failure. Pressing P pins the result as a tab named Failed today.

#Steps

Step	Key	What it does
1	b then Enter on platform	Scopes to the group and every subgroup beneath it
2	Tab to pipelines	Same scope, CI view
3	f then status:failed updated < 24h	Only what went red since this morning
4	P	Pins it as Failed today

Next time, press 2 instead of typing status:failed. Failed CI is one of the five quick filters.

#Reading the result

A group scope includes its subgroups. Selecting one node in Browse is what puts nine projects on screen from a group tree rather than from a list you maintain.

The failure reason column turns a list into a diagnosis. Three runner_system_failure and stuck_or_timeout_failure rows are a runner capacity problem, and they would be invisible if every row said only "failed".

The source column shows how much of the red is scheduled dependency bumps. On many teams that is most of it.

#Row actions

Key	Action	Key	Action
Enter	jobs by stage, plus the first failure inline	L	the failing job's log, scrolled to the error
R	retry	w	watch it, and get notified when it lands

Enter is usually enough. The failure summary puts the failing assertion in the preview without opening a log at all.

#Variations

#Human-triggered pipelines only

Add source:push,merge-request to the filter bar and the scheduled dependency bumps leave the table. On most teams this is the largest single change to the recipe.

#The branch that must stay green

Add ref:main, pin it as main is red, then set that tab's refresh interval to 30 seconds. Each pinned view polls on its own schedule.

#What is running right now

Filter to status:running,pending and pin it. Read the queued column: it separates CI that is slow from runners that are busy. The two feel identical and have different fixes.

#A group with hundreds of projects

labdash bounds the fan-out across a group's projects itself. You never set a limit, and a very wide group still returns. See group pipelines.